EP4330985A1

EP4330985A1 - Methods and systems for thermal capacity prediction of medical imaging devices

Info

Publication number: EP4330985A1
Application number: EP23766154.1A
Authority: EP
Inventors: Jie Yu
Original assignee: Shanghai United Imaging Healthcare Co Ltd
Current assignee: Shanghai United Imaging Healthcare Co Ltd
Priority date: 2022-03-11
Filing date: 2023-03-10
Publication date: 2024-03-06
Also published as: CN114596951A; WO2023169577A1

Abstract

A method implemented on at least one machine each of which has at least one processor and at least one storage device for thermal capacity prediction of a medical imaging device is provided. The method comprising: obtaining a first target prediction model, the first target prediction model including a machine learning model; obtaining property information of a tube of the medical imaging device and working status information of the medical imaging device, the working status information including a current thermal capacity of the tube and one or more scanning parameters; and determining a target predicted thermal capacity of the tube at a preset time point by processing, based on the first target prediction model, the property information and the working status information.

Description

METHODS AND SYSTEMS FOR THERMAL CAPACITY PREDICTION OF MEDICAL IMAGING DEVICES

CROSS-REFERENCE TO RELATED APPLICATIONS
This application is a Continuation of International Application No. 202210242007.9, filed on March 11, 2022, the contents of which are hereby incorporated by reference.

TECHNICAL FIELD

The present disclosure relates to the field of medical imaging devices, and in particular to methods and systems for thermal capacity prediction of medical imaging devices.

BACKGROUND

Medical imaging devices have become indispensable devices in the present medical field. When a medical imaging device is imaging a patient, the components in the medical imaging device can affect the quality of the imaging. The tube is a core part of a Computed Tomography (CT) scanning device, and reducing the loss of the tube is very important to ensure CT imaging. The thermal capacity of the tube can limit the scanning capability of the CT device, and too high thermal capacity can lead to loss of the tube.
Therefore, it is necessary to propose methods for thermal capacity prediction of the medical imaging device to predict the thermal capacity induced by the operation of the tube, and to take measures in advance based on the predicted thermal capacity to reduce the loss of the tube.
SUMMARY
One aspect of the present disclosure provides a method implemented on at least one machine each of which has at least one processor and at least one storage device for thermal capacity prediction of a medical imaging device, the method comprising: obtaining a first target prediction model, the first target prediction model including a machine learning model; obtaining property information of a tube of the medical imaging device and working status information of the medical imaging device, the working status information including a current thermal capacity of the tube and one or more scanning parameters; determining a target predicted thermal capacity of the tube at a preset time point by processing, based on the first target prediction model, the property information and the working status information.
Another aspect of the present disclosure provides a system implemented on at least one machine each of which has at least processor and at least one storage device for thermal capacity prediction of a medical imaging device, the system comprising: a model acquisition module, configured to obtain a first target prediction model, the first target prediction model including a machine learning model; an information acquisition module, configured to obtain property information of a tube of the medical imaging device and working status information of the medical imaging device, the working status information including a current thermal capacity of the tube and one or more scanning parameters; and a prediction module, configured to determine a target predicted thermal capacity of the tube at a preset time point by processing the property information and the working status information based on the first target prediction model.
A further aspect of the present disclosure provides a medical imaging device, including at least one processor and at least one storage; the at least one storage being configured to store computer instructions; the at least one processor being configured to execute at least a portion of the computer instructions to implement a method for thermal capacity prediction of a medical imaging device.
Some embodiments of the present disclosure have the following beneficial effects: (1) the accuracy of the thermal capacity prediction can be improved by predicting a thermal capacity through a machine learning model; (2) considering the property information of the tube in the prediction makes it possible to make the prediction based on the practical situation of the tube itself, with high practicability; (3) in practical applications, the parameters of the first prediction model can be updated based on augmented training which is based on the prediction result and the practical detected thermal capacity value, which can improve the prediction accuracy of the first prediction model and enable it to make a prediction based on the property information of the tube of the medical imaging device.

BRIEF DESCRIPTION OF THE DRAWINGS

This description will be further illustrated by means of exemplary embodiments which will be described in detail through accompanying drawings. These embodiments are not restrictive, in which the same numbering indicates the same structure, wherein:
FIG. 1 is a schematic diagram illustrating an application scenario of a thermal capacity prediction system according to some embodiments of the present disclosure;
FIG. 2 is a block diagram illustrating an exemplary thermal capacity prediction system according to some embodiments of the present disclosure;
FIG. 3 is a flowchart illustrating an exemplary process for thermal capacity prediction according to some embodiments of the present disclosure;
FIG. 4 is a flowchart illustrating an exemplary process for obtaining a first target prediction model according to some embodiments of the present disclosure;
FIG. 5 is another flowchart illustrating an exemplary process for obtaining a first target prediction model according to some embodiments of the present disclosure;
FIG. 6 is a schematic diagram illustrating an exemplary structure of a first target prediction model according to some embodiments of the present disclosure;
FIG. 7 is a schematic diagram illustrating an exemplary training process of a first target prediction model according to some embodiments of the present disclosure.

DETAILED DESCRIPTION

In order to more clearly explain the technical scheme of the embodiment of this description, a brief description of the accompanying drawings required for the embodiment description is given below. Obviously, the accompanying drawings below are only some examples or embodiments of this description, and it is possible for ordinary technicians skilled in the art to apply this description to other similar scenarios according to these accompanying drawings without creative effort. Unless obviously obtained from the context or the context illustrates otherwise, the same numeral in the drawings refers to the same structure or operation.
It should be understood that the “system” , “device” , “unit” and/or “module” used in this article are a method configured to distinguish different components, elements, parts, portions or assemblies of different levels. However, if other words serve the same purpose, the words may be replaced by other expressions.
As shown in this description and claims, the words “one” , “a” , “a kind” and/or “the” are not specially singular but may include the plural unless the context expressly suggests otherwise. In general, the terms “comprise” and “include” imply the inclusion only of clearly identified steps and elements that do not constitute an exclusive listing. A method or equipment may also include other steps or elements.
Flowcharts are used in this description to illustrate the operation performed according to the system of the embodiments of this description. It should be understood that the previous or subsequent operations may not be accurately implemented in order. Instead, each step may be processed in reverse order or simultaneously. Meanwhile, other operations may also be added to these processes, or a certain step or several steps may be removed from these processes.
FIG. 1 is a schematic diagram illustrating an application scenario of a thermal capacity prediction system according to some embodiments of the present disclosure.
As shown in FIG. 1, the thermal capacity prediction system 100 may include a processing device 110, a network 120, a user terminal 130, a storage device 140, and a medical imaging device 150.
In some embodiments, the thermal capacity prediction system 100 may predict a thermal capacity change of a tube during use of a medical imaging device by implementing the method and/or process disclosed in this present disclosure.
The processing device 110 may process the acquisition of data and/or information from the user terminal 130, the storage device 140, and/or the medical imaging device 150. The processing device 110 may access information and/or data from the user terminal 130, the storage device 140, and/or the medical imaging device 150 via the network 120. The processing device 110 may connect directly to the user terminal 130, the storage device 140, and/or the medical imaging device 150 to access the information and/or data. For example, the processing device 110 may obtain working status information of the medical imaging device 150 and property information of the tube of the medical imaging device 150 from the medical imaging device 150 and/or the user terminal 130. The processing device 110 may process the obtained data and/or information. For example, the processing device 110 may predict a target predicted thermal capacity of the tube at a preset time point based on the obtained working status information and the property information of the tube. As another example, the processing device 110 may be trained to obtain a first target prediction model and the first target prediction model may be used to predict the target predicted thermal capacity of the tube at the preset time point. In some embodiments, the processing device 110 may be a single server or a group of servers. The processing device 110 may be provided in the medical imaging device 150. The processing device 110 may be local or remote. The processing device 110 may be implemented on a cloud platform.
The network 120 may include any suitable network that provides information and/or data exchange capable of facilitating the thermal capacity prediction system 100. In some embodiments, the information and/or data may be exchanged between one or more components of the thermal capacity prediction system 100 (e.g., the processing device 110, the user terminal 130, the storage device 140, and/or the medical imaging device 150) via network 120. The network 120 may include a local area network (LAN) , a wide area network (WAN) , a wired network, a wireless network, or the like, or any combination thereof.
The user terminal 130 refers to one or more terminal devices or software used by a user. In some embodiments, the user terminal 130 may be a mobile device, a tablet computer, etc., or any combination thereof. In some embodiments, the user terminal 130 may interact with other components in the thermal capacity prediction system 100 via the network 120. For example, the user terminal 130 may send one or more control instructions to the medical imaging device 150 to control the processing device 110 to process the working status information of the medical imaging device 150 and the property information of the tube of the medical imaging device 150 to obtain the target predicted thermal capacity of the tube of the medical imaging device. In some embodiments, the user terminal 130 may be part of the processing device 110. In some embodiments, the user terminal 130 may be integrated with the processing device 110 as an operation platform for the medical imaging device 150.
The storage device 140 may be configured to store data, instructions, and/or any other information. In some embodiments, the storage device 140 may store data and/or information obtained from, for example, the server 110, the user terminal 130, the medical imaging device 150, etc. For example, the storage device 140 may store a pre-trained first target prediction model, second target prediction model, etc. As another example, the storage device 140 may store working parameters of the medical imaging device, the property information of the tube, a measured thermal capacity of the tube, etc. The storage device 140 may be provided in the medical imaging device 150. In some embodiments, the storage device 140 may include a mass storage, a removable storage, etc., or any combination thereof.
The medical imaging device 150 may be configured to obtain image data of a scanned object. The scanned object may include a biological object (e.g., human body, animal, etc. ) , a non-biological object (e.g., body model) , etc. In some embodiments, the medical imaging device 150 may be a CT imaging device, a PET-CT imaging device, a CT-RT device, etc. The CT imaging device may include a spiral CT device, a wide-cone CT device, a dual-source CT device, a dual-energy CT device, etc. In some embodiments, the medical imaging device 150 may include an X-ray tube ( “tube” ) that emits a radiation beam (e.g., X-rays) onto the scanned object to produce the image data. The tube may include a tube core and a tube shell. The tube core may include a cathode, a focal point, an anode target, a pivot bearing configured to support the anode target, etc. The tube shell may include a coolant, a thermal sink, etc. As previously described, the processing device 110 and the storage device 140 may be part of the medical imaging device 150. Correspondingly, the thermal capacity prediction may be performed by the medical imaging device 150.
FIG. 2 is a block diagram illustrating an exemplary thermal capacity prediction system according to some embodiments of the present disclosure.
In some embodiments, the thermal capacity prediction device 200 may include a model acquisition module 210, an information acquisition module 220, and a prediction module 230.
The model acquisition module 210 may be configured to obtain a first target prediction model. The first target prediction model may include a machine learning model. In some embodiments, the model acquisition module 210 may also be configured to: construct a second initial prediction model; obtain first training data; determine a second target prediction model by updating, based on the first training data parameters of the second initial prediction model; and determine a first target prediction model based on the second target prediction model. In some embodiments, the model acquisition module 210 may obtain second training data during a practical imaging, determine a first intermediate prediction model based on the second target prediction model, and obtain the first target prediction model by updating the first intermediate prediction model based on the second training data. In some embodiments, in practical imaging, the model acquisition module 210 may determine an evaluation result of the first intermediate prediction model by processing an intermediate predicted thermal capacity output by the first intermediate prediction model, a usage duration of the tube, a device type of the medical imaging device, and a count of updates of the first intermediate prediction model; in response to the evaluation result meeting a preset evaluation condition, terminate updating the first intermediate prediction model and obtain the first target prediction model. For more information about obtaining the first target prediction model, please refer to FIG. 4, FIG. 5, FIG. 7 and related descriptions thereof.
The information acquisition module 220 may be configured to obtain the property information of the tube of the medical imaging device and the working status information of the medical imaging device. In some embodiments, the information acquisition module 220 may obtain a measured thermal capacity of the tube at a preset time point based on a sensor. In some embodiments, the property information of the tube may include at least a current thermal dissipation parameter of the tube, and the current thermal dissipation parameter of the tube may be determined based on a thermal dissipation prediction model by: determining the current thermal dissipation parameter of the tube by processing usage data of the medical imaging device, a usage duration of the tube, and an initial thermal dissipation parameter of the tube based on the thermal prediction model. For more information about the property information of the tube, the working status information of the medical imaging device, and obtaining the measured thermal capacity, please refer to FIG. 3 and related descriptions thereof.
The prediction module 230 may determine the target predicted thermal capacity of the tube at the preset time point by processing the property information and working status information using the first target prediction model. The first target prediction model may be a machine learning model. For more information about determining the target predicted thermal capacity, please refer to FIG. 3 and related descriptions thereof, and for more information about the first target prediction model, please refer to FIG. 6 and related descriptions thereof.
In some embodiments, the thermal capacity prediction device 200 may further include a sending module (not shown in FIG. 2) configured to: determine whether the target predicted thermal capacity is greater than a thermal capacity threshold; in response to the target predicted thermal capacity being greater than the thermal capacity threshold, send an instruction for stopping the scanning to the medical imaging device, or send alert information to a user terminal. For more information about the sending module, please refer to FIG. 3 and related descriptions thereof.
It should be understood that the modules shown in FIG. 2 can be implemented by using various approaches. For example, in some embodiments, the system and its modules thereof may be implemented by hardware, software, or a combination of software and hardware. The hardware may be implemented by a specific logic. The software may be stored in a storage and executed by an appropriate instruction execution system, such as a microprocessor or a dedicated design hardware. It will be understood by those skilled in the art that the above-mentioned methods and systems may be implemented using computer-executable instructions and/or embedded in control codes of a processor. For example, the control codes may be provided by a carrier medium, such as a disk, a CD, or a DVD-ROM, a programmable storage such as a read-only memory (firmware) , or a data carrier such as an optical carrier or an electronic signal carrier. The system and its modules of the present disclosure may be implemented by a hardware circuit, which includes a semiconductor such as a very large-scale integration or gate array, a logic chip, a transistor, or the like, or a programmable hardware device such as a field programmable gate array, a programmable logic device, or the like. The system and its modules of the present disclosure may be implemented by a software, for example, a software executed by various types of processors. The system and its modules of the present disclosure may also be implemented by a combination of the hardware circuit and the software (e.g., a firmware) .
It should be noted that the above description of the system and its modules is merely provided for the purposes of illustration and does not limit the present disclosure to the scope of the cited embodiments. It will be understood that for those skilled in the art, after understanding the principle of the system, it is possible to arbitrarily combine various modules, or form subsystems to connect with other modules without departing from this principle. In some embodiments, the model acquisition module 210, the information acquisition module 220, and the prediction module 230 disclosed in FIG. 2 may be different modules in one system, or one module may implement the functions of two or more of the above modules. For example, each module may share a storage module, and each module may also have a respective storage module. Such deformations may be all within the scope of the protection of the present disclosure.
FIG. 3 is a flowchart illustrating an exemplary process for thermal capacity prediction according to some embodiments of the present disclosure. In some embodiments, process 300 may be performed by a processing device (e.g., the processing device 110) . For example, process 300 may be stored in a storage device in the form of a program or instructions, and process 300 may be implemented when the server or the module shown in FIG. 2 executes the program or instructions. In some embodiments, process 300 may be performed using one or more additional operations not described below, and/or performed without one or more of the operations discussed below. As shown in FIG. 3, process 300 may include one or more of the following operations.
In 310, a first target prediction model may be obtained. In some embodiments, operation 310 may be performed by the model acquisition module 210.
The first target prediction model may be configured to predict the thermal capacity of the tube. The first target prediction model may be a machine learning model. The parameters of the first target prediction model may be obtained by training an initial prediction model. The training data for the training may include first training data and second training data. The training process may include an initial training and/or an augmented training. The training based on the first training data may be called "initial training" and the training based on the second training data may be called "augmented training" . In some embodiments, the initial training may be performed offline in advance before the model application is released, and the augmented training may be performed online in real time during the practical imaging. For more information about "initial training" and "augmented training" , please refer to FIGs. 4, 5 and 7 and descriptions thereof.
In some embodiments, the first target prediction model may include one or more layers. For example, the first target prediction model may include an input layer, an incremental prediction layer, a fusion layer, and an output layer. For more information about the structure of the first target prediction model, please refer to FIG. 6 and related descriptions thereof.
In some embodiments, the type of the first target prediction model may include a convolutional neural network (CNN) , a deep neural network (DNN) , etc.
In some embodiments, the model acquisition module 210 may obtain a first target prediction model by training an initial prediction model. The trained first target prediction model may be stored in a storage device (e.g., the storage device 140) , and the model acquisition module 210 may obtain the first target prediction model from the storage device.
In 320, property information of a tube of the medical imaging device and working status information of the medical imaging device may be obtained. In some embodiments, operation 320 may be performed by the information acquisition module 220.
The medical imaging device may be configured to obtain image data of a scanned object, and the medical imaging device may include a tube. For more information about the medical imaging device and its tube, please refer to FIG. 1 and related description thereof.
The property information of the tube may be data information related to the tube itself. In some embodiments, the property information of the tube may include, but is not limited to, at least one of a target material of the tube, a target surface diameter of the tube, a target angle of the tube, a tube core material of the tube, a filtering equivalent excess of the tube wall, a thermal dissipation mode of the tube, a thermal dissipation coefficient (e.g., anode thermal dissipation coefficient) of the tube, a cooling rate (e.g., anode cooling rate) of the tube, and a support between a rotor and a bearing of the tube. For example, the target material of the tube may be tungsten, molybdenum, rhodium, rhenium, etc. ; the target surface diameter may be 50 mm, etc. ; the target angle may be 10° or 5°, etc. ; the tube core material may be ceramic, glass, etc. ; the filtering equivalent excess of the tube wall may be 5 mAL/140KV, etc. ; the cooling mode may be air cooling, water cooling, oil cooling, etc. ; the anode thermal coefficient may be 4 KW/min, etc. ; the anode cooling rate may be 10 KW/min, etc. ; the support between the rotor and the bearing may be mechanical ball bearing, liquid metal bearing, etc. It should be noted that the property information of the tube may also be or include other information related to the tube. In some cases, the thermal dissipation coefficient of the tube may change with usage of the tube. In some embodiments, the property information of the tube may include a current thermal dissipation coefficient of the tube. More descriptions of the current thermal dissipation coefficient of the tube may be found elsewhere in the present disclosure.
In some embodiments, the property information of the tube may be stored in a storage device, and the information acquisition module 220 may read the property information of the tube from the storage device. In some embodiments, the information acquisition module 220 may obtain the property information of the tube directly from the medical imaging device. In some embodiments, the property information of the tube may be different for different types of medical imaging devices. The property information of the tube of the same type of medical imaging devices may be different. For example, the property information of the tube in different CT devices from different manufacturers may be different.
The working status information of the medical imaging device may be data related to the imaging of the scanned object imaged by the medical imaging device. In some embodiments, the working status information may include a current thermal capacity of the tube and scanning parameters of the medical imaging device.
The current thermal capacity may be the thermal capacity of the tube at a current time point. The current time point may change according to different situations. For example, if the current time point is a time point ready to scan the object, the current thermal capacity may be the thermal capacity of the medical imaging device before the scanning begins. As another example, if the scanning has started, the current thermal capacity may be the thermal capacity of the medical imaging device during the scanning process.
In some embodiments, the information acquisition module 220 may obtain the current thermal capacity of the tube through a sensor. The sensor may be disposed in the medical imaging device and configured to determine the thermal capacity of the tube of the medical imaging device. The sensor may be a thermal capacity sensor, through which the thermal capacity may be directly obtained. The sensor may be other sensors, for example, a temperature sensor and a thermal sensor, and the temperature detected by the temperature sensor and the heat detected by the thermal sensor may be converted into the current thermal capacity of the tube. In some embodiments, the sensor may obtain the thermal capacity of the tube at a regular interval. For example, the sensor obtains the thermal capacity of the tube at an interval of 0.005 s and sends the thermal capacity to the storage device, etc. In some embodiments, the current thermal capacity of the tube may be obtained in other ways.
Scanning parameters may include information about parameters related to the scanning process of the medical imaging device. In some embodiments, the one or more scanning parameters may include, but are not limited to, a tube voltage of the tube, a tube current, a focal point size, an anode speed, a scanning time of the medical imaging device, etc.
In some embodiments, the information acquisition module 220 may obtain the one or more scanning parameters from the medical imaging device, which may automatically generate the scanning parameters based on feature parameters of the scanned object (e.g., age, weight, physical characteristics, medical conditions, etc. ) . In some embodiments, the information acquisition module 220 may obtain, from a terminal (e.g., terminal 130) , the scanning parameters set by medical personnel. The scanning parameters may be stored in a storage device (e.g., storage device 140) , and the information acquisition module 220 may obtain the scanning parameters from the storage device.
In 330, a target predicted thermal capacity of the tube at a preset time point may be determined by processing, based on the first target prediction model, the property information and the working status information. In some embodiments, operation 330 may be performed by the prediction module 230.
The preset time point may be after the current time and after the scanning of the medical imaging device is started. In some embodiments, the preset time point may be a time point at which the scanning ends, or a time point between the start and end of the scanning. The preset time point may be determined based on the scanning time of the medical imaging device. For example, if the scanning time is 1 min and the scanning starts at 16: 10, the preset time point may be determined to be 16: 11 based on this scanning time. As another example, the preset time point may be set to 40 s after the start of the scanning, 2 min after the start of the scanning, etc. In some embodiments, the preset time point may be set by medical personnel.
In some embodiments, the preset time point may be determined based on the practical situation (e.g., whether the tube is aging, the feature parameters of the object to be scanned, the scanning mode of the medical imaging device, the working environment, etc. ) . For example, the aging of the tube may reduce the thermal capacity of the tube (i.e., an indicator of the ability of the tube to carry heat) . When the tube is aging, the preset time point may be set to a relatively early time point after the start of the scanning process (e.g., 30 s after the start of the scanning process) to avoid interruptions in the scanning process due to the thermal capacity exceeding the thermal capacity. As another example, if the object to be scanned is an obese object, the tube is working under high load, and a long time high load can affect the thermal capacity change and the thermal dissipation coefficient of the tube, the preset time point may be set to a relatively early time point after the start of the scanning (for example, 40 s after the start of the scanning) to avoid the situation that the thermal capacity of the tube exceeds the thermal capacity due to the reduction of the thermal dissipation coefficient of the tube.
The target predicted thermal capacity may be the predicted value of the thermal capacity of the tube at the preset time point. In some embodiments, the prediction module 230 may determine the target predicted thermal capacity of the tube by using a first target prediction model. For more information about the first target prediction model, please refer to FIGs. 4, 5, 6, 7 and their related descriptions.
In order to gradually improve the prediction accuracy of the first target prediction model, the first target prediction model may be trained in the process of the practical imaging.
In some embodiments, the information acquisition module 220 may obtain a measured thermal capacity of the tube at the preset time point through the sensor during the practical imaging, and use the measured thermal capacity at the preset time point and the target predicted thermal capacity at the preset time point as the second training data for the update of the model parameters. For example, during the practical imaging of the first target prediction model, the model acquisition module 210 may update the parameters of the first target prediction model directly based on the measured thermal capacity at the preset time point and the target predicted thermal capacity at the preset time point, and use the updated first target prediction model for the next thermal capacity prediction. The next prediction may be performed for a tube of a medical imaging device that is the same as or different from the tube predicted currently or previously. For more information about updating the parameters of the first target prediction model based on the measured thermal capacity at the preset time point and the target predicted thermal capacity at the preset time point, please refer to FIG. 5 and related descriptions thereof.
In some embodiments, the information acquisition module 220 may obtain the current thermal capacity of the tube at the start of the working of the tube and the measured thermal capacity of the tube at the preset time point through the sensor during the practical imaging, and use the current thermal capacity of the tube at the start of the working of the tube, the scanning parameters, the property information, and the measured thermal capacity of the tube at the preset time point as the second training data for the update of the model parameters. For example, during the practical use of the first target prediction model, the model acquisition module 210 may update the parameters of the first target prediction model directly based on the current thermal capacity of the tube at the start of the working, the scanning parameters, the property information, and the measurement thermal capacity of the tube at the preset time point, and use the updated first target prediction model for the next thermal capacity prediction. The next prediction may be performed for a tube of a medical imaging device that is the same as or different from the tube predicted currently or previously. For more information about updating the parameters of the first target prediction model, please refer to FIG. 5 and its related description.
In some embodiments, the thermal capacity difference between the target predicted thermal capacity and the measured thermal capacity may be used to: determine whether the parameters of the first target first prediction model need to be updated, determine whether the target predicted thermal capacity and the measured thermal capacity can be used as the second training data, or determine whether the current thermal capacity, the scanning parameters, the property information, and the measured thermal capacity of the tube at the preset time point when the tube starts working can be used as the second training data. For example, if the thermal capacity difference between the target predicted thermal capacity and the measured thermal capacity is greater than a thermal capacity difference threshold, the model acquisition module 210 may update the parameters of the first target prediction model based on the target predicted thermal capacity and the measured thermal capacity, or may update the parameters of the first target prediction model based on the current thermal capacity (measured at the start of the working of the tube) , the scanning parameters, the property information, and the measured thermal capacity of the tube at the preset time point. If the thermal capacity difference between the target predicted thermal capacity and the measured thermal capacity is less than the thermal capacity difference threshold, the parameters of the first target prediction model may be not updated, or it may be selected not to update the parameters of the first target prediction model.
According to some embodiments of this present disclosure, the accuracy of the thermal capacity prediction may be improved by predicting the thermal capacity using a machine learning model. Considering the property information of the tube in the prediction makes the prediction performed based on the practical situation of each tube itself more practical.
In some embodiments, subsequent operations may also be determined based on the relationship between the target predicted thermal capacity and the thermal capacity threshold output by the first target prediction model during the practical use.
In some embodiments, the sending module may determine whether the target predicted thermal capacity is greater than the thermal capacity threshold. In response to the target predicted thermal capacity being greater than the thermal capacity threshold, the sending module may send an instruction to the medical imaging device to stop scanning or send alert information to the user terminal. For example, if the current scanning has started, i.e., the current thermal capacity is the thermal capacity at a certain time point in the scanning process, and if the target predicted thermal capacity at the preset time point is greater than the thermal capacity threshold, the sending module may send an instruction to stop scanning before the preset time point, or send alert information to the user terminal (e.g., terminal 130) before the preset time point (e.g., the alert information may be "the predicted thermal capacity at the preset time point will exceed the thermal capacity threshold" to avoid accidents in which the thermal capacity exceeds the threshold during the scanning process and results in damage to the tube or interruption of scanning. The thermal capacity threshold may be a system default value, an empirical value, a human preset value, or any combination thereof, and may be set according to practical needs, which is not limited in this present disclosure.
It should be noted that the thermal capacity threshold of the tube may change with the use of the medical imaging device. In some embodiments, the thermal capacity threshold may change for different tube materials or qualities and for tubes that have undergone different numbers of maintenance. Accordingly, the thermal capacity threshold at the preset time point may be predicted to determine a more accurate thermal capacity threshold.
In some embodiments, the sending module may determine a remaining life of the tube corresponding to a preset exposure time range based on a life prediction model, and in response to the remaining life being within a preset life range, generate the thermal capacity threshold based on the preset exposure time range.
The preset exposure time range may refer to a range of the maximum exposure time allowed by the tube under a certain tube current and a tube voltage condition. When the exposure time of the tube exceeds the predicted exposure time range, the focal point surface of the tube may be damaged by overheating due to heat accumulation. In some embodiments, the preset exposure time range may be a system default value, an empirical value, a manually preset value, or any combination thereof, or may be set according to practical needs, which is not limited in this present disclosure.
The life prediction model may be a machine learning model for determining the remaining life of the tube corresponding to the preset exposure time range. For example, the life prediction model may include a CNN model, a DNN model, etc., or any combination thereof.
In some embodiments, the life prediction model may be configured to input a usage duration of the medical imaging device, a usage duration of the tube, a tube material, a count of times the tube has been maintained, a preset exposure time range, and may be configured to output a remaining life of the tube corresponding to the preset exposure time range.
The usage duration of the medical imaging device may refer to the working duration of the medical imaging device since it was shipped to date. For example, the usage duration of the medical imaging device may refer to the count of seconds that the medical imaging device works.
The usage duration of the tube may refer to the working duration of the original or replacement tube since it was shipped to date. For example, the usage duration of the tube may refer to the count of seconds the tube works. The original tube may refer to the tube that comes standard with the medical imaging device when it leaves the factory. The replacement tube may refer to a tube replaced later.
In some embodiments, the tube material may include at least one of a metal-ceramic tube, a liquid metal bearing tube, a glass tube, etc.
In some embodiments, the life prediction model may be obtained by training based on a large count of third training samples with third labels. For example, multiple third training samples with third labels may be input to the initial life prediction model, a loss function may be determined based on the results of the initial life prediction model and the third labels, and the parameters of the initial life prediction model may be updated iteratively based on the loss function. When the loss function of the initial life prediction model meets a preset condition, the model training may be completed and the trained life prediction model may be obtained. The preset condition may be that the loss function converges, the count of iterations reaches a threshold, or the like.
In some embodiments, the third training samples may include a usage duration of a sample medical imaging device, a usage duration of a third sample tube, a tube material of the third sample tube, a maintenance frequency of the third sample tube, and a preset exposure time range of the third sample tube. The third label may be the practical remaining life of the sample tube. The first sample tube, the second sample tube and the third sample tube may be the same or different.
In some embodiments, the multiple sets of third training samples and third labels may be extracted from the historical usage data of the faulty or failed tube. The historical usage data of the faulty or failed tube may include a historical total usage duration of the medical imaging device, a historical total usage duration of the tube, a tube material, a historical total maintenance frequency of the tube, a historical total exposure time of the tube, a historical total life of the tube, etc. Each set of third training samples corresponds to the historical usage frequency of the medical imaging device, the historical usage frequency of the tube, the tube material, the maintenance frequency of the tube, and the historical total exposure time of the tube. In some embodiments, the historical usage data of the faulty or failed tube may be divided by multiple historical time periods. Each set of third training samples includes the historical usage frequency of the medical imaging device, the historical usage frequency of the tube, the tube material, the historical maintenance frequency of the tube, the historical exposure time of the tube, and the historical life of the tube. The third label corresponding to each set of third training samples is the historical practical remaining life of the tube. In some embodiments, the third label corresponding to each set of third training samples may be obtained by subtracting the total historical life of the tube from the historical life of the tube in the set of third training samples.
In some embodiments, it may be further determined whether the remaining life of the tube is within the preset life range. In response to the remaining life being within the preset life range, the thermal capacity threshold may be generated based on the preset exposure time range.
The preset life range may be a preset range of the remaining life of the tube. The preset life range may be a system default value, an empirical value, a manually preset value, or any combination thereof, and may be set according to practical needs, which is not limited in this present disclosure.
In some embodiments, the sending module may determine, based on the preset exposure time range, the current thermal capacity of the tube, the current tube voltage, and the current tube current, a thermal capacity range of the tube at the current moment as the thermal capacity threshold of the tube. Accordingly, the thermal capacity threshold of the tube may be a value range. An exemplary determination of the thermal capacity threshold may be shown as follows:
K=K₁+U×I×S, (1)
where K is the thermal capacity threshold and K₁ is the current thermal capacity of the tube, U is the tube voltage (kV) , I is the tube current (mA) , and S is any exposure time within the preset exposure time range.
In some embodiments of this present disclosure, the life prediction model can be configured to accurately and efficiently predict the remaining life of the tube, facilitating subsequent prediction of the changing thermal capacity threshold based on the remaining life of the tube to determine a more accurate thermal capacity threshold. At the same time, when determining the practical thermal capacity threshold of the tube, the tube material, the usage frequency and the maintenance frequency may be taken into account, which can make the results more accurate and provide a more accurate reference for the subsequent judgment of whether the target predicted thermal capacity exceeds the thermal capacity threshold.
As mentioned above, the thermal dissipation parameter of the tube may change, which may be caused by the loss of the tube as the use of the medical imaging device and the increment of the usage duration of the tube during practical use.
The current thermal dissipation parameter of the tube may refer to the thermal dissipation parameter of the tube in its current state. In some embodiments, the information acquisition module 220 may determine the current thermal dissipation parameter of the tube based on the thermal dissipation prediction model. In some embodiments, the information acquisition module 220 may determine the current thermal dissipation parameter of the tube based on a thermal dissipation prediction model that processes the usage data of the medical imaging device, the usage duration of the tube, and the initial thermal dissipation parameter of the tube.
The thermal dissipation prediction model may be a machine learning model for determining the current thermal dissipation parameter of the tube. For example, the thermal dissipation prediction model may include a CNN model, a DNN model, etc., or any combination thereof. In some embodiments, the input of the thermal dissipation prediction model may include usage data of the medical imaging device, a usage duration of the tube, an initial thermal dissipation parameter of the tube, and the output of the thermal dissipation prediction model may include a current thermal dissipation parameter of the tube.
In some embodiments, the usage data of the medical imaging device may include a usage duration of the medical imaging device, usage environment information, etc. In some embodiments, the usage environment information may include a temperature, a humidity, a barometric pressure, etc.
In some embodiments, the usage duration of the tube may refer to the usage duration of the original tube or the replacement tube since it was shipped from the factory to date. In some embodiments, the usage duration of the tube may refer to the usage duration of the tube so far during the current imaging.
In some embodiments, the initial thermal dissipation parameter of the tube may refer to the thermal dissipation parameter of the tube when it is shipped from the factory. In some embodiments, the initial thermal dissipation parameter of the tube may refer to the initial thermal dissipation parameter of the tube during the current imaging. The initial thermal dissipation parameter of the tube may be different for different practical use.
In some embodiments, the thermal dissipation prediction model may be obtained by training based on a large count of fifth training samples with fifth labels. For example, the multiple fifth training samples with the fifth labels may be input into the initial thermal dissipation prediction model, and a loss function may be determined based on the results of the initial thermal dissipation prediction model and the fifth labels, and the parameters of the initial thermal dissipation prediction model may be updated iteratively based on the loss function. When the loss function of the initial thermal dissipation prediction model meets a preset condition, the model training may be completed and the trained thermal dissipation prediction model may be obtained. The preset condition may be that the loss function converges, the count of iterations reaches a threshold, or the like.
In some embodiments, the fifth training samples may include multiple sets of sample usage data of different sample medical imaging devices, usage duration of sample tubes of the different sample medical imaging devices, and initial thermal dissipation parameters. In some embodiments, the fifth training samples may be determined by multiple different sets of historical usage data of the sample medical imaging devices. For example, the historical usage data of each sample medical imaging device, the usage duration of the sample tube of that sample medical imaging device, and the initial thermal dissipation parameter may be used as a set of fifth training samples.
In some embodiments, the fifth label may be the practical thermal dissipation parameter of the sample tube corresponding to each set of fifth training samples. In some embodiments, the fifth label may be that an practical thermal dissipation coefficient of the sample tube is obtained by measuring the thermal capacity of the sample tube at a certain scanning time point and measuring the thermal capacity of the sample tube at a cooling time point while each set of medical imaging devices are operating, based on the thermal capacity difference between the thermal capacity at the scanning time point and the thermal capacity at the cooling time point divided by the time difference between the preset time point and the scanning time point. The scanning time point is a time point in the scanning process of the tube, and the cooling time point is the time point when the tube cools down completely after the work is finished. In some embodiments, the cooling time point may be obtained based on an anodic cooling curve of the tube. For example, the horizontal axis of the anode cooling curve is time and the vertical axis is thermal capacity. The anode cooling curve may include the time change required to cool from 100%thermal capacity to 0%thermal capacity, and the cooling time point may be determined from the anode cooling curve based on the thermal capacity at the end of the working of the tube and the end time point.
In some embodiments of this present disclosure, the accuracy of the thermal capacity prediction may be effectively improved by predicting the current thermal dissipation parameter before the start of the scanning in order to predict the target predicted thermal capacity of the tube at the preset time point of this scanning based on the current thermal dissipation parameter as well as other parameters to prevent thermal capacity overrun.
FIG. 4 is a flowchart illustrating an exemplary process for obtaining a first target prediction model according to some embodiments of the present disclosure. In some embodiments, process 400 may be performed by a processing device (e.g., processing device 110) . For example, process 400 may be stored in a storage device in the form of a program or instructions, and process 400 may be implemented when the server or the module shown in FIG. 2 executes the program or instructions. In some embodiments, process 400 may be performed using one or more additional operations not described below, and/or performed without one or more of the operations discussed below. As shown in FIG. 4, process 400 may include one or more of the following operations.
In 410, a second initial prediction model may be obtained. In some embodiments, operation 410 may be performed by the model acquisition module 210.
The second initial prediction model may be a machine learning model. The second initial prediction model is a model that has not yet been trained or is not yet trained. In some embodiments, the parameters of the second initial prediction model may be obtained by initialization.
In some embodiments, similar to the structure of the first target prediction model, the second initial prediction model may include multiple layers, e.g., an input layer, an incremental prediction layer, a fusion layer, and an output layer. In some embodiments, the structure of the second prediction model may differ from the structure of the first target prediction model. For example, the input layer of the second initial prediction model may have two nodes to receive two inputs, e.g., the current thermal capacity and the scanning parameters. In some embodiments, the second initial prediction model may receive two inputs, e.g., the current thermal capacity and the scanning parameters, and output the target predicted thermal capacity.
In some embodiments, the second initial prediction model may have the same structure as the first target prediction model. For example, the second initial prediction model may include three input nodes to receive three inputs, e.g., the property information, the current thermal capacity and the scanning parameters of the tube. In some embodiments, the second initial prediction model may receive three inputs, e.g., the property information, the current thermal capacity, and the scanning parameters of the tube, and output the target predicted thermal capacity.
For more information about the structure of the first target prediction model, please refer to FIG. 6 and related descriptions thereof.
It should be noted that “three inputs” or “two inputs” in this present disclosure is not a limit on the count of inputs, but is used to indicate the type of data to be input. Similarly, “three input nodes” and “two input nodes” in this present disclosure are not limitations on the count of nodes, but rather on the correspondence with the input data types, stating that there are input nodes for each data type, but the exact number is not limited in this present disclosure.
In 420, first training data may be obtained, the first training data including a first training sample and a first label corresponding to the first training sample. In some embodiments, operation 420 may be performed by the model acquisition module 210.
The first training data may include one or more sets of data. In some embodiments, the first training sample may include the sample current thermal capacity of the first sample tube and the sample scanning parameters. That is, the second initial prediction model may be trained based on the sample current thermal capacity of the first sample tube and the sample scanning parameters.
In some embodiments, the first training sample may include the sample current thermal capacity of the first sample tube, the sample scanning parameters, and the sample property information. That is, the second initial prediction model may be trained based on the sample current thermal capacity of the first sample tube, the sample property information, and the sample scanning parameters.
In some embodiments, if the first training sample includes the sample current thermal capacity of the first sample tube, the sample scanning parameters, and the sample property information of the first sample tube, the sample property information of the first sample tube may be set to a preset value. For example, the preset value may be 0. Accordingly, the second initial prediction model may be practically trained only based on the sample current thermal capacity and the sample scanning parameters.
As mentioned above, the input nodes of the second initial prediction model may receive two inputs or three inputs. In some embodiments, if the second initial prediction model has a different structure from the first target prediction model, e.g., the input nodes of the second initial prediction model receive two inputs, then the first training sample may include the sample current thermal capacity of the first sample tube and the sample scanning parameters. If the second initial prediction model has the same structure as the first target prediction model, e.g., the input nodes of the second initial prediction model receive three inputs, then the first training sample may include the sample current thermal capacity of the first sample tube, the sample scanning parameters, and the sample property information of the first sample tube. The sample property information of the first sample tube may be a preset value (e.g., 0) , or may be the practical property information of the first sample tube.
The first label corresponding to the first training sample is the measured thermal capacity of the first sample tube at the first sample preset time point. The sample current thermal capacity of the first sample tube is the thermal capacity at a time point prior to the first sample preset time point, and similar to the current thermal capacity, the sample current thermal capacity may be the thermal capacity of the first sample tube under different circumstances. The first sample tube may be the same or different for different first training samples.
In some embodiments, the first training data may be stored in a storage device, and the model acquisition module 210 may obtain the first training data from the storage device. The first training data may be obtained based on a historical scanning, in which the first label of the first training sample may be obtained by the sensor during the historical scanning.
In 430, a second target prediction model may be determined by updating the second initial prediction model based on the first training data. In some embodiments, operation 430 may be performed by the model acquisition module 210.
The parameters of the second target prediction model are determined after training. The second target prediction model has the same structure as the second initial prediction model. It can be understood that, similar to the second initial prediction model, the structure of the second target prediction model may be the same as or different from the structure of the first target prediction model, and the structure of the second target prediction model being determined by the second initial prediction model.
In some embodiments, the model acquisition module 210 may input a first training sample with a first label into the second initial prediction model, determine a loss function based on the output of the second initial prediction model (i.e., the target predicted thermal capacity) and the first label, and iteratively update the parameters of the second initial prediction model based on the loss function. When the trained model meets a preset condition, the training ends and the second target prediction model is determined. The preset condition may be that the loss function converges, the count of iterations reaches a threshold, etc.
In 440, the first target prediction model may be determined based on the second target prediction model. In some embodiments, operation 440 may be performed by the model acquisition module 210.
In some embodiments, the first target prediction model and the second target prediction model may have the same structure, and the model acquisition module 210 may directly use the second target prediction model as the first target prediction model. In some embodiments, if the first target prediction model has a structure different from the second target prediction model, the model acquisition module 210 may migrate the parameters of the second target prediction model to the first initial prediction model, and use the first initial prediction model after parameter migration as the first target prediction model.
The first initial prediction model may be a model that has not been trained or has not been trained yet. The first initial prediction model may have the same structure as the first target prediction model. The parameters of the first initial prediction model may be obtained by initialization.
In some embodiments, the model acquisition module 210 may determine a first intermediate prediction model based on the second target prediction model, and further determine a first target prediction model based on the first intermediate prediction model.
The first intermediate prediction model may be a model determined by updating the parameters of the first initial prediction model based on the first training data. The first intermediate prediction model may not undergone an augmented training.
In some embodiments, the model acquisition module 210 may directly use the first intermediate prediction model as the first target prediction model for the thermal capacity prediction. In other words, the first target prediction model applied at this point has only undergone an initial training. In some embodiments, the model acquisition module 210 may update the first intermediate prediction model based on the second training data (i.e., enhanced training) and use the updated first intermediate prediction model as the first target prediction model for the thermal capacity prediction. In other words, the first target prediction model applied at this point underwent an initial training and augmented training. For more information about the augmented training, please refer to FIG. 5 and related descriptions thereof.
As mentioned above, the first target prediction model may have the same structure as the second target prediction model. In some embodiments, the model acquisition module 210 may directly use the second target prediction model obtained by training based on the first training data as the first intermediate prediction model.
As mentioned above, the first target prediction model may be structured differently from the second target prediction model. In some embodiments, the model acquisition module 210 may migrate the parameters of the second target prediction model to the first initial prediction model, using the first initial prediction model after parameter migration as the first intermediate prediction model. In some embodiments, the model acquisition module 210 may migrate all or some of the parameters of the second target prediction model to the first initial prediction model. For example, the model acquisition module 210 may update the parameters of the incremental prediction layer, the fusion layer, and the output layer of the first initial prediction model only based on migration. After migration, some or all of the initialized parameters in the first initial prediction model may be updated with the trained parameters.
The structure of the first target prediction model and the second target prediction model may be different. By migrating the parameters of the second target prediction model, some of the parameters in the first initial prediction model may be updated, and the other parameters may have the initialized values. For example, in the first target prediction model, the parameters of the input nodes corresponding to the property information of the tube may have the initialized values, and the other parameters may have updated parameters. On this basis, the model acquisition module 210 may update the parameters of the input nodes corresponding to the property information of the tube based on the first target prediction model in practical use by the augmented training.
The property information of the tube may be closely related to the medical imaging equipment itself, and in the practical use, the property information of the tube may differ from one medical imaging device to another. In some embodiments of this present disclosure, the initial training may be performed based only on the sample current thermal capacity of the sample tube and the sample scanning parameters, without considering the sample property information of the sample tube, so that the first target prediction model obtained after the initial training may be able to substantially achieve the thermal capacity prediction. In practical use, the parameters of the first target prediction model may be updated based on the prediction results and the practical detected thermal capacity values before the augmented training, and the prediction accuracy of the first target prediction model may be improved so that it can make predictions based on the property information of the tubes of each device. It can be understood that the first training data may be obtained without considering the property information of the tube of the medical imaging device, which can reduce the cost of obtaining the training data and ensure the accuracy of the model prediction at the same time.
In some embodiments, the model acquisition module 210 may determine a first target prediction model by training the first initial prediction model directly based on the first training data. The first training sample includes the sample current thermal capacity, the sample scanning parameters and the sample property information of the first sample tube. In some embodiments, the sample property information of the first sample tube may be the practical property information. The first label of the first training sample may be the same as that used in the training of the second initial prediction model. The model acquisition module 210 may determine a loss function based on the first label and the output of the first initial prediction model to update the parameters of the first initial prediction model.
In some embodiments, after the model acquisition module 210 obtains the first intermediate prediction model, it may directly use the first intermediate prediction model as the first target prediction model, or it may determine the first target prediction model after updating the parameters of the first intermediate prediction model based on the second training data (i.e., augmented training) .
It should be noted that the above description of the process 400 is for example and illustration purposes only and does not limit the scope of application of this present disclosure. For those skilled in the art, various modifications and changes may be made to the process 400 under the guidance of the present disclosure. However, these modifications and changes are still within the scope of the present disclosure.
FIG. 5 is another flowchart illustrating an exemplary process for obtaining a first target prediction model according to some embodiments of the present disclosure. In some embodiments, process 500 may be executed by a processor (e.g., processing device 110) . For example, process 500 may be stored in a storage device in the form of a program or instructions, and process 500 may be implemented when the server or the module shown in FIG. 2 executes the program or instructions. In some embodiments, process 500 may be performed using one or more additional operations not described below, and/or performed without one or more of the operations discussed below. As shown in FIG. 5, process 500 may include the following operations:
In 510, second training data may be obtained, the second training data including a second training sample and a second label corresponding to the second training sample. In some embodiments, operation 510 may be performed by the model acquisition module 210.
In some embodiments, the second training sample may include a sample target predicted thermal capacity of a second sample tube at a second sample preset time point. The second training sample may be obtained in the course of practical use. For example, the target predicted thermal capacity of the tube determined by using the first target prediction model in a practical use may be used as a second training sample.
In some embodiments, the second training sample may include a sample current thermal capacity of the second sample tube, sample scanning parameters, and sample property information. The second training sample may be obtained in the course of practical use. For example, real data of the second sample tube in a practical use (including the current thermal capacity, scanning parameters and property information of the second sample tube at a previous time point) may be used as the second training sample.
In some embodiments, the second label may be a measurement thermal capacity of the second sample tube at the preset time point of the second sample. When the second training sample is obtained at a previous time point, the second sample is preset at a time point after the previous time point. For example, the second sample preset time point may be the time point when scanning or radiation ends. For the current timeline, a certain previous time point and the second sample preset time point are both historical. The second label may be obtained during the practical use. For example, the measured thermal capacity of the device feedback at the end of the discharge is used as the second label. As an example, the measured thermal capacity at the end of the discharge may be obtained by the sensor as the second label.
The second sample tube may be the same or different for different second training samples. The first sample tube and the second sample tube may be the same or different.
In 520, a first intermediate prediction model may be determined based on the second target prediction model. In some embodiments, operation 520 may be performed by the model acquisition module 210.
For more information about the first intermediate prediction model, and the determination of the first intermediate prediction model based on the second target prediction model, please refer to step 440 and related descriptions thereof.
In 530, the first target prediction model may be obtained by updating the first intermediate prediction model based on the second training data. In some embodiments, operation 530 may be performed by the model acquisition module 210.
In some embodiments, if the second training sample includes a target predicted thermal capacity of the second sample tube at the second sample preset time point, the model acquisition module 210 may compare the target predicted thermal capacity of the second sample tube at the second sample preset time point with the measured thermal capacity of the tube at that preset time point, update the parameters of the first intermediate prediction model, and use the first intermediate prediction model with the updated parameters as the first target prediction model. For example, the model acquisition module 210 may determine a loss function based on the target predicted thermal capacity of the second sample tube at the second sample preset time point and the second label, and update the parameters of the first intermediate prediction model based on the loss function.
In some embodiments, if the second training sample includes a sample current thermal capacity of the second sample tube, sample scanning parameters, and sample property information, the model acquisition module 210 may input the second training sample into the first intermediate prediction model, update the parameters of the first intermediate prediction model based on the second training data, and use the first intermediate prediction model with the updated parameters as the first target prediction model. For example, the model acquisition module 210 may input the second training sample into the first intermediate prediction model, determine a loss function based on the output of the first intermediate prediction model and the second label, and update the parameters of the first intermediate prediction model based on the loss function. The intermediate prediction thermal capacity output by the first intermediate prediction model may refer to the output of the first intermediate prediction model in the updating process (performed based on the second training data) of the first intermediate prediction model.
In some embodiments, the model acquisition module 210 may determine an evaluation result of the first intermediate prediction model by processing an intermediate predicted thermal capacity output by the evaluation model, a usage duration of the tube, a device type of the medical imaging device, and a count of updates of the first intermediate prediction model; in response to the evaluation result meeting a preset evaluation condition, stop updating the first intermediate prediction model and obtain the first target prediction model. The device type of the medical imaging device may be a CT imaging device, a PET-CT imaging device, a CT-RT device, and the like. In a specific embodiment, when the medical imaging device is a CT imaging device, the device type may be a spiral CT device, a wide cone CT device, a dual source CT device, a dual energy CT device, etc. It should be noted that different types of devices have different losses to the tube, and the parameters of the tube vary, which indirectly affects the accuracy of the evaluation model.
The count of updates of the first intermediate prediction model may refer to the count of updates of the parameters of the first intermediate prediction model that are updated based on the second training data.
The evaluation result of the first intermediate prediction model may reflect the accuracy of the intermediate predicted thermal capacity output by the first intermediate prediction model. The higher the accuracy is, the closer the intermediate predicted thermal capacity is to the measured thermal capacity. In some embodiments, the evaluation result may be expressed as a real number between 0 and 1, in which a relatively small value may indicate a relatively high accuracy of the intermediate predicted thermal capacity output by the first intermediate prediction model.
In some embodiments, the evaluation model may be obtained by joint training with the first intermediate prediction model in an augmented training process.
In some embodiments, the first intermediate prediction model and the initial evaluation model may be trained based on a large count of fourth training samples with fourth labels. A fourth training sample may include a second training sample used to train the sample first intermediate prediction model (including the sample current thermal capacity of the second sample tube, the sample scanning parameters, and the sample property information) , the sample usage duration of the second sample tube, the device type of the sample medical imaging device, and the count of updates of the sample first intermediate prediction model. A fourth label may be an evaluation result of the sample first intermediate prediction model. The fourth training sample and the fourth label may be determined in historical practical use. For example, the fourth label may be determined by using the difference between the predicted result of the first intermediate prediction model and the historical measured thermal capacity during the historical practical use of the second sample tube. For example, a ratio of an absolute value of the difference to the historical measured thermal capacity is used as the fourth label.
An exemplary training process may include: inputting the second training sample into the first intermediate prediction model to obtain the intermediate predicted thermal capacity output by the first intermediate prediction model; inputting the intermediate predicted thermal capacity (used as training data) , the sample usage duration of the second sample tube, the device type of the sample medical imaging device, and the count of updates of the sample first intermediate prediction model into the initial evaluation model, to obtain the evaluation result output by the initial evaluation model; based on the evaluation result output by the initial evaluation model and the fourth label, determining a loss function to update the parameters of the first intermediate prediction model and the initial evaluation model simultaneously. The trained evaluation model is obtained by updating the parameters.
The preset evaluation condition may refer to a judgment condition related to the evaluation result. For example, the preset evaluation condition may be that the evaluation result is less than an evaluation threshold. The evaluation threshold may be a system default value, an empirical value, a manually pre-set value, or any combination thereof, and may be set according to practical needs, which is not limited in this present disclosure.
In some embodiments of this present disclosure, by determining the accuracy of the first intermediate prediction model, when the first intermediate prediction model meets the preset evaluation condition, no further augmented training of the model is performed, thereby reducing computational effort and improving data processing efficiency.
FIG. 6 is a schematic diagram illustrating an exemplary structure of a first target prediction model according to some embodiments of the present disclosure.
As shown in FIG. 6, in some embodiments, the first target prediction model includes an input layer 610, an incremental prediction layer 620, a fusion layer 630, and an output layer 640.
The input layer 610 may be configured to input the property information, the current thermal capacity and the scanning parameters of the tube. The input layer includes input nodes that receive the property information, the current thermal capacity and the scanning parameters of the tube.
The incremental prediction layer 620 may determine a thermal capacity increment of at least one position of the tube at the preset time point based on the property information and working status information (i.e., scanning parameters and current thermal capacity) . The thermal capacity increment is the amount of thermal capacity increased between the current time and the preset time point.
As described in FIG. 1, the tube includes multiple components. In some embodiments, the at least one position of the tube may be a position of a different part of the tube or a different position point on a different part. In some embodiments, thermal capacity increments at different positions of the tube may be different. For example, the thermal capacity increment of a filament is greater than the thermal capacity increment of a buffer tube. For example, the at least one position includes n positions, and the thermal capacity increments of the n positions are thermal capacity increment 1, thermal capacity increment 2, . . . . . ., and thermal capacity increment n.
The fusion layer 630 may determine an initial predicted thermal capacity of the at least one position at the preset time point based on the current thermal capacity and the thermal capacity increment of the at least one position of the tube at the preset time point.
In some embodiments, for a target position of the at least one position, the fusion layer 630 may fuse the thermal capacity increment of that target position, the thermal capacity increments of the other positions of the at least one position, and the current thermal capacity of the tube to determine the initial predicted thermal capacity of the target position at the preset time point. The other positions of the at least one position means one or more other positions of the at least one position excluding the target position. It can be understood that there may be differences in the thermal capacity increments at different positions of the tube, and the changes in thermal capacity between the different positions may be influenced by each other.
In some embodiments, for a target position of the at least one position, other positions that are fused with the target position may be determined by various means. Exemplarily, one or more positions that are closer to the target position (e.g., less than a threshold distance) may be used as the other positions for fusion. As another example, the other positions may be preset.
In some embodiments, the fusion operations may be performed in one or more ways. For example, averaging, weighting, taking the median, taking the maximum, etc. For example, when weighting based on a first weight, the first weight may be determined based on an influence factor of the other positions on the target position. If the influence factor is relatively large, the corresponding first weight may be relatively large.
In some embodiments, the influence factor may be related to a distance of other positions from the target position. For example, the further away from the target position the other positions are, the smaller the influence factor is. In some embodiments, the influence factor may also be related to a thermal capacity change rate at the other positions. For example, other positions with larger thermal capacity change rate correspond to smaller influence factor. Conversely, the corresponding influence factor is larger.
In some embodiments, the first weight of each of the other positions may be determined based on a ratio of its influence factor to a sum of the influence factors of all other positions.
In some embodiments, the first weight may be determined by the fusion layer 630.
In some embodiments, the fusion layer 630 may determine the first weight of each of the at least one position based on the thermal capacity change rate of each of the at least one position, a loss degree of each of the at least one position, and a distance of one of the at least one position from the other positions in the at least one position. The thermal capacity change rate of each position may reflect a thermal capacity change at the target position over a period of time. The thermal capacity change rate may be different for different positions. In some embodiments, the thermal capacity change rate of each position may be determined based on the amount of thermal capacity change in each of the at least one position over a preset time period divided by the preset time period. For example, if position 1 has a thermal capacity change of 300 KHU over a preset time period (e.g., 2 mins) , the thermal capacity change rate of position 1 may be determined to be 300/2=150 KHU/min.
The loss degree of each position may be expressed by a real number from 0 to 1. The higher the value is, the higher the loss degree is. In some embodiments, the loss degree of each position may be determined based on a preset rule, based on the usage duration of the tube and the count of maintenance of the tube. An exemplary preset rule may include: if the usage duration of the tube is in the duration range 1, the corresponding loss degree may be 0.1; if the usage duration of the tube is in the duration range 2, the corresponding loss degree may be 0.2, etc. Another exemplary preset rule may include: if the count of maintenances of the tube is in the count range 1, the corresponding loss degree may be 0.1; if the count of maintenances of the tube is in the count range 2, the corresponding loss degree may be 0.2. The division of the duration range and count range and their corresponding loss degrees may be preset manually.
In some embodiments, the loss degree of each position may be determined based on an image similarity between a current image of the tube and an initial image of the tube. The current image of the tube may refer to a currently captured image. The initial image of the tube may refer to an image taken at the time the tube was shipped from the factory. In some embodiments, the current image of the tube and the initial image of the tube may be processed by a similarity model to determine the image similarity. The loss degree of the each position may be determined according to a similarity range in which the image similarity is located and its corresponding loss degree. The similarity model may be a machine learning model configured to determine the image similarity. For example, the similarity model may be a GNN model. The similarity range and its corresponding loss degree may be preset manually.
In some embodiments, the first training sample for the initial training may further include a thermal capacity change rate at each of the at least one position of the first sample tube, a loss degree at each of the at least one position, a distance between one of the at least one position and the other positions of the at least one position. The second training sample for the augmented training may also include a thermal capacity change rate at each of the at least one position of the second sample tube, a loss degree at each of the at least one position, and a distance between one of the at least one position and the other positions of the at least one position.
In some embodiments, the fusion layer 630 may fuse the thermal capacity increment of the position, the thermal capacity increments of the other positions in the at least one position, and the current thermal capacity of the tube, based on the first weight of each position in the at least one position, to determine an initial predicted thermal capacity of the position at the preset time point.
Exemplarily, for a target position i of the at least one position, the initial predicted thermal capacity of the target position may be determined as follows:
y_i=m₁*x₁+m₂*x₂+…+m_j*x_j+m_i*x_i+k*C, (2)
where, y_i is the initial predicted thermal capacity of the target position i, x₁, x₂, …, x_j are the thermal capacity increments of other positions, x_i is the thermal capacity increment of the target position i, C is the current thermal capacity of the tube, m₁, m₂, …, m_j、 m_i are the first weights of x₁, x₂, …, x_j, x_i, respectively, and k is the coefficient. In some embodiments, m₁, m₂, …, m_j, m_i and k can be preset in advance or can be obtained in the aforementioned way.
As another example, the at least one position includes position 1, position 2 and position 3, and the initial predicted thermal capacities of the position 1, position 2 and position 3 are y1, y2 and y3, respectively, in which there are no other positions fused at position 1, other positions fused at position 2 include position 1, and other positions fused at position 3 include position 1 and position 2. The initial predicted thermal capacities of positions 1, 2 and 3 can be determined as follows:
y1=t₁*x₁+t₂*C, (3)
y2=t₃*y₁+t₄*x₂+t₅*C, (4)
y3=t₆*y1+t₇*y2+t₈*x₃+t₉*C, …… (5)
where, y1, y2, y3 are the initial predicted thermal capacities at positions 1, 2, and 3, respectively; x₁, x₂, x₂ are the thermal capacity increments at positions 1, 2, and 3, respectively; C is the current thermal capacity; and t₁, t₂, t₃, t₄, t₅, t₆, t₇, t₈, t₉ are the parameters obtained from model training.
It can be understood that there are other ways of computing the initial predicted thermal capacity after fusion, and the above equations are only examples for illustration.
In some embodiments, for a target position of the at least one position, the fusion layer 630 may fuse the thermal capacity increment of the position directly with the current thermal capacity to determine the initial predicted thermal capacity of the position at the preset time point. For example, for a position a, the fusion layer 630 may fuse the thermal capacity increment of position a with the current thermal capacity to determine the initial predicted thermal capacity of position a at a preset time.
The output layer 640 may determine a target predicted thermal capacity based on the initial predicted thermal capacity of at least one position at a preset time point. For example, the output layer 640 may fuse the initial predicted thermal capacity at the at least one position and use the fused value as the target predicted thermal capacity. Exemplary fusion algorithms may include averaging, weighted averaging, etc. As another example, the output layer 640 may select a maximum of multiple initial predicted thermal capacities as the target predicted thermal capacity of the tube at the preset time point.
In some embodiments, if the initial predicted thermal capacity of the at least one position is fused by means of weighted summation, the second weight may be determined in a variety of ways. For example, the second weight of each position may be determined based on the thermal capacity change rate of the at least one position. The position with a relatively large thermal capacity change rate may correspond to a relatively small second weight. Conversely, the corresponding second weight may be relatively large. As another example, the second weight of each position may be determined based on the loss degree of the at least one position. The position with a relatively high loss degree may correspond to a relatively small second weight. Conversely, the corresponding second weight may be relatively large.
In some embodiments, the second weight may be determined by the output layer 640.
In some embodiments, the thermal capacity change rate at each of the at least one position, the loss degree of each of the at least one position, may be used as an input of the output layer 640 to determine the second weight for each of the at least one position.
In some embodiments, the output layer 640 may fuse the initial predicted thermal capacity of each position based on the second weight of each of the at least one position to determine a target predicted thermal capacity of one of the at least one position at the preset time point.
In some embodiments, the model type of one or more layers of the first target prediction model may include a neural network model, e.g., CNN, DNN, etc. For example, the incremental prediction layer 620 may be a CNN network.
FIG. 7 is a schematic diagram illustrating an exemplary training process of a first target prediction model according to some embodiments of the present disclosure. In some embodiments, the training of the first target prediction model may be performed by the processing device 110 (e.g., model acquisition module 210) .
The training of the first target prediction model may include an initial training and an augmented training. For practical use, the first target prediction model may be obtained after the initial training or after the initial training and the augmented training.
As shown in FIG. 7, the initial training may include a first initial training 710 and a second initial training 720. The model acquisition module 210 may be initially trained to obtain the first target prediction model 1 using any training method. The first target prediction model 1 may be not trained through augmented training. The first target prediction model 1 is the intermediate prediction model.
The first initial training 710 determines the second target prediction model based on the training of the second initial prediction model based on the first training data and obtains the first target prediction model 1 based on the second target prediction model. For more information about the determination of the first target prediction model based on the second prediction model, please refer to FIG. 4 and related descriptions thereof, which will not be repeated here.
The first initial training 720 is performed to train the first initial prediction model based on the first training data to obtain the first target prediction model 1.
The parameters of the first target prediction model 1 may be updated by the augmented training 730 to improve the prediction ability of the model when it is applied in practice. When the processing device makes the first prediction using the first target prediction model 1, the obtained target predicted thermal capacity and the corresponding measured thermal capacity are used as a second set of training data, and the parameters of the first target prediction model 1 are updated based on the set of training data to obtain the first target prediction model 2. The first target prediction model 2 may be used as a model for the next prediction. When the processing device makes a second prediction using the first target prediction model 2, the target predicted thermal capacity obtained from the second prediction is used with the corresponding measured thermal capacity as a second set of training data, and the parameters of the first target prediction model 2 are updated based on this set of training data to obtain the first target prediction model 3. And so on, the first target prediction model may be trained through several augmented training.
In some embodiments, the augmented training may be stopped when the thermal capacity difference between the target predicted thermal capacity and the measured thermal capacity obtained by the first target prediction model after several augmented training sessions meets one or more conditions (e.g., the thermal capacity difference is less than the thermal capacity difference threshold) .
While using the first target prediction model to determine the target predicted thermal capacity of the tube, the parameters of the first target prediction model may be continuously updated according to the corresponding measured thermal capacity of the tube, so that the updated first target prediction model achieves more accurate prediction.
Having described the basic concepts above, it is clear that the above detailed disclosures are intended only as examples for technicians skilled in the art and do not constitute the qualification of this description. Although it is not explicitly stated herein, this description may be subject to various modifications, improvements and corrections by technicians skilled in the art. Such modifications, improvements and corrections are suggested in this description and therefore remain within the spirit and scope of the demonstration embodiments of this description.
Moreover, certain terminology has been configured to describe embodiments of the present disclosure. For example, the terms “one embodiment, ” “an embodiment, ” and “some embodiments” mean that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present disclosure. Therefore, it is emphasized and should be appreciated that two or more references to “an embodiment” or “one embodiment” or “an alternative embodiment” in various portions of this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined as suitable in one or more embodiments of the present disclosure.
Furthermore, unless expressly stated in the claims, the order or elements and sequences of treatment, the use of alphanumeric numbers, or other names described in this description shall not be configured to define the order of processes and methods in this description. Although the above disclosure discusses some embodiments of the invention currently considered useful by various examples, it should be understood that such details are for illustrative purposes only, and the additional claims are not limited to the disclosed embodiments. In stead, the claims are intended to cover all combinations of corrections and equivalents consistent with the substance and scope of the embodiments of the invention. For example, although the implementation of various components described above may be embodied in a hardware device, it may also be implemented as a software only solution, e.g., an installation on an existing server or mobile device.
Similarly, it should be appreciated that in the foregoing description of embodiments of the present disclosure, various features are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure aiding in the understanding of one or more of the various embodiments. This method of disclosure, however, is not to be interpreted as reflecting an intention that the claimed subject matter requires more features than are expressly recited in each claim. Rather, claimed subject matter may lie in less than all features of a single foregoing disclosed embodiment.
In some embodiments, the numbers expressing quantities, properties, and so forth, configured to describe and claim certain embodiments of the application are to be understood as being modified in some instances by the term “about” , “approximate” , or “substantially” . For example, “about” , “approximate” , or “substantially” may indicate ±20%variation of the value it describes, unless otherwise stated. Accordingly, in some embodiments, the numerical parameters set forth in the written description and attached claims are approximations that may vary depending upon the desired properties sought to be obtained by a particular embodiment. In some embodiments, the numerical parameters should be construed in light of the count of reported significant digits and by applying ordinary rounding techniques. Notwithstanding that the numerical ranges and parameters setting forth the broad scope of some embodiments of the application are approximations, the numerical values set forth in the specific examples are reported as precisely as practicable.
For each patent, patent application, patent application publication, and other material, such as article, book, specification, publication, document, etc., cited in the present specification, the entire contents of which are hereby incorporated into the present specification for reference. Except for application history documents that are inconsistent with or conflict with the content of this present disclosure, and except for documents that limit the broadest scope of the claims of this present disclosure (currently or subsequently attached to this present disclosure) . By way of example, should there be any inconsistency or conflict between the description, definition, and/or the use of a term associated with any of the incorporated material and that associated with the present document, the description, definition, and/or the use of the term in the present document shall prevail.
Finally, it should be understood that the embodiments described in the present disclosure are only configured to illustrate the principle of the embodiments of the present disclosure. Other deformation may also belong to the scope of the present disclosure. Therefore, as an example rather than restrictions, the replacement configuration of the embodiment of the present disclosure may be consistent with the teaching of the present disclosure. Correspondingly, the embodiments of the present disclosure are not limited to the implementation and description of the present disclosure.

Claims

A method implemented on at least one machine each of which has at least one processor and at least one storage device for thermal capacity prediction of a medical imaging device 150, the method comprising:

obtaining a first target prediction model, the first target prediction model including a machine learning model;

obtaining property information of a tube of the medical imaging device 150 and working status information of the medical imaging device 150, the working status information including a current thermal capacity of the tube and one or more scanning parameters; and

determining a target predicted thermal capacity of the tube at a preset time point by processing, based on the first target prediction model, the property information and the working status information.
The method of claim 1, wherein

the first target prediction model includes an input layer 610, an incremental prediction layer 620, a fusion layer 630 and an output layer 640,

the input layer 610 is configured to input the property information and the working status information;

the incremental prediction layer 620 is configured to determine, based on the property information and the working status information, a thermal capacity increment of at least one position of the tube at the preset time point;

the fusion layer 630 is configured to determine an initial predicted thermal capacity of the at least one position at the preset time point based on the current thermal capacity and the thermal capacity increment of the at least one position of the tube at the preset time point; and

the output layer 640 is configured to determine the target predicted thermal capacity of the at least one position based on the initial predicted thermal capacity of the at least one position at the preset time point.
The method of claim 1 or 2, wherein the property information includes at least a current thermal dissipation parameter of the tube, the current thermal dissipation parameter of the tube being determined based on a thermal dissipation prediction model by:

processing usage data of the medical imaging device 150, a usage duration of the tube, and an initial thermal dissipation parameter of the tube, based on the thermal dissipation prediction model, the thermal dissipation prediction model including a machine learning model.
The method of any one of claims 1-3, wherein to determine the initial predicted thermal capacity of the at least one position at the preset time point, the fusion layer 630 is further configured to:

for a target position of the at least one position, fuse the thermal capacity increment of the target position, the thermal capacity increment of other positions of the at least one position excluding the target position, and the current thermal capacity.
The method of claim 4, wherein

the input layer 610 is further configured to input a thermal capacity change rate of each position of the at least one position, a loss degree of the each position, a distance between the each position and the other positions of the at least one position; and

to fuse the thermal capacity increment of the target position, the thermal capacity increment of other positions of the at least one position excluding the target position, and the current thermal capacity, the fusion layer 630 is further configured to:

determine a first weight for the each position based on the thermal capacity change rate of the each position, the loss degree of the each position, and the distance between the each position and the other positions of the at least one position;

fuse the thermal capacity increment of the target position, the thermal capacity increment of the other positions, and the current thermal capacity, based on the first weight of the each position.
The method of any one of claims 1-5, wherein to determine the initial predicted thermal capacity of the at least one position at the preset time point, the fusion layer 630 is further configured to:

for a target position of the at least one position, fuse the thermal capacity increment of the target position and the current thermal capacity.
The method of any one of claims 1-6, wherein the obtaining a first target prediction model includes:

obtaining a second initial prediction model;

obtaining first training data, the first training data including a first training sample and a first label corresponding to the first training sample, the first training sample including a sample current thermal capacity of a first sample tube and one or more sample scanning parameters, the first label including a measurement thermal capacity of the first sample tube at a first sample preset time point;

determining a second target prediction model by updating the second initial prediction model based on the first training data; and

determining the first target prediction model based on the second target prediction model.
The method of claim 7, wherein the first training sample further includes sample property information of the first sample tube.
The method of claim 7 or 8, wherein the determining the first target prediction model based on the second target prediction model includes:

obtaining second training data, the second training data including a second training sample and a second label corresponding to the second training sample, the second training sample including a sample target predicted thermal capacity of a second sample tube at a second sample preset time point, the second label including a measurement thermal capacity of the second sample tube at the second sample preset time point;

determining a first intermediate prediction model based on the second target prediction model; and

obtaining the first target prediction model by updating the first intermediate prediction model based on the second training data.
The method of claim 9, wherein the obtaining the first target prediction model by updating the first intermediate prediction model based on the second training data includes:

determining an evaluation result of the first intermediate prediction model by:

processing an intermediate predicted thermal capacity output by the first intermediate prediction model, a usage duration of the tube, a device type of the medical imaging device 150 and a count of updates of the first intermediate prediction model based on an evaluation model;

in response to the evaluation result meeting a preset evaluation condition, stopping updating the first intermediate prediction model to obtain the first target prediction model.
The method of any one of claims 1-5, further comprising:

determining whether the target predicted thermal capacity is greater than a thermal capacity threshold;

in response to the target predicted thermal capacity being greater than the thermal capacity threshold, sending an instruction for stopping scanning to the medical imaging device 150, or sending alert information to a user of a terminal.
The method of claim 11, wherein the thermal capacity threshold is determined by:

determining, based on a life prediction model, a remaining life of the tube corresponding to a preset exposure time range, wherein the life prediction model including a machine learning model;

in response to the remaining life being within a preset life range, generating the thermal capacity threshold based on a preset exposure time range.
A device implemented on at least one machine each of which has at least processor and at least one storage device for thermal capacity prediction of a medical imaging device 150, the system comprising:

a model acquisition module 210, configured to obtain a first target prediction model, the first target prediction model including a machine learning model;

an information acquisition module 220, configured to obtain property information of a tube of the medical imaging device 150 and working status information of the medical imaging device 150, the working status information including a current thermal capacity of the tube and one or more scanning parameters; and

a prediction module 230, configured to determine a target predicted thermal capacity of the tube at a preset time point by processing the property information and the working status information based on the first target prediction model.
The device of claim 13, wherein the first target prediction model includes an input layer 610, an incremental prediction layer 620, a fusion layer 630 and an output layer 640, wherein

the input layer 610 is configured to input the property information and the working status information;

the incremental prediction layer 620 is configured to determine, based on the property information and the working status information, a thermal capacity increment of at least one position of the tube at the preset time point;

the fusion layer 630 is configured to determine an initial predicted thermal capacity of the at least one position at the preset time point based on the current thermal capacity and the thermal capacity increment of the at least one position of the tube at the preset time point; and

the output layer 640 is configured to determine the target predicted thermal capacity of the at least one position based on the initial predicted thermal capacity of the at least one position at the preset time point.
The device of claim 13 or 14, wherein to determine the initial predicted thermal capacity of the at least one position at the preset time point, the fusion layer 630 is further configured to:

for a target position of the at least one position, fuse the thermal capacity increment of the target position, the thermal capacity increment of other positions of the at least one position excluding the target position, and the current thermal capacity.
The device of any one of claims 13-15, wherein to determine the initial predicted thermal capacity of the at least one position at the preset time point, the fusion layer 630 is further configured to:

for a target position of the at least one position, fuse the incremental thermal capacity of the position and the current thermal capacity.
The device of any one of claims 13-16, wherein in order to obtain a first target prediction model, the model acquisition module 210 is configured to:

obtain a second initial second prediction model;

obtain first training data, the first training data including a first training sample and a first label corresponding to the first training sample, the first training sample including a sample current thermal capacity of a first sample tube and one or more sample scanning parameters, the first label including a measurement thermal capacity of the first sample tube at a first sample preset time point;

determine a second target prediction model by updating the second initial prediction model based on the first training data; and

determine the first target prediction model based on the second target prediction model.
The device of claim 17, wherein in order to determine the first target prediction model based on the second target prediction model, the model acquisition module 210 is configured to:

obtain second training data, the second training data including a second training sample and a second label corresponding to the second training sample, the second training sample including a sample target predicted thermal capacity of a second sample tube at a second sample preset time point, the second label including a measurement thermal capacity of the second sample tube at the second sample preset time point;

determine a first intermediate prediction model based on the second target prediction model;

obtain the first target prediction model by updating the first intermediate prediction model based on the second training data.
A medical imaging device 150 including at least one processor and at least one storage;

the at least one storage being configured to store computer instructions;

the at least one processor being configured to execute at least a portion of the computer instructions to implement a method of any one of claims 1 to 12.
A computer storage medium, comprising a set of instructions, wherein and when the computer reads the instructions of the storage medium, the computer performs the method as claimed in any one of claims 1 to 12.