CN118159322A

CN118159322A - Medical device and ventilation state recognition method

Info

Publication number: CN118159322A
Application number: CN202180103623.XA
Authority: CN
Inventors: 李响; 刘京雷; 黄志文; 万聪颖; 周小勇
Original assignee: Shenzhen Mindray Bio Medical Electronics Co Ltd
Current assignee: Shenzhen Mindray Bio Medical Electronics Co Ltd
Priority date: 2021-12-09
Filing date: 2021-12-09
Publication date: 2024-06-07
Also published as: WO2023102820A1

Abstract

A medical device (100) and a ventilation status recognition method, the medical device (100) comprising: a memory (60) for storing executable program instructions; a processor (50) for executing program instructions stored in the memory (60), causing the processor (50) to perform the steps of: acquiring current respiration waveform data of a target subject (S910), wherein the respiration waveform data includes basic respiration waveform data and auxiliary waveform data; based on the trained recognition model, the breathing waveform data of the target object are recognized, and the recognition result of the current ventilation state of the target object is determined (S920), wherein the recognition result comprises normal ventilation or abnormal ventilation events, so that the accuracy of recognition of abnormal ventilation events in the aspect of man-machine synchronization can be improved, a doctor is assisted in judging man-machine countermeasure events, occurrence of man-machine countermeasure is further reduced, and the comfort and safety of breathing of a mechanically ventilated patient are improved.

Description

Medical device and ventilation state recognition method

Description

Technical Field

The invention relates to the technical field of medical equipment, in particular to medical equipment and a ventilation state identification method.

Background

In mechanical ventilation, when a patient breathes spontaneously, the inhalation and exhalation efforts of the patient are required to reach the set trigger sensitivity, and the ventilator can be switched to the corresponding inhalation phase or exhalation phase. For example, the inspiratory trigger may be set to a flow rate trigger that diverts the inspiratory phase when the flow rate exceeds a trigger sensitivity (e.g., 2L/min), or a pressure trigger that diverts the inspiratory phase when the airway pressure is below a positive end expiratory pressure (Positive End Expiratory Pressure, PEEP) -trigger sensitivity (e.g., 2cmH 2O). The exhalation trigger sensitivity is typically a percentage of the inhalation peak flow rate, such as when the inhalation flow rate decreases to 25% of the inhalation peak flow rate, the ventilator switches to the exhalation phase. Because the inspiration or expiration triggering sensitivity is set empirically by the doctor, there are situations in which the sensitivity setting is not compatible with the needs of the patient in clinic, resulting in an event of human-machine incompatibility. Such as inspiration trigger delay, expiration trigger, expiration switch advance or switch delay, etc. In addition, according to the difference of patients, the situations of accumulated water in the pipeline, overlarge resistance, overlarge compliance and the like can also occur during mechanical ventilation.

In view of the above, the present application provides a new medical device and ventilation status recognition method.

Disclosure of Invention

The present invention has been made in order to solve at least one of the above problems. In particular, an aspect of the present invention provides a medical device comprising:

a memory for storing executable program instructions;

A processor for executing the program instructions stored in the memory, causing the processor to perform the steps of:

Acquiring current respiration waveform data of a target object, wherein the respiration waveform data comprises basic respiration waveform data and auxiliary waveform data;

And carrying out recognition processing on the breathing waveform data of the target object based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object, wherein the recognition result comprises normal ventilation or ventilation abnormal events.

In yet another aspect, the present application provides a ventilation status identification method, the method comprising:

Still another aspect of the present application provides a medical device, the device comprising:

a memory for storing executable program instructions;

Acquiring current respiration waveform data and patient data of a target object, wherein the respiration waveform data comprises basic respiration waveform data and/or auxiliary waveform data;

And carrying out recognition processing on the breathing waveform data and the patient data based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object, wherein the recognition result comprises normal ventilation or ventilation abnormal events.

Optionally, the trained recognition model is based on training by a machine learning method.

Optionally, the trained recognition model comprises a trained neural network model, and the processor is further configured to: based on the patient data of the past patient and the historical data information of the respiratory waveform data of the past patient, training to obtain the trained recognition model by using a deep learning method, wherein the historical data information of the respiratory waveform data of the past patient comprises the historical data information of the basic respiratory waveform data and/or the historical data information of the auxiliary waveform data.

Optionally, the processor performs an identification process on the respiratory waveform data and the patient data based on a trained identification model, and determines an identification result of the current ventilation state of the target object, including: inputting respiratory waveform data of the target object to the trained neural network model to obtain a network output result; inputting the patient data of the target object into the trained neural network model, and mapping the patient data of the target object through transformation to obtain a parameter matrix; and in the trained neural network model, carrying out information fusion on the parameter matrix and the network output result to obtain the identification result.

Optionally, the processor performs an identification process on the respiratory waveform data and the patient data, and determines an identification result of the current ventilation state of the target object, including: preprocessing the respiration waveform data of the target object to obtain processed respiration waveform data, wherein the preprocessing comprises normalization processing and/or filtering processing; and carrying out recognition processing on the processed respiratory waveform data and the patient data of the target object based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object.

Optionally, the processor performs an identification process on the respiratory waveform data and the patient data based on a trained identification model, and determines an identification result of a current ventilation state of the target subject, including: judging whether the quality of the respiration waveform data meets a preset condition or not; when the quality of the breathing waveform data meets the preset condition, inputting the breathing waveform data and the patient data into the trained recognition model for recognition processing, and determining a recognition result of the current ventilation state of the target object; and when the quality of the breathing waveform data is determined not to meet the preset condition, not inputting the breathing waveform data and the patient data into the trained recognition model for recognition processing, and determining a recognition result of the current ventilation state of the target object.

Optionally, the processor performs an identification process on the respiratory waveform data and the patient data based on a trained identification model, and determines an identification result of the current ventilation state of the target subject, including: and after the recognition processing is carried out on the breathing waveform data of the target object and the patient data of the target object based on the trained recognition model, an expert system is utilized to carry out recognition processing on the breathing waveform data of the target object and the patient data, and a recognition result of the current ventilation state of the target object is determined.

Optionally, the processor performs obtaining patient data of the target object, including: acquiring electronic medical record information of the target object; acquiring at least part of patient data of the target object based on the electronic medical record information; or acquiring patient data of the target object input by a user through a man-machine interaction interface.

Optionally, the patient data includes at least one of the following information: patient medical record information, mechanical ventilation parameters, monitoring parameters, blood gas value parameters, ultrasound parameters, disease severity scores, fluid balance information, wherein the patient medical record information comprises at least one of the following information: age, height, weight, BMI index, co-morbid information, genetic disease information, medication information, the mechanical ventilation parameters including at least one of the following: tidal volume, minute ventilation, respiratory rate, inspiratory time, PEEP, mean airway pressure, inhaled oxygen concentration, pressure support level, maximum inspiratory pressure, maximum expiratory pressure, end tidal carbon dioxide, lung compliance, airway resistance, mechanical energy, work of breathing, driving pressure, peak pressure, plateau pressure, the monitored parameters comprising at least one of the following parameters: blood oxygen, heart rate, blood pressure, mean arterial pressure, central venous pressure, and temperature; the blood gas value parameter includes at least one of the following information: arterial blood carbon dioxide partial pressure, arterial blood oxygen partial pressure, arterial blood gas pH; the disease severity score includes at least one of the following scores: acute physiology and chronic health assessment IV, glasgang coma score, oxford acute disease severity score, sequential organ failure assessment, simplified acute physiology score II, acute physiology score, logical organ dysfunction score; the liquid balance information includes at least one of the following: urine volume, crystal volume.

Optionally, the processor is configured to: selecting a corresponding identification mode according to an instruction input by a user; when the recognition mode is a first recognition mode, carrying out recognition processing on the breathing waveform data of the target object and the patient data of the target object based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object;

When the identification mode is a second identification mode, an expert system is utilized to carry out identification processing on the breathing waveform data of the target object and the patient data of the target object, and an identification result of the current ventilation state of the target object is determined; and when the recognition mode is a third recognition mode, the trained recognition model and the expert system are utilized to perform recognition processing on the breathing waveform data of the target object and the patient data of the target object, and a recognition result of the current ventilation state of the target object is determined.

Optionally, the processor is further configured to: and outputting prompt information when the ventilation abnormal event is determined to exist.

Optionally, the medical device further comprises a display, and the display is used for acquiring and displaying the prompt information.

Optionally, the display is configured to display the prompt information in a preset display manner, where the preset display manner includes one or more of the following manners: highlighting, additional symbology, distinctive color, distinctive shading, flashing.

Optionally, the medical device includes a display, a key is disposed on a display interface of the display, and the processor is configured to: and when a user instruction input by the user through the key is acquired, controlling a display to display the prompt information.

Optionally, the medical equipment further comprises an alarm device, wherein the alarm device is used for giving an alarm prompt when the frequency or the frequency of ventilation abnormal events occurring in the preset time meets preset conditions.

Optionally, the processor is further configured to: and when the identification result is a ventilation abnormal event, automatically adjusting the setting parameters of the breathing machine for ventilating the target object according to the identification result.

Optionally, the processor is further configured to: and when the identification result is a ventilation abnormal event, outputting adjustment suggestion information of the setting parameters of the breathing machine for ventilating the target object according to the identification result.

Optionally, the medical device includes a display for acquiring and displaying the adjustment suggestion information.

Optionally, the processor is further configured to: when the identification result is a ventilation abnormal event, evaluating the lung injury degree of the target object according to the identification result and the patient data of the target object; and the medical device further comprises a display for displaying the result of the evaluation of the degree of lung injury.

Optionally, the processor evaluates the lung injury degree of the target object according to the identification result and the patient data of the target object, including: and assessing the degree of lung injury of the target subject based on the identification result and the tidal volume in the patient data of the target subject.

Optionally, the base respiratory waveform data includes one or more of the following waveform data: pressure, flow rate, or volumetric waveform; the auxiliary waveform data includes one or more of the following waveforms: trans-pulmonary pressure, esophageal pressure, intragastric pressure, or diaphragmatic electricity.

Optionally, the ventilation abnormality event comprises one or more of the following events: ineffective triggering, double inspiration, false triggering, too low flow rate, reverse triggering, early switching, delayed switching, slow pressure rise, too fast pressure rise, endogenous positive end expiratory pressure, short inspiration time, long inspiration time, pipeline water accumulation, too high resistance, too low compliance and inversely proportional ventilation.

Optionally, the medical device is a ventilator, anesthesia machine, monitor or central station.

Optionally, when the medical device is a monitor, the monitor includes a communication interface for communication connection with a ventilator that ventilates the target subject.

In another aspect, the present application further provides a ventilation status identifying method, which includes:

Optionally, the trained recognition model includes a trained neural network model, the trained recognition model being obtained by training using a deep learning method based on patient data of a past patient and historical data information of respiratory waveform data of the past patient, the historical data information of respiratory waveform data of the past patient including historical data information of basic respiratory waveform data and/or historical data information of auxiliary waveform data.

Optionally, the identifying the respiratory waveform data of the target object and the patient data of the target object based on the trained identifying model, determining the identifying result of the current ventilation state of the target object includes: judging whether the quality of the respiration waveform data meets a preset condition or not; when the quality of the respiration waveform data of the target object is determined to meet the preset condition, inputting the respiration waveform data of the target object and the patient data of the target object into the trained recognition model for recognition processing, and determining a recognition result of the current ventilation state of the target object; when the respiration waveform data of the target object is determined not to meet the preset condition, the respiration waveform data and the patient data are not input into the trained recognition model to be recognized, and a recognition result of the current ventilation state of the target object is determined.

Optionally, the method further comprises: and outputting prompt information when the ventilation abnormal event is determined to exist.

Optionally, the method further comprises:

When the identification result is a ventilation abnormal event, evaluating the lung injury degree of the target object according to the identification result and the patient data of the target object;

And displaying the evaluation result of the lung injury degree through a display.

Optionally, the evaluating the lung injury degree of the target object according to the identification result and the patient data of the target object includes:

and assessing the degree of lung injury of the target subject based on the identification result and the tidal volume in the patient data of the target subject.

According to the medical equipment and the ventilation state identification method, the basic waveform and the auxiliary waveform data of the target object can be identified based on the trained identification model, and the identification result of the current ventilation state of the target object is determined, so that the accuracy of ventilation abnormal event identification in the aspect of man-machine synchronization can be improved, a doctor is assisted in judging man-machine countermeasure events, occurrence of man-machine countermeasure is reduced, and the comfort and safety of breathing of a mechanically ventilated patient are improved.

In a second aspect, the medical device according to the present application is capable of performing recognition processing on respiratory waveform data and patient data of a target object based on a trained recognition model, and determining a recognition result of a current ventilation state of the target object, so that accuracy in recognition of ventilation abnormal events in man-machine synchronization can be improved, a doctor can be assisted in judging man-machine countermeasure events, occurrence of man-machine countermeasure is reduced, and comfort and safety of breathing of a mechanically ventilated patient are improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort to a person skilled in the art.

FIG. 1 shows a schematic block diagram of a medical device in one embodiment of the invention;

FIG. 2 shows a schematic block diagram of a ventilator in one embodiment of the invention;

FIG. 3 shows a schematic flow diagram of a processor in one embodiment of the invention as it recognizes via a neural network model;

FIG. 4 shows a schematic flow chart of the recognition of a processor by a neural network model and expert system in one embodiment of the invention;

FIG. 5 shows a schematic flow chart of the processor in one embodiment of the invention as it recognizes by a neural network model and expert system in combination with patient data;

FIG. 6 shows a schematic flow chart diagram when a processor identifies by multiple neural network models in one embodiment of the invention;

FIG. 7 shows a schematic flow chart diagram of a processor in one embodiment of the invention as it recognizes by multiple neural network models and expert systems;

FIG. 8 shows a schematic flow chart of the processor in one embodiment of the invention as it recognizes through multiple neural network models and expert systems in combination with patient data;

FIG. 9 illustrates a flow chart of a ventilation abnormality identification method in one embodiment of the invention;

FIG. 10 shows a schematic flow chart of a processor in another embodiment of the invention as it recognizes by a neural network model;

FIG. 11 shows a schematic flow chart of a processor in yet another embodiment of the invention as it recognizes by a neural network model;

FIG. 12 shows a schematic flow chart of the recognition by the neural network model and expert system by the processor in one embodiment of the invention;

FIG. 13 shows a schematic flow chart of the processor in one embodiment of the invention as it recognizes through a neural network model and expert system in combination with patient data;

fig. 14 shows a flowchart of a ventilation abnormality identification method in another embodiment of the invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some embodiments of the present invention and not all embodiments of the present invention, and it should be understood that the present invention is not limited by the example embodiments described herein. Based on the embodiments of the invention described in the present application, all other embodiments that a person skilled in the art would have without inventive effort shall fall within the scope of the invention.

In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the present invention. It will be apparent, however, to one skilled in the art that the invention may be practiced without one or more of these details. In other instances, well-known features have not been described in detail in order to avoid obscuring the invention.

It should be understood that the present invention may be embodied in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. As used herein, the term "and/or" includes any and all combinations of the associated listed items.

In view of the foregoing technical problems, the present application provides a medical device, including: a memory for storing executable program instructions; a processor for executing program instructions stored in the memory, causing the processor to perform the steps of: acquiring current respiration waveform data of a target object, wherein the respiration waveform data comprises basic respiration waveform data and auxiliary waveform data; and carrying out recognition processing on the breathing waveform data of the target object based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object, wherein the recognition result comprises normal ventilation or ventilation abnormal events.

According to the medical equipment provided by the application, the basic waveform and auxiliary waveform data of the target object can be identified based on the trained identification model, and the identification result of the current ventilation state of the target object is determined, so that the accuracy of ventilation abnormal event identification in the aspect of man-machine synchronization can be improved, a doctor is assisted in judging man-machine countermeasure events, the occurrence of man-machine countermeasure is reduced, and the comfort and safety of breathing of a mechanically ventilated patient are improved.

In order to provide a thorough understanding of the present application, detailed structures will be presented in the following description in order to illustrate the technical solutions presented by the present application. Alternative embodiments of the application are described in detail below, however, the application may have other implementations in addition to these detailed descriptions. Specifically, the medical device and the ventilation status recognition method according to the present application will be described in detail with reference to the accompanying drawings. The features of the examples and embodiments described below may be combined with each other without conflict.

By way of example, the medical device of the present application may be a ventilator, anesthesia machine, monitor or central station, or the like. When the medical device is a monitor or a central station, the medical device comprises a communication interface, and the communication interface can be in communication connection with a breathing machine, so that various data output by the breathing machine or the anesthesia machine, such as breathing waveform data, various set parameter information and the like, can be obtained, and in the application, the breathing waveform data refers to waveform data related to breathing. In one example, when the medical device is a central station, the central station may directly acquire various data output from the ventilator or the anesthesia machine through a communication interface provided therein, or the central station may be further connected to a monitor, and the monitor is connected to the ventilator or the anesthesia machine so as to acquire various data output therefrom, and transmit the data to the central station. In some examples, the ventilator or anesthesia machine may further include a communication interface through which the monitor or central station is communicatively connected to obtain various monitored parameters of the patient for which it is ventilated, such as blood oxygen (SpO 2), heart Rate (HR), blood Pressure (BP), mean Arterial Pressure (MAP), central Venous Pressure (CVP), temperature (T), etc., from the monitor or central station.

In one embodiment, as shown in FIG. 1, the medical device 100 of the present application includes one or more processors 50, a display 70, a memory 60, and a communication interface, among others. These components are interconnected by a bus system and/or other forms of connection mechanisms (not shown). It should be noted that the components and structures of the medical device 100 shown in fig. 1 are exemplary only and not limiting, and that the medical device 100 may have other components and structures as desired.

The memory 60 is used to store various data and executable programs that are generated during use of the associated medical device, such as system programs for storing medical devices, various application programs, or algorithms for performing various specific functions. May include one or more computer program products that may include various forms of computer-readable storage media, such as volatile memory and/or non-volatile memory. Volatile memory can include, for example, random Access Memory (RAM) and/or cache memory (cache) and the like. The non-volatile memory may include, for example, read Only Memory (ROM), hard disk, flash memory, and the like. During use of the medical device, data stored locally, if desired, may be stored in memory 60.

The processor 50 may be a Central Processing Unit (CPU), an image processing unit (GPU), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), or other form of processing unit with data processing and/or instruction execution capabilities, and may control other components in the medical device to perform desired functions. For example, the processor 50 can include one or more embedded processors, processor cores, microprocessors, logic circuits, hardware Finite State Machines (FSMs), digital Signal Processors (DSPs), image processing units (GPUs), or combinations thereof.

The medical device may further include a man-machine interaction means, which may include a display 70 for displaying breathing waveform data when the ventilator ventilates the patient, displaying state information of the patient, recognition results of ventilation states, various prompt information or alarm information, etc., and displaying specific contents may include characters, diagrams, numbers, colors, waveforms, characters, etc., for intuitively displaying various kinds of information. In practical application, the man-machine interaction device may further include an input device, through which a medical staff may set various parameters, select and control a display interface of the display 70, and so on, so as to implement information interaction between man-machines. The display 70 may also be a touch display. The man-machine interface may be referred to herein as a display interface of the display 70.

In some embodiments, the medical device may be a ventilator, which is an artificial mechanical ventilator to assist or control the respiratory motion of the patient to effect gas exchange within the lungs, reducing work done by the patient's breathing to facilitate recovery of respiratory function. Referring to fig. 2, in some embodiments the ventilator may further include a respiratory interface 211 (i.e., patient interface), a source interface 212, a respiratory circuit (i.e., respiratory circuit), a respiratory assistance device (i.e., an airflow providing device), a ventilation detection device for detecting ventilation parameters such as respiratory waveform data, a processor 50, a memory 60, and a display 70, among others.

Ventilation detection devices (not shown) are provided on the respiratory line or patient interface for detecting various respiratory waveform data, etc., which may include flow rate of ventilation, airway pressure, respiratory rate, tidal volume, inspiration time, respiratory system or pulmonary compliance, etc. It should be noted that, the detection of the respiration waveform data may be obtained by direct detection, or may be obtained by detecting certain basic parameters and then calculating.

The breathing circuit selectively communicates the air supply interface 212 with the patient's respiratory system. In some embodiments the breathing circuit includes an exhalation limb 213a and an inhalation limb 213b, the exhalation limb 213a being connected between the breathing interface 211 and the exhaust port 213c for directing exhaled gases from the patient to the exhaust port 213c. The exhaust port 213c may be opened to the outside environment or may be provided in a gas recovery device dedicated to the passage. The gas source interface 212 is used to connect to a gas source (not shown) for providing a gas, which may typically be oxygen, air, or the like; in some embodiments, the air source can adopt a compressed air bottle or a central air supply source, and the air source can supply air to the breathing machine through an air source interface 212, wherein the air source interface 212 can comprise conventional components such as a pressure gauge, a pressure regulator, a flowmeter, a pressure reducing valve, an air-oxygen ratio regulating and protecting device and the like, and the conventional components are respectively used for controlling the flow of various gases (such as oxygen and air). An inhalation branch 213b is connected between the respiratory interface 211 and the source interface 212 for providing oxygen or air to the patient, e.g. gas fed from the source interface 212 enters the inhalation branch 213b and then through the respiratory interface 211 into the lungs of the patient. The breathing interface 211 is used to connect the patient to a breathing circuit, and in addition to the gases delivered by the inspiratory limb 213b being directed to the patient, gases exhaled by the patient may be directed to the exhaust port 213c via the expiratory limb 213 a; the respiratory interface 211 may be a nasal cannula or a mask for wearing over the mouth and nose, or the respiratory interface 211 may also be a nasal mask, nasal cannula, or tracheal tube, as appropriate. The breathing assistance device is connected with the air source interface 212 and the breathing circuit and controls the air provided by the external air source to be delivered to a patient through the breathing circuit; in some embodiments the breathing assistance apparatus may comprise an exhalation controller 214a and an inhalation controller 214b, the exhalation controller 214a being arranged on the exhalation limb 213a for switching on the exhalation limb 213a or switching off the exhalation limb 213a, or controlling the flow rate or pressure of the patient's exhaled gas, depending on the control instructions. In particular implementations, the exhalation controller 214a may include one or more of an exhalation valve, a one-way valve, a flow controller, a PEEP valve, etc. that enable flow or pressure control. An inhalation controller 214b is provided on the inhalation branch 213b for switching on the inhalation branch 213b or switching off the inhalation branch 213b or controlling the flow rate or pressure of the output gas according to the control instruction. In particular implementations, the inspiration controller 214b may include one or more of an exhalation valve, a one-way valve, or a flow controller, among other devices capable of effecting flow or pressure control.

In some embodiments, the processor 50 may also be configured to execute instructions or programs to control various control valves in the breathing assistance device, the air supply interface 212, and/or the breathing circuit, or to process received data to generate desired calculations or determinations, or to generate visual data or graphics and output the visual data or graphics to the display 70 for display.

Wherein the processor 50 of the medical device is configured to execute the program instructions stored in the memory, such that the processor 50 performs the steps of: acquiring current respiration waveform data of a target object, wherein the respiration waveform data comprises basic respiration waveform data and auxiliary waveform data; and carrying out recognition processing on the breathing waveform data of the target object based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object, wherein the recognition result comprises normal ventilation or ventilation abnormal events.

It should be noted that, in the embodiment of the present application, respiration waveform data refers to data related to respiration, which may be waveforms or data capable of characterizing waveforms, and the like.

In one example, the base respiratory waveform data includes one or more of the following waveform data: pressure, flow rate or volume. The auxiliary waveform data includes one or more of the following waveforms: trans-pulmonary pressure, esophageal pressure, intragastric pressure, or diaphragmatic electricity, or other waveform data that is capable of reflecting the respiratory state of a patient. The above-described respiration waveform data may be obtained based on detection of a ventilation detecting means of a ventilator for detecting a ventilation parameter such as respiration waveform data, for example, based on acquisition of diaphragm electricity by, for example, an myoelectricity acquisition sensor, etc., and for example, esophageal pressure may be measured by inserting a catheter for measuring esophageal pressure into the esophagus of a patient, and for gastric pressure may be measured by inserting a catheter for measuring gastric pressure into the stomach of a patient.

In one example, one or more of the above respiratory waveform data may also be obtained based on one or more data calculations, e.g., the cross-lung pressure is the difference between intra-alveolar pressure and intra-thoracic pressure, where intra-alveolar pressure may be characterized by measuring airway pressure, measuring esophageal pressure to estimate intra-thoracic pressure, and further obtaining cross-lung pressure based on airway pressure and intra-thoracic pressure calculations, and, for example, volume may be obtained by flow rate integration. For another example, new auxiliary waveform data may also be calculated based on one or more of the respiratory waveform data (e.g., the base waveform data and the auxiliary waveform data), i.e., other auxiliary waveform data that is responsive to the respiratory state of the patient in addition to the enumerated trans-pulmonary pressure, esophageal pressure, intragastric pressure, or diaphragmatic electricity, such as combining waveforms of pressure, flow rate, volume, and esophageal pressure to derive new auxiliary waveform data.

The trained recognition model may be trained by machine learning methods including, but not limited to, methods based on decision trees, random forests, bayesian learning, deep learning, and the like.

For example, the trained recognition model may be obtained by training a neural network model using deep learning, i.e., the trained recognition model includes a trained neural network model, optionally including but not limited to a CNN multi-layer convolutional neural network, a RNN convolutional neural network, or a cnn+rnn network, wherein the trained recognition model is obtained by training with a deep learning method based on historical data information of respiratory waveform data of a past patient, the historical data information of respiratory waveform data of the past patient including historical data information of basic respiratory waveform data and historical data information of auxiliary waveform data.

Taking the recognition model as a neural network model as an example, the training process of the neural network model can include the following steps: firstly, a waveform database is established, wherein samples in the database comprise historical data information of respiratory waveform data of a past patient, such as historical data information comprising basic respiratory waveform data, wherein the historical data information of the basic respiratory waveform data comprises historical data information of pressure, historical data information of flow rate and historical data information of volume information, and further comprises different combinations of the historical data information of the basic respiratory waveform data and historical data information of auxiliary waveform information such as esophageal pressure, intragastric pressure or diaphragmatic muscle electricity. The database may contain training sets, test sets. Optionally, the database further includes a validation set (also called tuning set), and the training set adjusts the weight occupied by each neuron in order to train the parameters of the neural network; the verification set (also called tuning set) is used for verifying whether the model is fitted or not, so as to adjust super parameters (such as the number of network layers, the number of network neurons of each layer, the learning rate and the like); test set in order to finally evaluate the performance of the model, an evaluation is given to the generalization ability of the trained neural network model. The training set, the verification set (tuning set) and the test set form a breathing waveform together, and are mutually exclusive and uniformly distributed.

In one example, the duration of the respiratory waveform data of each sample in the database may include at least one or several complete respiratory cycles, where the database includes both normal waveforms and ventilation abnormal waveforms, and the proportion of ventilation abnormal waveforms in the total waveform data is within a preset threshold range, where the probability of occurrence of ventilation abnormal waveforms in actual situations may be simulated by setting the proportion of the number of ventilation abnormal waveforms, so as to ensure that the sample distribution conforms to clinical actual situations, so that a trained recognition model, such as a neural network model, can still give reasonable predictions and judgments when the information is incomplete. In the database, each sample has its own tag, whether it is a normal waveform or a ventilation abnormal waveform. These labels may be given by a number of clinical professionals.

For possible missing problems of basic waveform data such as pressure, flow rate and volumetric waveform input into a trained recognition model in practical application, in one example, before training a neural network model by using a deep learning method to obtain the trained recognition model, first data processing is performed on historical data information of respiratory waveform data to obtain abnormal waveforms with waveform missing so as to obtain more abnormal waveforms, and then training the neural network model by using a deep learning method to obtain the trained recognition model, the abnormal waveforms with waveform missing are put into a database so as to expand samples in the database, and thus the adaptability of the neural network model to waveform missing is enhanced. Optionally, the first data processing method includes, but is not limited to: data occlusion, noise addition, partial waveform removal, any type of waveform removal, and the like.

In order to make the recognition result of the trained recognition model better and more accurate, during the training process, the breathing waveform data of the samples in the database may be preprocessed to improve the recognition accuracy of the neural network model, for example, the preprocessing method includes, but is not limited to, data normalization, for example, pressure, flow rate, volume and esophageal pressure, and the data normalization process may be performed based on the following formulas:

Wherein t represents time, P ₀(t),F ₀(t),V ₀(t),Pes ₀ (t) is pressure, flow rate, volume and esophageal pressure at time t, P (t), F (t), V (t), pes (t) is normalized pressure, flow rate, volume and esophageal pressure, The average value can be adopted, a fixed value can be given empirically, and P _std,F _std,V _std,Pes _std is the standard deviation. Other auxiliary waveform information is normalized, for example, in substantially the same way as the esophageal pressure.

After preprocessing, the neural network model may be trained by a deep learning algorithm to complete the mapping from respiratory waveform data (as well as baseline waveform data and auxiliary waveform data) to ventilation abnormality types.

The process of the processor 50 performing the recognition processing on the breathing waveform data of the target subject based on the trained recognition model, and determining the recognition result of the current ventilation state of the target subject will be described below.

In one example, the processor 50 performs an identification process on the respiratory waveform data based on the trained identification model, determining an identification of the current ventilation status of the target subject, including: preprocessing respiratory waveform data of a target object to obtain processed respiratory waveform data; and carrying out recognition processing on the processed respiratory waveform data based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object.

Optionally, the preprocessing method includes, but is not limited to, one or more of data normalization and filtering, wherein when the data normalization process is performed, the process is substantially the same as the method of data normalization process in the training process described in the foregoing, and a description thereof will not be repeated.

In one example, the processor 50 performs an identification process on the respiration waveform data of the target subject based on the trained identification model, determining an identification of the current ventilation status of the target subject, including: judging whether the quality of the auxiliary waveform data and the basic waveform data meets a preset condition or not; when the quality of the auxiliary waveform data and the basic waveform data meets the preset conditions, inputting the auxiliary waveform data and the basic respiration waveform data meeting the requirements into a trained recognition model for recognition processing, and determining a recognition result of the current ventilation state of the target object; when the quality of the auxiliary waveform data and the basic waveform data is determined not to meet the preset condition, the auxiliary waveform data and the basic respiratory waveform data are not input into a trained recognition model for recognition processing. The accuracy of the identification result can be ensured through quality judgment, and the problem that an error result is output or no result is output due to unqualified quality of the breathing waveform data is avoided.

The processor 50 may determine whether the quality of the auxiliary waveform data and the basic waveform satisfies a preset condition by a time domain method, a frequency domain method, a waveform template, or the like. The time domain method is to judge through preset characteristics of waveforms in the time domain, such as Maximum Inspiratory Pressure (MIP), mean airway pressure, slope or time of pressure rise, inspiratory time, maximum flow rate, mean flow rate, tidal volume (Vt) and other preset characteristic parameter indexes, wherein each preset parameter index corresponds to a normal range, waveform quality can be judged through the parameters, if the waveform quality is good in a threshold range, otherwise, the waveform is judged to be unqualified; for the frequency domain method: converting the waveform into a frequency domain range for spectrum analysis to obtain the duty ratio of different frequencies, comparing the duty ratio with the frequency spectrum of a normal waveform, judging through a threshold value, for example, when the duty ratio of a high-frequency part reaches a preset threshold value, considering that the quality of the waveform is unqualified due to excessive noise, and when the duty ratio of the high-frequency part is lower than the preset threshold value, judging that the quality of the waveform is qualified, or for some waveforms, when the duty ratio of the low-frequency part reaches the preset threshold value, judging that the quality of the waveform is unqualified due to excessive noise, and when the duty ratio of the low-frequency part is lower than the preset threshold value, judging that the quality of the waveform is qualified. For the waveform template analysis method: the waveform template can be established according to the normal waveform (for example, the pressure template is established according to a plurality of normal pressure waveforms), breathing waveform data which is input into the trained recognition model in a preset mode is compared with the waveform template corresponding to the breathing waveform data, the difference degree of the whole period is calculated, the quality is judged through the threshold value, for example, when the difference degree is larger than a preset threshold value, the quality of the breathing waveform data is judged to be unqualified, and when the difference degree is smaller than or equal to the preset threshold value, the quality of the breathing waveform data is judged to be qualified.

It should be noted that, when there are a plurality of basic waveform data and a plurality of auxiliary waveform data, and the quality of a part of the basic waveform data and a part of the auxiliary waveform data in the plurality of basic waveform data satisfy the preset condition, the basic waveform data and the auxiliary waveform data satisfying the preset condition may also be input to the trained recognition model for recognition.

In one example, as shown in fig. 3, the number of trained recognition models may be 1, for example, the trained recognition model is a trained neural network model, and the processor 50 is further configured to: performing fusion processing on the basic respiration waveform data of the target object and/or the auxiliary waveform data of the target object to obtain fusion data; and inputting the fusion data into a trained recognition model for recognition processing, and determining a recognition result of the current ventilation state of the target object. For example, when the quality of the basic respiration waveform data and the auxiliary waveform data is acceptable, the basic respiration waveform data and/or the auxiliary waveform data of different types and numbers may be combined into a plurality of types of combined waveform data, for example, a combination of at least two types of waveform data in the basic respiration waveform data, or a combination of at least one type of waveform data in the basic respiration waveform data and at least one type of waveform data in the auxiliary waveform data, or a combination of at least two types of waveform data in the auxiliary waveform data. The fusion process refers to fusing each type of combined waveform data (at least two types of waveform data are included in the combined waveform data), for example, size alignment, etc., and the purpose of the fusion process is to determine the format that is finally input to the trained recognition model.

Optionally, the recognition result of the current ventilation status includes, but is not limited to, normal ventilation or ventilation abnormality, wherein the ventilation abnormality includes one or more of the following events: ineffective triggering, double inhalation (specifically, double inhalation may include double triggering and breath overlapping), false triggering, too low a flow rate, reverse triggering, switching advance, switching delay, slow pressure rise, too fast pressure rise, endogenous positive end expiratory pressure, short inhalation time, long inhalation time, line water accumulation, too high resistance, too low compliance, inversely proportional ventilation, etc., or may also include other ventilation anomalies that can be identified.

In one example, the processor 50 performs an identification process on the respiration waveform data of the target subject based on the trained identification model, determining an identification of the current ventilation status of the target subject, including: and carrying out recognition processing on the breathing waveform data of the target object based on the trained recognition model, and determining the recognition result of the current ventilation state of the target object and the reliability (also called confidence) corresponding to the recognition result. Through the credibility, an evaluation of the occurrence probability of the identification result can be given to the user, so that the user is assisted in judging whether the identification result is credible or not.

In one example, the number of trained recognition models may also be multiple, for example, when the trained recognition models include multiple trained neural network models, as shown in fig. 6, the respiratory waveform data includes at least one of a plurality of combined waveform data, each trained neural network model corresponding to one of the combined waveform data, optionally one or more of the following combinations: the combination of the at least one type of basal respiration waveform data, the at least one type of auxiliary waveform data, and the at least one type of basal respiration waveform data, for example, the basal respiration waveform data includes one or more of the following waveform data: pressure, flow rate, or volume, the auxiliary waveform data includes one or more of the following waveforms: the cross-lung pressure, esophageal pressure, intragastric pressure or diaphragmatic electricity, the at least one type of basic respiration waveform data may be pressure, flow rate or solvent, may further include any two types of combinations of pressure, flow rate and solvent and combinations of three types thereof, and the at least one type of auxiliary waveform data and the at least one type of basic respiration waveform data may further include the following combinations: a combination of one type of auxiliary waveform data and one type of basic respiratory waveform data (e.g., pressure and esophageal pressure, flow rate and intragastric pressure, etc.), a combination of two types of auxiliary waveform data and one type of basic respiratory waveform data (e.g., esophageal pressure, intragastric pressure and pressure), a combination of two types of auxiliary waveform data and two types of basic respiratory waveform data (e.g., esophageal pressure, intragastric pressure, volume and flow rate), a combination of three types of auxiliary waveform data and three types of basic respiratory waveform data (e.g., esophageal pressure, trans-pulmonary pressure, intragastric pressure, volume and flow rate), a combination of four types of auxiliary waveform data and three types of basic respiratory waveform data (e.g., diaphragmatic power, esophageal pressure, trans-pulmonary pressure, intragastric pressure, volume and flow rate), and the like.

As shown in fig. 6, the processor 50 is further configured to: and obtaining at least one kind of combined waveform data based on the obtained basic respiratory waveform data and auxiliary waveform data, respectively inputting various kinds of combined waveform data into the trained neural network model corresponding to the combined waveform data for recognition so as to respectively obtain a group of recognition results through recognition, wherein the recognition results can comprise normal ventilation or abnormal ventilation events and correspond to the reliability of the obtained recognition results. In this way, the user may be provided with one or more recognition results, which may be different or partially identical, the user may choose from his experience the result that is trusted, or may determine the current ventilation status of the patient based on the degree of confidence.

In one example, the processor 50 performs an identification process on the respiration waveform data of the target subject, determines an identification result of the current ventilation status of the target subject, and further includes: after the respiratory waveform data of the target object is subjected to the recognition processing based on the trained recognition model, the respiratory waveform data of the target object is subjected to the recognition processing by using an expert system, for example, when the number of trained neural network models is 1 as shown in fig. 4, the recognition result of the current ventilation state of the patient is determined by using the expert system according to the output values of the trained neural network models and by performing the recognition processing in combination with the respiratory waveform data (for example, by reasoning using a knowledge base of the expert system), and for example, when the number of trained neural network models is a plurality of, the recognition result of the current ventilation state of the patient is determined by using the expert system according to the output values of the respective trained neural network models and by performing the recognition processing in combination with the respiratory waveform data as shown in fig. 7.

In one example, the processor 50 performs an identification process of respiratory waveform data of a target subject using an expert system, including: acquiring patient data of a target object; the respiratory waveform data and the patient data are identified by using an expert system, for example, as shown in fig. 5 and 8, the identification result (including the type of the identified identification result) and the confidence coefficient P which are output by identifying the trained identification model (for example, the neural network model) are input to the expert system, the acquired patient data and respiratory waveform data are identified by using the expert system, the knowledge base in the expert system is used for reasoning, and the output result of the acquired neural network model is combined to give a final identification result and a final confidence coefficient P1.

Alternatively, the patient data may be input by the user through a human-computer interface, or may be obtained by the processor 50 by obtaining information such as an electronic medical record of the patient. Alternatively, the patient data may include one or more of the following: patient medical record information, mechanical ventilation parameters, monitoring parameters, blood gas value parameters, ultrasound parameters, disease severity scores, fluid balance information, and the like, wherein the patient medical record information includes at least one of the following: age, height, weight, BMI index (where BMI index is weight (kg) divided by height (m) square), comorbidities information, genetic disease information, medication information (e.g., antibiotic type), etc.; the mechanical ventilation parameters include at least one of the following information: tidal volume (Vt), minute Ventilation (MV), respiratory Rate (RR), inspiratory time, PEEP, mean airway pressure, inhaled oxygen concentration (FiO 2), pressure support level (PS level), maximum Inspiratory Pressure (MIP), maximum Expiratory Pressure (MEP), end tidal carbon dioxide (ETCO 2), lung compliance, airway resistance, mechanical energy, respiratory effort, driving pressure, peak pressure, plateau pressure, etc., and the monitored parameters include at least one of the following parameters: blood oxygen (SpO 2), heart Rate (HR), blood Pressure (BP), mean Arterial Pressure (MAP), central Venous Pressure (CVP), temperature (T), etc.; the blood gas value parameter includes at least one of the following information: arterial blood carbon dioxide partial pressure (PaCO 2), arterial blood oxygen partial pressure (PaO 2), arterial blood gas pH; the disease severity score includes at least one of the following scores: acute physiological and chronic health assessment IV (APACHE-IV, acute Physiology and Chronic Health Evaluation IV), glasgow coma score (GCS, glasgow coma scale), oxford acute disease severity score (OASIS, oxford Acute Severity of Illness Score), sequential organ failure assessment (Sequential Organ Failure Assessment), simplified acute physiological score (SAPS, SIMPLIFIED ACUTE PHYSIOLOGY SCORE), simplified acute physiological score II (sapiii, SIMPLIFIED ACUTE PHYSIOLOGY SCORE II), acute physiological score (APSIII, acute Physiology Score III), logical organ dysfunction score (LODS, logistic Organ Dysfunction Score), and the like; the liquid balance information includes at least one of the following: urine volume, crystal volume.

It is worth mentioning that in the present application, the expert system is a computer program designed to model the problem solving capabilities of human experts. It is an intelligent computer program system that contains within it the knowledge and experience of one or more domain experts that can be used to solve the domain problem using the knowledge of human experts and the method of solving the problem, the expert system being pre-stored in memory and the processor 50 retrieving it for use when required. The expert system in the method combines the expertise, experience and problem solving method of a plurality of ICU clinicians to identify, judge and process the breathing waveform, and further confirm the identification result of the neural network model, so that the obtained identification result can be accepted by users, and the accuracy of the identification result is higher.

In one example, the processor 50 is configured to: selecting a corresponding identification mode according to an instruction input by a user; when the recognition mode is the first recognition mode, carrying out recognition processing on the breathing waveform data of the target object and the patient data of the target object based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object; when the recognition mode is the second recognition mode, the expert system is utilized to recognize the breathing waveform data of the target object and the patient data of the target object, and the recognition result of the current ventilation state of the target object is determined; and when the recognition mode is a third recognition mode, performing recognition processing on the breathing waveform data of the target object and the patient data of the target object by using the trained recognition model and the expert system, and determining a recognition result of the current ventilation state of the target object. In this way, more choices can be provided to the user so that the user can select the appropriate recognition mode according to his own needs.

In one example, the processor 50 is further configured to: and outputting prompt information when the ventilation abnormal event is determined to exist. The display 70 is used for acquiring and displaying the prompt information, and the prompt information can include text description of the event type of the ventilation abnormal event, or analog image prompt, etc., through which the user can be intuitively prompted, so that the user can acquire the ventilation abnormal event of the patient in time and take corresponding treatment measures for the ventilation abnormal event, thereby improving the breathing comfort of the patient.

In order to enable the display of the prompt information described above to quickly draw the attention of the user, in one example, the display 70 is configured to display the prompt information in a preset display manner including one or more of the following manners: highlighting, symbology, distinctive color, distinctive shading, flashing, or other suitable display.

In one example, keys are provided on a display interface of the display 70 of the medical device, and the processor 50 is configured to: when a user instruction input by a user through a key is acquired, the display 70 is controlled to display or hide the prompt information, and by setting the key, the user can display the prompt information on a display interface of the display 70 only when the user needs to display the prompt information, so that the influence on the use experience of the user caused by excessive display information on the display interface is avoided, and the prompt information can be hidden through the key when the user does not need to display the prompt information.

In one example, the medical device of the present application further comprises an alarm device, and the medical device further comprises an alarm device for alerting when the number or frequency of ventilation abnormal events occurring within a preset time satisfies a preset condition, for example, when the man-machine countermeasure index exceeds a preset threshold value within the preset time, the man-machine countermeasure index is a ratio of the number of periods of ventilation abnormal events to the total number of periods within the preset time. When the preset condition is met, the ventilation abnormal event is severe, and the negative influence on the patient is likely to be caused, so that the doctor is immediately attracted by means of alarming prompt, and reasonable treatment measures are carried out on the ventilation abnormal event.

The processor 50101 may generate alarm information (e.g., an alarm signal) when the number or frequency of ventilation abnormal events occurring within a preset time satisfies a preset condition. The alarm device is configured to acquire the alarm information and carry out alarm prompt, and the alarm modes of the alarm device include, but are not limited to, light, sound and other alarm modes, and the specific forms of the alarm device can be a flashing LED lamp, a buzzer, a loudspeaker and the like, for example, when the alarm device is the loudspeaker, the alarm device can also be used for ventilating prompt tones of abnormal events, such as voice broadcasting and the like, so that the requirements on the intensity of alarm signals and the like are met, and the alarm device is sufficient for attracting attention and vigilance of observers. In this way, real-time alerting may be implemented to prompt the user.

In one example, the processor 50 is further configured to: when the identification result is an abnormal ventilation event, according to the identification result, setting parameters of a breathing machine for ventilating a target object are automatically adjusted, such as adjusting setting parameters of tidal volume (Vt), minute Ventilation (MV), respiratory Rate (RR), inspiration time, PEEP, mean airway pressure, inhaled oxygen concentration (FiO 2), PS level (pressure support level), maximum Inspiratory Pressure (MIP), maximum Expiratory Pressure (MEP) and the like, so that a patient can quickly recover to a normal ventilation state from the abnormal ventilation event, and the safety and comfort of mechanical ventilation of the patient are improved.

In one example, the processor 50 is further configured to: when the recognition result is a ventilation abnormal event, according to the recognition result, the adjustment advice information of the setting parameters of the ventilator that ventilates the target object is output, and the display 70 is configured to acquire and display the adjustment advice information, in such a manner that the adjustment advice information can be given to the user to assist the user in handling the ventilation abnormal event.

The description of the medical device of the present application has been completed so far, but it will be understood that the medical device of the present application may have other functions in addition to the components and functions described, and will not be described here.

Next, a ventilation status recognition method according to the present application will be described with reference to fig. 9, which can be based on the aforementioned medical device as an execution subject, and the various technical features herein can be combined with each other without collision.

As an example, as shown in fig. 9, the ventilation status recognition method of the present application includes the following steps S910 to S920:

First, in step S910, current respiration waveform data of a target subject is acquired, wherein the respiration waveform data includes basic respiration waveform data and auxiliary waveform data. Details of this step may be found in the previous relevant description.

Next, in step S920, the breathing waveform data of the target subject is subjected to recognition processing based on the trained recognition model, and a recognition result of the current ventilation state of the target subject is determined, the recognition result including normal ventilation or a ventilation abnormality.

Optionally, the trained recognition model is trained by a machine learning method. For example, the trained recognition model includes a trained neural network model, and the trained recognition model is trained by a deep learning method based on historical data information of respiratory waveform data of a past patient, wherein the historical data information of respiratory waveform data includes historical data information of basic respiratory waveform data and historical data information of auxiliary waveform data.

In one example, the identifying the respiration waveform data, determining an identification of the current ventilation status of the target subject, includes: preprocessing respiratory waveform data of a target object to obtain processed respiratory waveform data, wherein the preprocessing comprises normalization processing; and carrying out recognition processing on the processed respiratory waveform data based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object.

In one example, identifying respiratory waveform data of a target subject based on a trained identification model, determining an identification of a current ventilation status of the target subject includes: judging whether the quality of the auxiliary waveform data and the basic respiration waveform data meet the preset condition or not; when the quality of the auxiliary waveform data and the basic respiration waveform data meet the preset conditions, inputting the auxiliary waveform data and the basic respiration waveform data meeting the preset conditions into a trained recognition model for recognition processing, and determining a recognition result of the current ventilation state of the target object; when the quality of the auxiliary waveform data and the basic waveform is determined not to meet the preset condition, the auxiliary waveform data and the basic respiration waveform data are not input into a trained recognition model for recognition processing. The accuracy of the identification result can be ensured through quality judgment, and the problem that an error result is output or no result is output due to unqualified quality of the breathing waveform data is avoided.

In one example, identifying respiratory waveform data based on a trained identification model, determining an identification of a current ventilation status of a target subject includes: performing fusion processing on the basic respiration waveform data of the target object and/or the auxiliary waveform data of the target object to obtain fusion data; and inputting the fusion data into a trained recognition model for recognition processing, and determining a recognition result of the current ventilation state of the target object.

In one example, performing an identification process on respiratory waveform data of a target subject based on a trained identification model, determining an identification of a current ventilation status of the target subject includes: and carrying out recognition processing on the breathing waveform data of the target object based on the trained recognition model, and determining the recognition result of the current ventilation state of the target object and the credibility corresponding to the recognition result.

In one example, the identifying the respiration waveform data, determining an identification of the current ventilation status of the target subject, includes: after the breathing waveform data of the target object is identified based on the trained identification model, the expert system is utilized to identify the breathing waveform data of the target object, and the identification result of the current ventilation state of the target object is determined.

In one example, the method of the present application further comprises: when the ventilation abnormal event is determined to exist, a prompt message is output, and the prompt message is displayed through a display.

In one example, the method of the present application further comprises: and when the frequency or the frequency of the ventilation abnormal events in the preset time meets the preset condition, alarming and prompting are carried out. The alarm indication may be based on the alarm device described above.

In one example, the method of the present application further comprises: and when the identification result is a ventilation abnormal event, automatically adjusting the setting parameters of the breathing machine for ventilating the target object according to the identification result.

According to the medical equipment and the method, the basic waveform and the auxiliary waveform data of the target object can be identified based on the trained identification model, and the identification result of the current ventilation state of the target object can be determined, so that the accuracy of ventilation abnormal event identification in the aspect of man-machine synchronization can be improved, a doctor can be assisted in judging man-machine countermeasure events, the occurrence of man-machine countermeasure is reduced, and the comfort and safety of breathing of a mechanically ventilated patient are improved.

A medical device in another embodiment of the present application will be described below with reference to fig. 1,2, and 10 to 13.

As an example, the medical device of the embodiment of the present application may be a ventilator, an anesthesia machine, a monitor, a central station, or the like. When the medical device is a monitor or a central station, the medical device comprises a communication interface, and the communication interface can be in communication connection with a breathing machine, so that various data output by the breathing machine or the anesthesia machine, such as breathing waveform data, various set parameter information and the like, can be obtained, and in the application, the breathing waveform data refers to waveform data related to breathing. In one example, when the medical device is a central station, the central station may directly acquire various data output from the ventilator or the anesthesia machine through a communication interface provided therein, or the central station may be further connected to a monitor, and the monitor is connected to the ventilator or the anesthesia machine so as to acquire various data output therefrom, and transmit the data to the central station. In some examples, the ventilator or anesthesia machine may further include a communication interface through which the monitor or central station is communicatively connected to obtain various monitored parameters of the patient for which it is ventilated, such as blood oxygen (SpO 2), heart Rate (HR), blood Pressure (BP), mean Arterial Pressure (MAP), central Venous Pressure (CVP), temperature (T), etc., from the monitor or central station.

In one embodiment, as shown in FIG. 1, the medical device 100 of the present application includes one or more processors 50, a display 70, a memory 60, and a communication interface, among others. These components are interconnected by a bus system and/or other forms of connection mechanisms (not shown). It should be noted that the components and structures of the medical device 100 shown in fig. 1 are exemplary only and not limiting, and that the medical device 100 may have other components and structures as desired. Reference is also made to the foregoing description for some details of the medical device 100, which are not repeated here.

In some embodiments, the medical device may be a ventilator, which is an artificial mechanical ventilator to assist or control the respiratory motion of the patient to effect gas exchange within the lungs, reducing work done by the patient's breathing to facilitate recovery of respiratory function. Referring to fig. 2, in some embodiments the ventilator may further include a respiratory interface 211 (i.e., patient interface), a source interface 212, a respiratory circuit (i.e., respiratory circuit), a respiratory assistance device (i.e., an airflow providing device), a ventilation detection device for detecting ventilation parameters such as respiratory waveform data, a processor 50, a memory 60, and a display 70, among others. For details on the ventilator of fig. 2, reference is made to the foregoing.

Wherein the processor 50 of the medical device is configured to execute the program instructions stored in the memory, such that the processor 50 performs the steps of: acquiring current respiration waveform data and patient data of a target object, wherein the respiration waveform data comprises basic respiration waveform data and/or auxiliary waveform data; and carrying out recognition processing on the breathing waveform data and the patient data based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object, wherein the recognition result comprises normal ventilation or ventilation abnormal events.

It should be noted that, in the embodiment of the present application, breathing waveform data refers to data related to breathing, which may be a waveform of a parameter or data capable of characterizing the waveform of the parameter, and the like.

In one example, the base respiratory waveform data includes one or more of the following waveform data: pressure, flow rate or volume. The auxiliary waveform data includes one or more of the following waveforms: trans-pulmonary pressure, esophageal pressure, intragastric pressure, or diaphragmatic electricity, or other waveform data that is capable of reflecting the respiratory state of a patient. The above-described respiration waveform data may be obtained based on detection of a ventilation detecting means of a ventilator for detecting a ventilation parameter such as respiration waveform data, for example, based on acquisition of diaphragm myoelectricity or the like by, for example, a myoelectricity acquisition sensor, and for example, esophageal pressure may be measured by inserting a catheter for measuring esophageal pressure into the esophagus of a patient, and for gastric pressure may be measured by inserting a catheter for measuring gastric pressure into the stomach of a patient.

In one example, one or more of the above respiratory waveform data may also be obtained based on one or more data calculations, e.g., the cross-lung pressure is the difference between intra-alveolar pressure and intra-thoracic pressure, where intra-alveolar pressure may be characterized by measuring airway pressure, measuring esophageal pressure to estimate intra-thoracic pressure, and further obtaining cross-lung pressure based on airway pressure and intra-thoracic pressure calculations, and, for example, volume may be obtained by flow rate integration. For another example, new auxiliary waveform data may also be calculated based on one or more of the respiratory waveform data (e.g., the base respiratory waveform data and auxiliary waveform data), i.e., other auxiliary waveform data that is responsive to the respiratory state of the patient in addition to the enumerated trans-pulmonary pressure, esophageal pressure, intragastric pressure, or diaphragmatic power, such as combining waveforms of pressure, flow rate, volume, and esophageal pressure to derive new auxiliary waveform data.

For example, the trained recognition model may be obtained by training a neural network model using deep learning, i.e., the trained recognition model includes a trained neural network model, optionally including but not limited to a CNN multi-layer convolutional neural network, a RNN convolutional neural network, or a cnn+rnn network, wherein the trained recognition model is obtained by training using a deep learning method based on patient data of a previous patient and historical data information of respiratory waveform data of the previous patient, the historical data information of respiratory waveform data of the previous patient including historical data information of basic respiratory waveform data and historical data information of auxiliary waveform data.

For possible missing problems of basic respiratory waveform data such as pressure, flow velocity and volume waveforms input into a trained recognition model in practical application, in one example, before training a neural network model by using a deep learning method to obtain the trained recognition model, historical data information of respiratory waveform data is subjected to first data processing to obtain abnormal waveforms with waveform missing so as to obtain more abnormal waveforms, and then training the neural network model by using a deep learning method so as to obtain the trained recognition model, the abnormal waveforms with waveform missing are put into a database, so that samples in the database are expanded, and the adaptability of the neural network model to waveform missing is further enhanced. Optionally, the first data processing method includes, but is not limited to: data occlusion, noise addition, partial waveform removal, any type of waveform removal, and the like.

In order to make the recognition result of the trained recognition model better and more accurate, during the training process, the breathing waveform data of the samples in the database may be preprocessed to improve the recognition accuracy of the neural network model, for example, the preprocessing method includes, but is not limited to, data normalization and/or filtering, and taking pressure, flow rate, volume and esophageal pressure as examples, the data normalization may be performed based on the following formulas:

After preprocessing, the neural network model may be trained by a deep learning algorithm to complete the mapping from respiratory waveform data (e.g., basal respiratory waveform data and auxiliary waveform data) to ventilation abnormality types.

In the training process, patient data of the previous patient can be input into the neural network model, for example, the patient data of the previous patient is mapped through F transformation to obtain a parameter matrix, and the parameter matrix and the neural network output of the training respiratory waveform data are subjected to information fusion, for example, inner product and the like, so that the recognition result is finally output. It should be noted that the neural network model may include a plurality of layers, including an input layer, an hidden layer and an output layer, where for the convolutional neural network, the hidden layer may include a convolutional layer, a pooled layer and a fully connected layer, for example, breathing waveform data of the convolutional neural network is input to the input layer of the neural network model and is processed by the hidden layer of the neural network model to obtain a neural network output (the neural network output may refer to a feature of an extracted output of a preset layer in the hidden layer), while patient data of a patient in the past is mapped by F-transformation to obtain a parameter matrix, and the parameter matrix and the obtained neural network output are information fused, for example, inner product, in the neural network model (for example, the parameter matrix and the neural network output are input together to a next layer of the preset layer of the hidden layer, which may be the hidden layer or may also be the fully connected layer), so as to output a final recognition result.

The process of the processor 50 for identifying the respiratory waveform data of the target subject and the patient data of the target subject based on the trained identification model and determining the identification result of the current ventilation state of the target subject will be described below.

In one example, the processor 50 performs an identification process on the respiratory waveform data based on the trained identification model, determining an identification of the current ventilation status of the target subject, including: and carrying out recognition processing on the breathing waveform data and the patient data based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object, wherein the recognition result comprises normal ventilation or ventilation abnormal events.

In one example, the processor 50 performs an identification process on the respiration waveform data of the target subject based on the trained identification model, determining an identification of the current ventilation status of the target subject, including: determining whether the quality of the respiratory waveform data (e.g., including the base respiratory waveform data and/or the auxiliary waveform data) meets a preset condition; when the quality of the respiration waveform data meets the preset condition, inputting the respiration waveform data meeting the requirement and the patient data of the target object into a trained recognition model for recognition processing, and determining a recognition result of the current ventilation state of the target object; when the quality of the respiration waveform data is determined not to meet the preset condition, the respiration waveform data and the patient data of the target object are not input into the trained recognition model for recognition processing. The accuracy of the identification result can be ensured through quality judgment, and the problem that an error result is output or no result is output due to unqualified quality of the breathing waveform data is avoided.

The processor 50 may determine whether the quality of the auxiliary waveform data and the basic respiration waveform data satisfies a preset condition by a time domain method, a frequency domain method, a waveform template, or the like. The time domain method is to judge through preset characteristics of waveforms in the time domain, such as Maximum Inspiratory Pressure (MIP), mean airway pressure, slope or time of pressure rise, inspiratory time, maximum flow rate, mean flow rate, tidal volume (Vt) and other preset characteristic parameter indexes, wherein each preset parameter index corresponds to a normal range, waveform quality can be judged through the parameters, if the waveform quality is good in a threshold range, otherwise, the waveform is judged to be unqualified; for the frequency domain method: converting the waveform into a frequency domain range for spectrum analysis to obtain the duty ratio of different frequencies, comparing the duty ratio with the frequency spectrum of a normal waveform, judging through a threshold value, for example, when the duty ratio of a high-frequency part reaches a preset threshold value, considering that the quality of the waveform is unqualified due to excessive noise, and when the duty ratio of the high-frequency part is lower than the preset threshold value, judging that the quality of the waveform is qualified, or for some waveforms, when the duty ratio of the low-frequency part reaches the preset threshold value, judging that the quality of the waveform is unqualified due to excessive noise, and when the duty ratio of the low-frequency part is lower than the preset threshold value, judging that the quality of the waveform is qualified. For the waveform template analysis method: the waveform template can be established according to the normal waveform (for example, the pressure template is established according to a plurality of normal pressure waveforms), breathing waveform data which is input into the trained recognition model in a preset mode is compared with the waveform template corresponding to the breathing waveform data, the difference degree of the whole period is calculated, the quality is judged through the threshold value, for example, when the difference degree is larger than a preset threshold value, the quality of the breathing waveform data is judged to be unqualified, and when the difference degree is smaller than or equal to the preset threshold value, the quality of the breathing waveform data is judged to be qualified.

It should be noted that, when there are a plurality of basic respiration waveform data and a plurality of auxiliary waveform data, and the quality of a part of the basic respiration waveform data and a part of the auxiliary waveform data in the plurality of basic respiration waveform data satisfy the preset condition, the basic respiration waveform data and the auxiliary waveform data satisfying the preset condition may also be input to the trained recognition model for recognition.

In one example, as shown in fig. 11, performing an identification process on the respiratory waveform data and the patient data based on a trained identification model, determining an identification result of a current ventilation status of the target subject includes: inputting respiratory waveform data of the target subject to the trained neural network model to obtain a network output result (the network output result may be a feature extracted from an underlying layer of the neural network model); inputting patient data of the target object into the trained neural network model and mapping the patient data of the target object to a parameter matrix via a transformation map, which may alternatively include, but is not limited to, full connection layer mapping, fourier transforms (also referred to herein as F-transforms), or other functional transforms; in the trained neural network model, the parameter matrix and the network output result are subjected to information fusion to obtain the identification result, and optionally, the information fusion comprises, but is not limited to, inner product.

In one example, the number of trained recognition models may be 1, e.g., the trained recognition model is a trained neural network model, and the processor 50 is further configured to: performing fusion processing on the basic respiration waveform data of the target object and/or the auxiliary waveform data of the target object to obtain fusion data; and inputting the fusion data and the patient data of the target object into a trained recognition model for recognition processing, and determining a recognition result of the current ventilation state of the target object. For example, when the quality of the respiratory waveform data is acceptable, the different kinds and amounts of the basic respiratory waveform data and/or the different kinds of auxiliary waveform data may be combined to form a plurality of types of combined waveform data, for example, a combination of at least two types of waveform data in the basic respiratory waveform data, or a combination of at least one type of waveform data in the basic respiratory waveform data and at least one type of waveform data in the auxiliary waveform data, or may also be a combination of at least two types of waveform data in the auxiliary waveform data. The fusion process refers to fusing each type of combined waveform data (at least two types of waveform data are included in the combined waveform data), for example, size alignment, etc., and the purpose of the fusion process is to determine the format that is finally input to the trained recognition model.

Optionally, the recognition result of the current ventilation status includes, but is not limited to, normal ventilation or ventilation abnormality, wherein the ventilation abnormality includes one or more of the following events: ineffective triggering, double inspiration, false triggering, too low a flow rate, reverse triggering, early switching, delayed switching, slow pressure rise, too high a pressure rise, endogenous positive end expiratory pressure, short inspiration time, long inspiration time, line water accumulation, too high resistance, too low a compliance, inversely proportional ventilation, etc., or may also include other ventilation anomalies that can be identified.

In one example, the processor 50 performs an identification process on the respiration waveform data of the target subject and the patient data of the target subject based on the trained identification model, determining an identification of the current ventilation status of the target subject, including: and carrying out recognition processing on the breathing waveform data of the target object and the patient data of the target object based on the trained recognition model, and determining the recognition result of the current ventilation state of the target object and the reliability (also called confidence) corresponding to the recognition result. Through the credibility, an evaluation of the occurrence probability of the identification result can be given to the user, so that the user is assisted in judging whether the identification result is credible or not.

In one example, the number of trained recognition models may also be a plurality, for example, when the trained recognition models include a plurality of trained neural network models, the respiratory waveform data includes at least one of a plurality of combined waveform data, each trained neural network model corresponding to one of the combined waveform data, optionally the combined waveform data includes one or more of the following combinations: the combination of the at least one type of basal respiration waveform data, the at least one type of auxiliary waveform data, and the at least one type of basal respiration waveform data, for example, the basal respiration waveform data includes one or more of the following waveform data: pressure, flow rate, or volume, the auxiliary waveform data includes one or more of the following waveforms: the cross-lung pressure, esophageal pressure, intragastric pressure or diaphragmatic electricity, the at least one type of basic respiration waveform data may be pressure, flow rate or solvent, may further include any two types of combinations of pressure, flow rate and solvent and combinations of three types thereof, and the at least one type of auxiliary waveform data and the at least one type of basic respiration waveform data may further include the following combinations: a combination of one type of auxiliary waveform data and one type of basic respiratory waveform data (e.g., pressure and esophageal pressure, flow rate and intragastric pressure, etc.), a combination of two types of auxiliary waveform data and one type of basic respiratory waveform data (e.g., esophageal pressure, intragastric pressure and pressure), a combination of two types of auxiliary waveform data and two types of basic respiratory waveform data (e.g., esophageal pressure, intragastric pressure, volume and flow rate), a combination of three types of auxiliary waveform data and three types of basic respiratory waveform data (e.g., esophageal pressure, trans-pulmonary pressure, intragastric pressure, volume and flow rate), a combination of four types of auxiliary waveform data and three types of basic respiratory waveform data (e.g., diaphragmatic power, esophageal pressure, trans-pulmonary pressure, intragastric pressure, volume and flow rate), and the like. Patient data may also be input into each neural network model separately.

The processor 50 is also configured to: at least one kind of combined waveform data is obtained based on the obtained basic respiration waveform data and auxiliary waveform data, various kinds of combined waveform data are respectively input into the corresponding trained neural network models for identification, and patient data of a target object can be respectively input into each trained neural network model so as to respectively obtain a group of identification results through identification, wherein the identification results can comprise normal ventilation or abnormal ventilation events and correspond to the reliability of the obtained identification results. In this way, the user may be provided with one or more recognition results, which may be different or partially identical, the user may choose from his experience the result that is trusted, or may determine the current ventilation status of the patient based on the degree of confidence. Where recognition result 1 includes, but is not limited to, ineffective triggering, double inspiration, too low a flow rate, endogenous PEEP, short inspiration time, line water accumulation, too high a resistance or too low a compliance, and the like. The recognition result 2 and the recognition result n include, but are not limited to, ineffective triggering, double inspiration, false triggering, too low a flow rate, reverse triggering, endogenous PEEP, short inspiration time, pipeline water accumulation, too high resistance, too low compliance and the like.

In one example, the processor 50 performs an identification process on the respiration waveform data of the target subject, determines an identification result of the current ventilation status of the target subject, and further includes: after the recognition processing is performed on the respiratory waveform data of the target object and the patient data of the target object based on the trained recognition model, the recognition result of the current ventilation state of the target object is determined by performing recognition processing on the respiratory waveform data and the patient data by using an expert system, for example, as shown in fig. 12, by performing recognition processing (for example, reasoning by using a knowledge base of the expert system) by using the expert system according to the output value of the trained neural network model and combining the respiratory waveform data and the patient data, so as to determine the recognition result of the current ventilation state of the patient (this result may be occurrence probability of each abnormal event), and for example, when the number of trained neural network models is plural, recognition processing is performed by using the expert system according to the output value of each trained neural network model and combining the respiratory waveform data and the patient data, so as to determine the recognition result of the current ventilation state of the patient.

In one example, the processor 50 performs an identification process on the respiration waveform data of the target subject using an expert system, including: acquiring patient data of a target object (e.g., acquiring patient data entered via a human-machine interaction interface); the respiratory waveform data and the patient data are identified by using an expert system, for example, as shown in fig. 13, an identification result (including the type of the identified identification result) and a confidence coefficient P which are output by identifying a trained identification model (such as a neural network model) are input to the expert system, the acquired patient data and respiratory waveform data are identified by using the expert system, a knowledge base in the expert system is used for reasoning, and a final identification result and a final confidence coefficient P1 are given by combining the output result of the acquired neural network model.

Alternatively, the patient data may be input by the user through a human-computer interaction interface, or may be electronic medical record information of the target object obtained by the processor 50; based on the electronic medical record information, at least part of the patient data of the target object is acquired, and further, for some parameters, such as mechanical ventilation parameters (e.g. parameters measured by a ventilator or an anesthesia machine), monitoring parameters (typically parameters monitored by a monitor), ultrasound parameters (typically measured by an ultrasound device), etc. may be acquired from a monitoring device of each parameter, or for ultrasound parameters may also be an ultrasound device communicatively connected to a central station, and a ventilator is communicatively connected to the central station, and a ventilator acquires ultrasound parameters measured by the ultrasound device from the central station, etc. Alternatively, the patient data may include one or more of the following: patient medical record information, mechanical ventilation parameters, monitoring parameters, blood gas value parameters, ultrasound parameters, disease severity scores, fluid balance information, and the like, wherein the patient medical record information includes at least one of the following: age, height, weight, BMI index (where BMI index is weight (kg) divided by height (m) square), comorbidities information, genetic disease information, medication information (e.g., antibiotic type), etc.; the mechanical ventilation parameters include at least one of the following information: tidal volume (Vt), minute Ventilation (MV), respiratory Rate (RR), inspiratory time, PEEP, mean airway pressure, inhaled oxygen concentration (FiO 2), pressure support level (PS level), maximum Inspiratory Pressure (MIP), maximum Expiratory Pressure (MEP), end tidal carbon dioxide (ETCO 2), lung compliance, airway resistance, mechanical energy, respiratory effort, driving pressure, peak pressure, plateau pressure, etc., and the monitored parameters include at least one of the following parameters: blood oxygen (SpO 2), heart Rate (HR), blood Pressure (BP), mean Arterial Pressure (MAP), Central Venous Pressure (CVP), temperature (T), etc.; the blood gas value parameter includes at least one of the following information: arterial blood carbon dioxide partial pressure (PaCO 2), arterial blood oxygen partial pressure (PaO 2), arterial blood gas pH; the disease severity score includes at least one of the following scores: acute physiological and chronic health assessment IV (APACHE-IV, acute Physiology and Chronic Health Evaluation IV), grassgo coma score (GCS, glasgow coma scale), oxford acute disease severity score (OASIS, oxford Acute Severity of Illness Score), sequential organ failure assessment (Sequential Organ Failure Assessment), Simplified acute physiology score (SAPS, SIMPLIFIED ACUTE PHYSIOLOGY SCORE), simplified acute physiology score II (sapiii, SIMPLIFIED ACUTE PHYSIOLOGY SCORE II), acute physiology score (APSIII, acute Physiology Score III), logical organ dysfunction score (LODS, logistic Organ Dysfunction Score), and the like; the liquid balance information includes at least one of the following: urine volume, crystal volume.

It should be noted that in the present application, the expert system is a computer program designed to model the problem solving capabilities of human experts. It is an intelligent computer program system that contains within it the knowledge and experience of one or more domain experts that can be used to solve the domain problem using the knowledge of human experts and the method of solving the problem, the expert system being pre-stored in memory and the processor 50 retrieving it for use when required. The expert system in the method combines the expertise, experience and problem solving method of a plurality of ICU clinicians to identify, judge and process the breathing waveform, and further confirm the identification result of the neural network model, so that the obtained identification result can be accepted by users, and the accuracy of the identification result is higher.

The neural network model may be used alone, the expert system may be used alone, or both models may be used simultaneously throughout the recognition system, and in one example, the processor 50 is configured to: selecting a corresponding identification mode according to an instruction input by a user; when the recognition mode is the first recognition mode, carrying out recognition processing on the breathing waveform data of the target object and the patient data of the target object based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object; when the recognition mode is the second recognition mode, the expert system is utilized to recognize the breathing waveform data of the target object and the patient data of the target object, and the recognition result of the current ventilation state of the target object is determined; and when the recognition mode is a third recognition mode, performing recognition processing on the breathing waveform data of the target object and the patient data of the target object by using the trained recognition model and the expert system, and determining a recognition result of the current ventilation state of the target object. In this way, more choices can be provided to the user so that the user can select the appropriate recognition mode according to his own needs.

In one example, keys are provided on a display interface of the display 70 of the medical device, and the processor 50 is configured to: when a user instruction input by a user through a key is acquired, the display 70 is controlled to display or hide the prompt information, and by setting the key, the user can display the prompt information on a display interface of the display 70 only when the user needs to display the prompt information, so that the influence on the use experience of the user caused by excessive information displayed on the display interface is avoided, and the prompt information can be hidden through the key when the user does not need the prompt information.

In one example, the medical device of the present application further comprises an alarm device, and the medical device further comprises an alarm device for alerting when the number or frequency of ventilation abnormal events occurring within a preset time satisfies a preset condition, for example, when the man-machine countermeasure index exceeds a preset threshold value within the preset time, the man-machine countermeasure index is a ratio of the number of periods of ventilation abnormal events to the total number of periods within the preset time. When the preset condition is met, the ventilation abnormal event is severe, and the negative influence on the patient is likely to be caused, so that the doctor is immediately attracted by means of alarming prompt, and reasonable treatment measures are carried out on the ventilation abnormal event. Optionally, the preset time may be set reasonably according to actual needs, for example, it may be 30s, 60s, 90s or any other suitable duration, or it may also be a time set by the respiratory cycle of the patient, for example, it may be a time length corresponding to 10 respiratory cycles of the patient, or a time length corresponding to any other integer multiple of respiratory cycles.

The processor 50 may generate an alarm message (e.g., an alarm signal) when the number or frequency of ventilation anomalies occurring within a preset time satisfies a preset condition. The alarm device is configured to acquire the alarm information and carry out alarm prompt, and the alarm modes of the alarm device include, but are not limited to, light, sound and other alarm modes, and the specific forms of the alarm device can be a flashing LED lamp, a buzzer, a loudspeaker and the like, for example, when the alarm device is the loudspeaker, the alarm device can also be used for ventilating prompt tones of abnormal events, such as voice broadcasting and the like, so that the requirements on the intensity of alarm signals and the like are met, and the alarm device is sufficient for attracting attention and vigilance of observers. In this way, real-time alerting may be implemented to prompt the user.

In one example, the processor 50 is further configured to: when the identification result is an abnormal ventilation event, according to the identification result, setting parameters of a breathing machine for ventilating a target object, such as triggering sensitivity, adjusting tidal volume (Vt), pressure rising time, minute ventilation volume (MV), respiratory Rate (RR), inspiration time, PEEP, mean airway pressure, inhaled oxygen concentration (FiO 2), PS level (pressure supporting level), maximum Inspiratory Pressure (MIP), maximum Expiratory Pressure (MEP) and the like, are automatically adjusted, so that a patient can quickly recover to a normal ventilation state from the abnormal ventilation event, a better man-machine synchronization effect is achieved, and the safety and comfort of mechanical ventilation of the patient are improved.

In one example, the processor 50 is further configured to: when the identification result is a ventilation abnormal event, evaluating the lung injury degree of the target object according to the identification result and the patient data of the target object; and the display is used for displaying the evaluation result of the lung injury degree, so that a user can judge the lung injury degree of the patient according to the evaluation result. For example, the degree of lung injury of the target subject may be estimated based on the recognition result and a tidal volume (e.g., minute tidal volume) in patient data of the target subject, and specifically, for example, when the recognition result is a double inspiratory event, the degree of lung injury may be predicted in combination with the minute tidal volume.

Next, a ventilation status recognition method according to the present application will be described with reference to fig. 14, which can be based on the aforementioned medical device as an execution subject, and the various technical features herein can be combined with each other without collision.

As an example, as shown in fig. 14, the ventilation status recognition method 700 of the present application includes the following steps S710 to S720:

First, in step S710, current respiratory waveform data and patient data of a target subject are acquired, wherein the respiratory waveform data includes basic respiratory waveform data and/or auxiliary waveform data. Details of this step may be found in the previous relevant description.

Next, in step S720, the respiratory waveform data and the patient data are subjected to an identification process based on the trained identification model, and an identification result of the current ventilation state of the target subject, which includes normal ventilation or a ventilation abnormality, is determined.

Optionally, the trained recognition model is trained by a machine learning method. For example, the trained recognition model includes a trained neural network model that is trained using a deep learning method based on patient data of a past patient and historical data information of respiratory waveform data of the past patient, the historical data information of respiratory waveform data of the past patient including historical data information of base respiratory waveform data and/or historical data information of auxiliary waveform data.

In one example, the identifying the respiration waveform data, determining an identification of the current ventilation status of the target subject, includes: preprocessing respiratory waveform data of a target object to obtain processed respiratory waveform data, wherein the preprocessing comprises normalization processing and/or filtering processing; and carrying out recognition processing on the processed respiratory waveform data and the patient data of the target object based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object.

In one example, identifying respiratory waveform data of a target subject based on a trained identification model, determining an identification of a current ventilation status of the target subject includes: judging whether the quality of the respiration waveform data meets a preset condition or not; when the quality of the respiration waveform data of the target object is determined to meet the preset condition, inputting the respiration waveform data of the target object and the patient data of the target object into the trained recognition model for recognition processing, and determining a recognition result of the current ventilation state of the target object; when the respiration waveform data of the target object is determined not to meet the preset condition, the respiration waveform data and the patient data are not input into the trained recognition model to be recognized, and a recognition result of the current ventilation state of the target object is determined. The accuracy of the identification result can be ensured through quality judgment, and the problem that an error result is output or no result is output due to unqualified quality of the breathing waveform data is avoided.

In one example, identifying respiratory waveform data based on a trained identification model, determining an identification of a current ventilation status of a target subject includes: performing fusion processing on the basic respiration waveform data of the target object and/or the auxiliary waveform data of the target object to obtain fusion data; and inputting the fusion data and the patient data into a trained recognition model for recognition processing, and determining a recognition result of the current ventilation state of the target object.

In one example, identifying the respiratory waveform data and the patient data of the target subject based on the trained identification model, determining an identification of the current ventilation status of the target subject includes: and carrying out recognition processing on the breathing waveform data and the patient data of the target object based on the trained recognition model, and determining the recognition result of the current ventilation state of the target object and the credibility corresponding to the recognition result.

In one example, the identifying the respiration waveform data, determining an identification of the current ventilation status of the target subject, includes: after the respiratory waveform data and the patient data of the target object are identified based on the trained identification model, the respiratory waveform data and the patient data of the target object are identified by an expert system, and the identification result of the current ventilation state of the target object is determined.

In one example, the method of the present application further comprises: when the identification result is a ventilation abnormal event, evaluating the lung injury degree of the target object according to the identification result and the patient data of the target object; and displaying the evaluation result of the lung injury degree through a display. Optionally, the degree of lung injury of the target subject is assessed based on the identification (e.g., double inhalation event) and a tidal volume in patient data of the target subject.

In addition, the embodiment of the invention also provides a computer storage medium, on which the computer program is stored. One or more computer program instructions may be stored on a computer readable storage medium, in which a processor may execute the program instructions stored by the storage device to perform the functions of (and/or other desired functions of) embodiments of the present invention herein, e.g., to perform the corresponding steps of ventilation status identification methods according to embodiments of the present invention, various applications and various data, e.g., various data used and/or generated by the applications, etc., may also be stored.

For example, the computer storage medium may include, for example, a memory card of a smart phone, a memory component of a tablet computer, a hard disk of a personal computer, read-only memory (ROM), erasable programmable read-only memory (EPROM), portable compact disc read-only memory (CD-ROM), USB memory, or any combination of the foregoing storage media.

Although the illustrative embodiments have been described herein with reference to the accompanying drawings, it is to be understood that the above illustrative embodiments are merely illustrative and are not intended to limit the scope of the present invention thereto. Various changes and modifications may be made therein by one of ordinary skill in the art without departing from the scope and spirit of the invention. All such changes and modifications are intended to be included within the scope of the present invention as set forth in the appended claims.

Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

In the several embodiments provided by the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described device embodiments are merely illustrative, e.g., the division of the elements is merely a logical functional division, and there may be additional divisions when actually implemented, e.g., multiple elements or components may be combined or integrated into another device, or some features may be omitted or not performed.

In the description provided herein, numerous specific details are set forth. However, it is understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.

Similarly, it should be appreciated that in order to streamline the invention and aid in understanding one or more of the various inventive aspects, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof in the description of exemplary embodiments of the invention. However, the method of the present invention should not be construed as reflecting the following intent: i.e., the claimed invention requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of this invention.

It will be understood by those skilled in the art that all of the features disclosed in this specification (including any accompanying claims, abstract and drawings), and all of the processes or units of any method or apparatus so disclosed, may be combined in any combination, except combinations where the features are mutually exclusive. Each feature disclosed in this specification (including any accompanying claims, abstract and drawings), may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.

Furthermore, those skilled in the art will appreciate that while some embodiments described herein include some features but not others included in other embodiments, combinations of features of different embodiments are meant to be within the scope of the invention and form different embodiments. For example, in the claims, any of the claimed embodiments may be used in any combination.

Various component embodiments of the invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art will appreciate that some or all of the functions of some of the modules according to embodiments of the present invention may be implemented in practice using a microprocessor or Digital Signal Processor (DSP). The present invention can also be implemented as an apparatus program (e.g., a computer program and a computer program product) for performing a portion or all of the methods described herein. Such a program embodying the present invention may be stored on a computer readable medium, or may have the form of one or more signals. Such signals may be downloaded from an internet website, provided on a carrier signal, or provided in any other form.

It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The use of the words first, second, third, etc. do not denote any order. These words may be interpreted as names.

Claims

A medical device, the medical device comprising:

a memory for storing executable program instructions;

A processor for executing the program instructions stored in the memory, causing the processor to perform the steps of:

Acquiring current respiration waveform data of a target object, wherein the respiration waveform data comprises basic respiration waveform data and auxiliary waveform data;

And carrying out recognition processing on the breathing waveform data of the target object based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object, wherein the recognition result comprises normal ventilation or ventilation abnormal events.
The medical device of claim 1, wherein the device comprises a plurality of sensors,

The trained recognition model is trained by a machine learning method.
The medical device of any one of claims 1-2, wherein the trained recognition model comprises a trained neural network model, the trained recognition model being trained using a deep learning method based on historical data information of past patient's respiratory waveform data, the historical data information of past patient's respiratory waveform data comprising historical data information of base respiratory waveform data and historical data information of auxiliary waveform data.
The medical device of any one of claims 1-3, wherein the processor performs an identification process on the respiratory waveform data based on a trained identification model, determining an identification of the current ventilation status of the target subject, comprising:

Preprocessing the respiration waveform data of the target object to obtain processed respiration waveform data;

And carrying out recognition processing on the processed respiratory waveform data based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object.
The medical device of claim 4, wherein the preprocessing comprises normalization processing and/or filtering processing.
The medical device of any one of claims 1-5, wherein the processor performs an identification process on the respiration waveform data of the target subject based on the trained identification model, determining an identification of the current ventilation status of the target subject, comprising:

Judging whether the quality of the auxiliary waveform data and the basic respiration waveform data meet the preset condition or not;

When the quality of the auxiliary waveform data and the basic respiration waveform data meet the preset conditions, inputting the auxiliary waveform data and the basic respiration waveform data meeting the preset conditions into the trained recognition model for recognition processing, and determining the recognition result of the current ventilation state of the target object;

And when the quality of the auxiliary waveform data and the basic waveform is determined not to meet the preset condition, the auxiliary waveform data and the basic respiratory waveform data are not input into the trained recognition model for recognition processing.
The medical device of any one of claims 1 to 6, wherein the processor is further configured to:

Fusing the basic respiration waveform data of the target object and/or the auxiliary waveform data of the target object to obtain fused data;

and inputting the fusion data into the trained recognition model for recognition processing, and determining a recognition result of the current ventilation state of the target object.
The medical device of any one of claims 1-7, wherein when the trained identification model comprises a plurality of trained neural network models, the respiratory waveform data comprises at least one of a plurality of combined waveform data, each trained neural network model corresponding to a respective one of the combined waveform data.
The medical device of any one of claims 1-8, wherein the base respiratory waveform data comprises one or more of the following waveform data: pressure, flow rate, or volume, the auxiliary waveform data comprising one or more of the following waveforms: trans-pulmonary pressure, esophageal pressure, intragastric pressure, or diaphragmatic electricity.
The medical device of any one of claims 1 to 9, wherein the processor performs an identification process on the respiration waveform data of the target subject based on the trained identification model, determining an identification of the current ventilation status of the target subject, comprising:

And carrying out recognition processing on the breathing waveform data of the target object based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object and the credibility corresponding to the recognition result.
The medical device of any one of claims 1 to 10, wherein the processor performs an identification process on the respiration waveform data of the target subject, determines an identification of the current ventilation status of the target subject, further comprising:

and after the recognition processing is carried out on the breathing waveform data of the target object based on the trained recognition model, an expert system is utilized to carry out recognition processing on the breathing waveform data of the target object, and a recognition result of the current ventilation state of the target object is determined.
The medical device of claim 11, wherein the processor performs an identification process of respiratory waveform data of the target subject using an expert system, comprising:

acquiring patient data of the target object;

The respiratory waveform data and the patient data are identified using the expert system.
The medical device of claim 1, wherein the processor is further configured to:

Acquiring current respiratory waveform data and patient data of a target object, wherein the respiratory waveform data comprises basic respiratory waveform data and/or auxiliary waveform data;

And carrying out recognition processing on the breathing waveform data and the patient data based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object.
The medical device of claim 13, wherein the trained recognition model comprises a trained neural network model, the processor further to: based on the patient data of the past patient and the historical data information of the respiratory waveform data of the past patient, training to obtain the trained recognition model by using a deep learning method, wherein the historical data information of the respiratory waveform data of the past patient comprises the historical data information of the basic respiratory waveform data and/or the historical data information of the auxiliary waveform data.
The medical device of claim 14, wherein the processor performs an identification process on the respiratory waveform data and the patient data based on a trained identification model, determining an identification of a current ventilation status of the target subject, comprising:

Inputting respiratory waveform data of the target object to the trained neural network model to obtain a network output result;

Inputting the patient data of the target object into the trained neural network model, and mapping the patient data of the target object through transformation to obtain a parameter matrix;

And in the trained neural network model, carrying out information fusion on the parameter matrix and the network output result to obtain the identification result.
The medical device of claim 13, wherein the processor performs an identification process on the respiratory waveform data and the patient data based on a trained identification model, determining an identification of the current ventilation status of the target subject comprising:

And after the recognition processing is carried out on the breathing waveform data of the target object and the patient data of the target object based on the trained recognition model, an expert system is utilized to carry out recognition processing on the breathing waveform data of the target object and the patient data, and a recognition result of the current ventilation state of the target object is determined.
The medical device of any one of claims 12-16, wherein the processor performs acquiring patient data of the target object, comprising:

acquiring electronic medical record information of the target object;

acquiring at least part of patient data of the target object based on the electronic medical record information;

Or alternatively

Patient data of the target object input by a user through a man-machine interaction interface are obtained.
The medical device of any one of claims 12 to 17, wherein the patient data includes at least one of the following information: patient medical record information, mechanical ventilation parameters, monitoring parameters, blood gas value parameters, ultrasound parameters, disease severity scores, fluid balance information, wherein the patient medical record information comprises at least one of the following information: age, height, weight, BMI index, co-morbid information, genetic disease information, medication information, the mechanical ventilation parameters including at least one of the following: tidal volume, minute ventilation, respiratory rate, inspiratory time, PEEP, mean airway pressure, inhaled oxygen concentration, pressure support level, maximum inspiratory pressure, maximum expiratory pressure, end tidal carbon dioxide, lung compliance, airway resistance, mechanical energy, work of breathing, driving pressure, peak pressure, plateau pressure, the monitored parameters comprising at least one of the following parameters: blood oxygen, heart rate, blood pressure, mean arterial pressure, central venous pressure, and temperature; the blood gas value parameter includes at least one of the following information: arterial blood carbon dioxide partial pressure, arterial blood oxygen partial pressure, arterial blood gas pH; the disease severity score includes at least one of the following scores: acute physiology and chronic health assessment IV, glasgang coma score, oxford acute disease severity score, sequential organ failure assessment, simplified acute physiology score II, acute physiology score, logical organ dysfunction score; the liquid balance information includes at least one of the following: urine volume, crystal volume.
The medical device of any one of claims 1 to 18, wherein the processor is further configured to: and outputting prompt information when the ventilation abnormal event is determined to exist.
The medical device of claim 19, further comprising a display for acquiring and displaying the reminder information.
The medical device of claim 20, wherein the display is configured to display the reminder information in a preset display manner, the preset display manner including one or more of: highlighting, additional symbology, distinctive color, distinctive shading, flashing.
The medical device of claim 19, wherein the medical device comprises a display having keys disposed on a display interface of the display, the processor to:

And when a user instruction input by the user through the key is acquired, controlling a display to display or hide the prompt information.
The medical device of any one of claims 1 to 22, further comprising an alarm means for alerting when the number or frequency of ventilation anomalies occurring within a preset time meets a preset condition.
The medical device of any one of claims 1 to 23, wherein the processor is further configured to:

and when the identification result is a ventilation abnormal event, automatically adjusting the setting parameters of the breathing machine for ventilating the target object according to the identification result.
The medical device of any one of claims 1 to 24, wherein the processor is further configured to:

And when the identification result is a ventilation abnormal event, outputting adjustment suggestion information of the setting parameters of the breathing machine for ventilating the target object according to the identification result.
The medical device of claim 25, wherein the medical device comprises a display for acquiring and displaying the adjustment advice information.
The medical device of any one of claims 1 to 26, wherein the processor is further configured to:

When the identification result is a ventilation abnormal event, evaluating the lung injury degree of the target object according to the identification result and the patient data of the target object; and

The medical device further comprises a display for displaying the result of the evaluation of the degree of lung injury.
The medical device of claim 27, wherein the processor evaluates the extent of lung injury in the target subject based on the identification and patient data of the target subject, comprising:

and assessing the degree of lung injury of the target subject based on the identification result and the tidal volume in the patient data of the target subject.
The medical device of any one of claims 1 to 28, wherein the ventilation abnormality event comprises one or more of the following events: ineffective triggering, double inspiration, false triggering, too low flow rate, reverse triggering, early switching, delayed switching, slow pressure rise, too fast pressure rise, endogenous positive end expiratory pressure, short inspiration time, long inspiration time, pipeline water accumulation, too high resistance, too low compliance and inversely proportional ventilation.
The medical device of any one of claims 1 to 29, wherein the medical device is a ventilator, anesthesia machine, monitor or central station.
The medical device of any one of claims 1 to 30, wherein when the medical device is a monitor or central station, the monitor or central station includes a communication interface for communication connection with a ventilator that ventilates the target subject to obtain the respiratory waveform data output by the ventilator.
The medical device of any one of claims 1 to 31, wherein the processor is configured to:

Selecting a corresponding identification mode according to an instruction input by a user;

when the recognition mode is a first recognition mode, carrying out recognition processing on the breathing waveform data of the target object based on the trained recognition model or carrying out recognition processing on the breathing waveform data of the target object and the patient parameters of the target object based on the trained recognition model, and determining the recognition result of the current ventilation state of the target object;

when the identification mode is a second identification mode, an expert system is utilized to carry out identification processing on the breathing waveform data of the target object or an expert system is utilized to carry out identification processing on the breathing waveform data of the target object and the patient parameters of the target object, and an identification result of the current ventilation state of the target object is determined;

and when the recognition mode is a third recognition mode, performing recognition processing on the breathing waveform data of the target object by using the trained recognition model and the expert system or performing recognition processing on the breathing waveform data of the target object and the patient parameters of the target object by using the trained recognition model and the expert system, and determining a recognition result of the current ventilation state of the target object.
A method of ventilation status identification, the method comprising:

Acquiring current respiration waveform data of a target object, wherein the respiration waveform data comprises basic respiration waveform data and auxiliary waveform data;

And carrying out recognition processing on the breathing waveform data of the target object based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object, wherein the recognition result comprises normal ventilation or ventilation abnormal events.
The method of claim 33, wherein the method further comprises:

acquiring patient data of a target object;

And carrying out recognition processing on the breathing waveform data and the patient data based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object.
The method of claim 33 or 34, wherein the trained recognition model is trained by a machine learning method.
The method of claim 33 or 34, wherein when the trained recognition model is used to recognize the respiration waveform data of the target subject, the trained recognition model comprises a trained neural network model that is trained using a deep learning method based on historical data information of respiration waveform data of a past patient, the historical data information of respiration waveform data of the past patient comprising historical data information of basal respiration waveform data and historical data information of auxiliary waveform data; or alternatively

When a trained recognition model is used to recognize the respiratory waveform data and the patient data, then the trained recognition model includes a trained neural network model, the trained recognition model being trained using a deep learning method based on patient data of a past patient and historical data information of the respiratory waveform data of the past patient, the historical data information of the respiratory waveform data of the past patient including historical data information of the basal respiratory waveform data and/or historical data information of the auxiliary waveform data.
The method of any of claims 33 to 36, wherein the identifying the respiratory waveform data to determine an identification of the current ventilation status of the target subject comprises:

preprocessing the respiration waveform data of the target object to obtain processed respiration waveform data, wherein the preprocessing comprises normalization processing;

And carrying out recognition processing on the processed respiratory waveform data based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object.
The method of any of claims 33 to 37, wherein the identifying the respiratory waveform data of the target subject based on the trained identification model, determining an identification of the current ventilation status of the target subject, comprises:

Judging whether the quality of the auxiliary waveform data and the basic respiration waveform data meet the preset condition or not;

When the quality of the auxiliary waveform data and the basic respiration waveform data meet the preset conditions, inputting the auxiliary waveform data and the basic respiration waveform data meeting the preset conditions into the trained recognition model for recognition processing, and determining a recognition result of the current ventilation state of the target object;

And when the quality of the auxiliary waveform data and the basic waveform is determined not to meet the preset condition, the auxiliary waveform data and the basic respiratory waveform data are not input into the trained recognition model for recognition processing.
The method of any of claims 33 to 38, wherein the identifying the respiratory waveform data based on the trained identification model, determining an identification of the current ventilation status of the target subject, comprises:

Fusing the basic respiration waveform data of the target object and/or the auxiliary waveform data of the target object to obtain fused data;

and inputting the fusion data into the trained recognition model for recognition processing, and determining a recognition result of the current ventilation state of the target object.
The method of any one of claims 33 to 39, wherein the identifying the respiratory waveform data of the target subject based on the trained identification model, determining an identification of the current ventilation status of the target subject, comprises:

And carrying out recognition processing on the breathing waveform data of the target object based on the trained recognition model, and determining a recognition result of the current ventilation state of the target object and the credibility corresponding to the recognition result.
The method of any one of claims 33 to 40, wherein said identifying said respiratory waveform data to determine an identification of a current ventilation status of said subject comprises:

and after the recognition processing is carried out on the breathing waveform data of the target object based on the trained recognition model, an expert system is utilized to carry out recognition processing on the breathing waveform data of the target object, and a recognition result of the current ventilation state of the target object is determined.
The method of any one of claims 33 to 41, further comprising: and outputting prompt information when the ventilation abnormal event is determined to exist.
The method of any one of claims 33 to 42, further comprising: and when the frequency or the frequency of the ventilation abnormal events in the preset time meets the preset condition, alarming and prompting are carried out.
The method of any one of claims 33 to 43, further comprising:

and when the identification result is a ventilation abnormal event, automatically adjusting the setting parameters of the breathing machine for ventilating the target object according to the identification result.
The method of any one of claims 33 to 44, further comprising:

When the identification result is a ventilation abnormal event, evaluating the lung injury degree of the target object according to the identification result and the patient data of the target object;

And displaying the evaluation result of the lung injury degree through a display.
The method of claim 45, wherein said assessing the extent of lung injury in said subject based on said identification and patient data of said subject comprises:

and assessing the degree of lung injury of the target subject based on the identification result and the tidal volume in the patient data of the target subject.