CN116153483A

CN116153483A - Medical data analysis processing method and system based on machine learning

Info

Publication number: CN116153483A
Application number: CN202310015243.1A
Authority: CN
Inventors: 周校平; 陈竹; 章有智
Original assignee: Wuhan Boke Guotai Information Technology Co ltd
Current assignee: Wuhan Boke Guotai Information Technology Co ltd
Priority date: 2023-01-03
Filing date: 2023-01-03
Publication date: 2023-05-23
Anticipated expiration: 2043-01-03
Also published as: CN116153483B

Abstract

The embodiment of the specification provides a medical data analysis processing method and system based on machine learning, wherein the method comprises the following steps: acquiring first medical data related to target medical equipment, wherein the first medical data comprises diagnosis and treatment data generated by the target medical equipment in a preset time period; determining multi-dimensional probability feature information of the target medical device based on the first medical data, wherein the multi-dimensional probability feature information comprises first probability feature information and second probability feature information; and responding to the multi-dimensional probability characteristic information meeting the preset condition, and sending out early warning to the user.

Description

Medical data analysis processing method and system based on machine learning

Technical Field

The present disclosure relates to the field of medical data analysis, and in particular, to a medical data analysis processing method and system based on machine learning.

Background

When many medical devices are operated under high load or are not maintained enough for long-term use, various problems often occur, such as errors in data output by the medical devices, serious image artifacts generated by the medical devices, ageing parts of the medical devices and the like, and if the problems are not found and solved in time, the problems can have serious influence on subsequent diagnosis and treatment.

In general, it is difficult to accurately judge whether a medical device has a fault based on medical data through naked eyes and personal experience, so there is a need for an efficient and accurate method for judging the probability of the medical device to have a fault based on the medical data and giving an early warning in time.

Disclosure of Invention

One or more embodiments of the present specification provide a machine learning-based medical data analysis processing method, the method including: acquiring first medical data related to target medical equipment, wherein the first medical data comprises diagnosis and treatment data generated by the target medical equipment in a preset time period; determining multi-dimensional probability feature information of the target medical device based on the first medical data, wherein the multi-dimensional probability feature information comprises first probability feature information and second probability feature information; and responding to the multi-dimensional probability characteristic information meeting the preset condition, and sending out early warning to the user.

One of the embodiments of the present specification provides a medical data analysis processing system based on machine learning, the system including: the acquisition module is used for acquiring first medical data related to the target medical equipment, wherein the first medical data comprise diagnosis and treatment data generated by the target medical equipment in a preset time period; the determining module is used for determining multi-dimensional probability characteristic information of the target medical equipment based on the first medical data, wherein the multi-dimensional probability characteristic information comprises first probability characteristic information and second probability characteristic information; and the early warning module is used for sending early warning to the user in response to the multi-dimensional probability characteristic information meeting the preset condition.

One or more embodiments of the present specification provide a machine-learning-based medical data analysis processing apparatus including a processor configured to perform the machine-learning-based medical data analysis processing method of any one of the above embodiments.

One or more embodiments of the present specification provide a computer-readable storage medium storing computer instructions that, when read by a computer, perform the machine-learning-based medical data analysis processing method of any one of the above embodiments.

Drawings

The present specification will be further elucidated by way of example embodiments, which will be described in detail by means of the accompanying drawings. The embodiments are not limiting, in which like numerals represent like structures, wherein:

FIG. 1 is an exemplary block diagram of a machine learning based medical data analysis processing system according to some embodiments of the present description;

FIG. 2 is an exemplary flow chart of a machine learning based medical data analysis processing method according to some embodiments of the present description;

FIG. 3 is an exemplary flow chart for determining multi-dimensional probabilistic feature information according to some embodiments of the present description;

FIG. 4 is an exemplary diagram of a first probabilistic information determination model as shown in some embodiments of the present description;

FIG. 5 is an exemplary diagram of a second probabilistic information determination model as shown in some embodiments of the present description;

FIG. 6 is an exemplary schematic diagram of a failure probability determination model, shown in accordance with some embodiments of the present description.

Detailed Description

In order to more clearly illustrate the technical solutions of the embodiments of the present specification, the drawings that are required to be used in the description of the embodiments will be briefly described below. It is apparent that the drawings in the following description are only some examples or embodiments of the present specification, and it is possible for those of ordinary skill in the art to apply the present specification to other similar situations according to the drawings without inventive effort. Unless otherwise apparent from the context of the language or otherwise specified, like reference numerals in the figures refer to like structures or operations.

It will be appreciated that "system," "apparatus," "unit" and/or "module" as used herein is one method for distinguishing between different components, elements, parts, portions or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.

As used in this specification and the claims, the terms "a," "an," "the," and/or "the" are not specific to a singular, but may include a plurality, unless the context clearly dictates otherwise. In general, the terms "comprises" and "comprising" merely indicate that the steps and elements are explicitly identified, and they do not constitute an exclusive list, as other steps or elements may be included in a method or apparatus.

A flowchart is used in this specification to describe the operations performed by the system according to embodiments of the present specification. It should be appreciated that the preceding or following operations are not necessarily performed in order precisely. Rather, the steps may be processed in reverse order or simultaneously. Also, other operations may be added to or removed from these processes.

FIG. 1 is an exemplary block diagram of a machine learning based medical data analysis processing system according to some embodiments of the present description.

As shown in fig. 1, in some embodiments, a machine learning based medical data analysis processing system 100 may include an acquisition module 110, a determination module 120, and an early warning module 130.

The acquiring module 110 is configured to acquire first medical data related to the target medical device, where the first medical data includes diagnosis and treatment data generated by the target medical device during a preset period of time. For more on the target medical device, the first medical data see fig. 2 and its related description.

The determining module 120 is configured to determine multi-dimensional probability feature information of the target medical device based on the first medical data, the multi-dimensional probability feature information including the first probability feature information and the second probability feature information. For more on the multi-dimensional probability feature information, the first probability feature information, the second probability feature information see fig. 2 and its associated description.

In some embodiments, the determining module 120 is further configured to perform feature classification on the first medical data to obtain a plurality of first medical sub-data sets; determining a plurality of first probability feature sub-information based on the plurality of first medical sub-data sets; determining a plurality of second probabilistic characteristic sub-information based on the plurality of first medical sub-data sets; and determining the first probability feature information and the second probability feature information based on the plurality of first probability feature sub-information and the plurality of second probability feature sub-information. For more on the first medical sub-data set, the first probabilistic characteristic sub-information, the second probabilistic characteristic sub-information see fig. 3 and its associated description.

In some embodiments, the determination module 120 is further configured to obtain second medical data and a second set of statements, the second set of statements including related statements on the authenticity of the second medical data; determining a plurality of second medical sub-data sets based on the second medical data, the plurality of second medical sub-data sets corresponding to the plurality of first medical sub-data sets; a plurality of second probabilistic characteristic sub-information is determined based on the plurality of first medical sub-data sets, the plurality of second medical sub-data sets, and the second set of statements. For more on the second medical data, the second set of statements, the second medical sub-data set see fig. 3 and its related description.

The early warning module 130 is configured to send an early warning to the user in response to the multidimensional probability feature information meeting a preset condition. See fig. 2 and the associated description for more details regarding preset conditions.

In some embodiments, the pre-warning module 130 is further configured to determine a probability that the target medical device is malfunctioning based on the multi-dimensional probabilistic feature information; and responding to the probability that the target medical equipment has faults to meet a preset condition, and sending the early warning to a user. For more details on determining the probability of a failure of the target medical device, determining whether preset conditions are met, see fig. 6 and its associated description.

It should be noted that the above description of the system and its modules is for convenience of description only and is not intended to limit the present description to the scope of the illustrated embodiments. It will be appreciated by those skilled in the art that, given the principles of the system, various modules may be combined arbitrarily or a subsystem may be constructed in connection with other modules without departing from such principles. For example, the acquisition module and the determination module may be integrated in one module. For another example, each module may share one storage device, or each module may have a respective storage device. Such variations are within the scope of the present description.

FIG. 2 is an exemplary flow chart of a machine learning based medical data analysis processing method according to some embodiments of the present description. As shown in fig. 2, the process 200 includes the steps of:

at step 310, first medical data associated with a target medical device is acquired. In some embodiments, step 210 may be performed by the acquisition module 110.

The target medical device is a medical device for which it is necessary to determine whether or not there is a failure. Such as CT scanning apparatuses, etc. whose lifetime exceeds a threshold of years.

The first medical data refers to diagnosis and treatment data generated when the target medical device diagnoses the patient. For example, a nuclear magnetic resonance image generated by a nuclear magnetic resonance apparatus, and an X-ray image generated after the X-ray apparatus irradiates a patient.

In some embodiments, the first medical data includes medical data generated by the target medical device for a predetermined period of time. The preset time period refers to a time period set manually in advance. For example, diagnosis and treatment data generated when a patient is diagnosed by a target medical device in the past month.

In some embodiments, the first medical data may be generated by the target medical device and stored in the storage device. In some embodiments, the acquisition module may acquire the first medical data from the storage device.

Step 220, determining multidimensional probability feature information of the target medical device based on the first medical data. In some embodiments, step 220 may be performed by determination module 120.

The multidimensional probability characteristic information refers to probability information for characterizing a problem generated by the target medical device and causing a failure of the problem. For example, the multi-dimensional probability feature information may be a probability that there is a problem of image blurring in a nuclear magnetic resonance image generated by the nuclear magnetic resonance apparatus, and for example, the multi-dimensional probability feature information may be a probability of a failure that causes a decrease in efficiency of the nuclear magnetic resonance image generated by the nuclear magnetic resonance apparatus.

In some embodiments, the multi-dimensional probabilistic characteristic information includes first probabilistic characteristic information and second probabilistic characteristic information.

The first probability characteristic information refers to probability information for characterizing that the target medical device has generated a fault that affects the quality of the data and that causes the problem. For example, the first probability characteristic information may be a probability that there is a macroscopic problem with the image quality generated by the contrast device. As another example, the first probability characteristic information may be a probability of a failure causing a problem with the quality of an image generated by the contrast device that is visible to the naked eye.

The second probability characteristic information refers to probability information for characterizing a problem other than the target medical device that has an influence on the quality of data and a failure that causes the problem. For example, the second probability feature information may be a probability that it takes longer time to generate an image of the current diagnosis and treatment than before, and for example, the second probability feature information may be a probability that a fault occurs in image data generated of the current diagnosis and treatment (the noise is different from noise of image data generated in the past but cannot be distinguished by naked eyes).

In some embodiments, the determination module may model or employ various data analysis algorithms, such as regression analysis, discriminant analysis, etc., to analyze the first medical data to determine multi-dimensional probability characteristic information of the target medical device.

In some embodiments, the determining module may perform feature classification on the first medical data to obtain a plurality of first medical sub-data sets, process the first medical sub-data sets based on a first probabilistic feature information determination model, determine first probabilistic feature information, process the first medical sub-data sets based on a second probabilistic feature information determination model, and determine second probabilistic feature information.

See fig. 3 and its associated description for more content of the first medical sub-data set, see fig. 4 and its associated description for more content of the first probabilistic information determination model, and fig. 5 and its associated description for more content of the second probabilistic information determination model.

And 230, responding to the multi-dimensional probability characteristic information meeting the preset condition, and sending out early warning to the user. In some embodiments, step 230 may be performed by the pre-warning module 130.

The preset condition refers to a judgment condition for determining whether or not the target medical device has a failure. For example, the preset adjustment may be a manually set threshold, and if the probability of the target medical device failing satisfies the threshold, it is determined that the target medical device fails. See fig. 6 and its associated description for more details regarding preset conditions.

In some embodiments, the early warning module may analyze the multidimensional probability feature information by modeling or adopting various data analysis algorithms, such as a regression analysis method, a discriminant analysis method, and the like, and determine whether the multidimensional probability feature information meets a preset condition, and send early warning to the user in response to the multidimensional probability feature information meeting the preset condition.

In some embodiments, the early warning module may determine a failure probability of the target medical device based on the multi-dimensional probability feature information, and send an early warning to the user in response to the failure probability meeting a preset condition. For more on determining the probability of failure see fig. 6 and its associated description.

In some embodiments of the present disclosure, the multi-dimensional probability feature information is determined by processing the first medical data acquired from the target medical device, whether the multi-dimensional probability feature information meets a preset condition is determined, and an early warning is sent to the user in response to the multi-dimensional probability feature information meeting the preset condition, so that the user can be helped to acquire the fault condition of the target medical device timely and accurately, and the accuracy of the subsequent diagnosis and treatment result is ensured.

Fig. 3 is an exemplary flow chart for determining multidimensional probability feature information in accordance with some embodiments of the present description. In some embodiments, the process 300 may be performed by the determination module 120. As shown in fig. 3, the process 300 includes the steps of:

At step 310, the first medical data is feature classified to obtain a plurality of first medical sub-data sets.

The first medical sub-data set refers to a diagnosis and treatment data set containing different patient characteristic categories, which is obtained by carrying out refinement classification on the first medical data according to different patient characteristics. For example, the first medical sub-data may be a diagnosis and treatment data set related to the same disease (e.g., heart disease).

In some embodiments, the determination module may determine the first medical data as different sets of medical sub-data according to different patient characteristic categories, which may include personal information and condition information of the patient, such as age, gender, disease, severity (e.g., early, mid, late, etc.). The patient characteristic category may be determined by clustering the historical first medical data using a clustering algorithm (e.g., K-Means clustering, mean shift clustering, etc.), and the clustered diagnosis and treatment data corresponding to the patient in each category may be used as a first medical sub-data set.

In some embodiments of the present disclosure, the first medical data is divided into a plurality of first medical sub-data sets by means of feature classification, so that the following prediction of the failure probability is guaranteed to be more fit with the actual situation, the accuracy of the prediction is improved, and the efficiency of processing a large amount of data can be effectively improved.

Step 320, determining a plurality of first probability feature sub-information based on the plurality of first medical sub-data sets.

The first probabilistic characteristic sub-information refers to information characterizing the probability that each statement in the first set of statements is a true proposition. The first set of statements refers to a set containing a plurality of statements related to quality aspects of data generated by the target medical device and to failure aspects that lead to quality problems.

In some embodiments, the first Chen Shuji can be constructed based on historical experience (e.g., an expert makes suggestions based on the historical experience). For example, the first statement set {1,2,3} represents { the generated image is severely artifact-the generated image has severe white noise, and there is aging of some structure inside the device }, where "aging of some structure inside the device" is the cause of "the generated image is severely artifact" and "the generated image has severe white noise". The first probability feature sub-information (0.9,0.5,0.1) indicates that the probability that "the generated image artifact is serious" is 90%, the probability that "the generated image has serious white noise" is true is 50%, and the probability that "a certain structure inside the device has aging" is true is 10%.

In some embodiments, the determining module may analyze the plurality of first medical sub-data sets to determine the plurality of first probability feature sub-information by modeling or employing various data analysis algorithms, such as regression analysis, discriminant analysis, and the like.

In some embodiments, the determining module may process the plurality of first medical sub-data sets by a first probabilistic characteristic determination model to determine a plurality of first probabilistic characteristic sub-information. For more details on the first probabilistic information determination model, see FIG. 4 and its associated description.

Step 330, determining a plurality of second probabilistic characteristic sub-information based on the plurality of first medical sub-data sets.

The second probabilistic characteristic sub-information refers to information characterizing the probability that each statement in the second set of statements is a true proposition. The second set of statements refers to a set containing a plurality of statements related to non-quality aspects of data generated by the target medical device and to failure aspects that lead to non-quality problems.

In some embodiments, the second set of statements may be constructed based on historical experience (e.g., suggestions made by an expert based on the historical experience). For example, the second statement set {4,5,6} represents { image generation time unevenness, image generation speed is slow, and there is aging of a structure inside the apparatus, wherein "aging of a structure inside the apparatus" is a cause of "image generation time unevenness" and "image generation speed is slow". The second probability characteristic sub-information (0.8,0.4,0.2) indicates that the probability that "image generation time unevenness" is true is 80%, the probability that "image generation speed is slow" is true is 40%, and the probability that "a certain structure inside the apparatus has aging" is true is 20%.

In some embodiments, step 330 may be implemented by sub-steps 331-333 described below (not shown). In some embodiments, steps 331-333 may be performed by determination module 120.

In a substep 331, second medical data and a second set of statements are obtained.

The second medical data refers to diagnosis and treatment data generated when each patient (or the same kind of patient) in the first medical data receives diagnosis and treatment of other devices of the same type as the target medical device in the same period. For example, a nuclear magnetic resonance image generated when a patient in the first medical data is diagnosed by another hospital using a nuclear magnetic resonance apparatus at the same time.

In some embodiments, the second set of statements includes related statements of the authenticity of the second medical data. For example, the second statement set {1,3,2,4} indicates that there is a slight blurring problem in the nuclear magnetic resonance image generated by the nuclear magnetic resonance facility diagnosis from a certain patient in the first medical data to hospital 1 (assumed 1 to hospital 1) and that there is a serious blurring problem in the nuclear magnetic resonance image generated by the nuclear magnetic resonance facility diagnosis from hospital 2 (assumed 2 to hospital 2). If there is a significant difference in image data generated when two hospitals diagnose using similar diagnosis and treatment equipment, the image data generated by the equipment of hospital 1 (or hospital 2) is in doubt, and the equipment may have faults.

In some embodiments of the present disclosure, the second medical data corresponding to the first medical data is acquired, and the authenticity of the second medical data is judged by the second statement set, so that misdiagnosis can be effectively prevented, and accuracy and scientificity of the second probability characteristic sub-information are improved.

A substep 332 of determining a plurality of second medical sub-data sets based on the second medical data.

The second medical sub-data set refers to a diagnosis and treatment data set containing different patient characteristic categories, which is obtained by carrying out refinement classification on the second medical data according to different patient characteristics. For example, the second medical sub-data set may be a diagnosis and treatment data set related to the same disease (e.g., heart disease).

In some embodiments, the plurality of second medical sub-data sets corresponds to the plurality of first medical sub-data sets. For example, the second medical sub-data set comprising diagnosis and treatment data related to heart disease corresponds to the first medical sub-data set comprising diagnosis and treatment data related to heart disease, and for another example, the second medical sub-data set comprising diagnosis and treatment data of patients with advanced cancer corresponds to the first medical sub-data set comprising diagnosis and treatment data of patients with advanced cancer.

The way of determining the plurality of second medical sub-data sets is the same as the way of determining the plurality of first medical sub-data sets, and reference is made to the classification of the first medical data in the foregoing for more content of determining the plurality of second medical sub-data sets.

Substep 333 determines a plurality of second probabilistic characteristic sub-information based on the plurality of first medical sub-data sets, the plurality of second medical sub-data sets, and the second set of statements.

In some embodiments, the determining module may determine the plurality of second probabilistic characteristic sub-information by modeling or analyzing the plurality of first medical sub-data sets, the plurality of second medical sub-data sets, and the second statement sets using various data analysis algorithms, such as regression analysis, discriminant analysis, and the like.

In some embodiments, the determining module may process the plurality of second medical sub-data sets, and the second set of statements by a second probabilistic characteristic information determining model to determine a plurality of second probabilistic characteristic sub-information. For more on the second probabilistic information determination model see fig. 5 and its associated description.

Step 340, determining the first probability feature information and the second probability feature information based on the plurality of first probability feature sub-information and the plurality of second probability feature sub-information.

For more on the first probabilistic feature information, the second probabilistic feature information see fig. 2 and its related description.

In some embodiments, the determining module may determine the first probabilistic characteristic information and the second probabilistic characteristic information by modeling or using various data analysis algorithms, such as regression analysis, discriminant analysis, and the like, to analyze the plurality of first probabilistic characteristic information and the plurality of second probabilistic characteristic information.

In some embodiments, the determining module may process the plurality of first probability feature sub-information through a first probability feature information determining model to determine a plurality of first probability feature information; the determining module may process the plurality of second probability feature sub-information through a second probability feature information determining model to determine a plurality of second probability feature information. For more on the first probabilistic characteristic information determination model see fig. 4 and its associated description, for more on the second probabilistic characteristic information determination model see fig. 5 and its associated description.

It should be noted that the above description of the flow 200 and the flow 300 is for illustration and description only, and is not intended to limit the scope of applicability of the present description. Various modifications and changes to flow 200 and flow 300 may be made by those skilled in the art under the guidance of this specification. However, such modifications and variations are still within the scope of the present description.

FIG. 4 is an exemplary diagram of a first probabilistic information determination model as shown in some embodiments of the present specification.

The first probability feature information determination model refers to a model for determining the first probability feature information. In some embodiments, the determination module may process the first medical sub-data set based on the first probabilistic characteristic information determination model to determine the first probabilistic characteristic information. In some embodiments, the input of the first probabilistic information determination model comprises a first medical sub-data set and a first statement set, and the output comprises the first probabilistic information.

For more on the first medical sub-data set and the first statement set see fig. 3 and its associated description, for more on the first probability feature information see fig. 2 and its associated description.

In some embodiments, the input of the first probabilistic characterization information determination model further includes a patient characterization category, the patient characterization category corresponding to each of the first medical sub-data sets. For example, the patient characteristic category is the category of the disease, and the corresponding first medical sub-data set is the first medical sub-data set related to the disease.

In some embodiments of the present disclosure, the model may be trained based on sample data of different patient feature classes using the patient feature type as an input to the first probabilistic feature information determination model, making model training more independent and model prediction more accurate.

In some embodiments, the output of the first probabilistic information determination model further comprises a first confidence. The first confidence level refers to a degree of confidence in the authenticity of the first probability feature information. For example, the first statement set {1,2,3} indicates that { the generated image artifact is serious, the generated image has serious white noise, a certain structure inside the device is aged }, the first probability feature information is (0.6,0.2,0.4), the first confidence is (0.8,0.2,0.7), the probability of the generated image artifact of the target medical device being serious is 0.8, the probability of the generated image of the target medical device having serious white noise is 0.2, the probability of the existence of the certain structure inside the target medical device being aged is 40%, and the probability of the existence of the certain structure inside the target medical device being aged is 0.7.

In some embodiments, the first confidence may be determined based on the size of each of the first medical sub-data sets, the number of statements of the first Chen Shuji, and the degree of discretization (e.g., variance, standard deviation, etc.) of the first probability feature sub-information. The larger the scale of the first medical sub-data set, the larger the number of statements of the first Chen Shuji, the lower the degree of dispersion of the first probability sub-information, the higher the first confidence; conversely, the lower the first confidence.

In some embodiments of the present disclosure, the first confidence coefficient corresponding to the first probability feature information is output, so that the degree of reality of the first probability feature information can be further determined, and the reality and the rationality of the target medical device fault probability determination result are ensured.

As shown in fig. 4, the first probabilistic characteristic information determination model may include an embedding layer 440, a sub-information determination layer 460, and a sub-information integration layer 480.

In some embodiments, the embedding layer 440 may process the plurality of first medical sub-data sets, the plurality of patient feature categories, the plurality of first statement sets, and obtain a plurality of embedding vectors. In some embodiments, the embedding layer 440 may be a variety of viable neural network models. For example, convolutional neural networks (Convolutional Neural Networks, CNN), deep neural networks (Deep Neural Networks, DNN), and the like, or combinations thereof.

As shown in FIG. 4, the inputs to the embedded layer 440-1 may include a first medical sub-data set 410-1 (e.g., image data generated by a target medical device for an intestinal disorder), a patient characteristic category 420-1 (e.g., the type of disorder that is an intestinal disorder), and a first Chen Shuji-1 (e.g., {1,2} indicates that problem number 1 is "serious image artifact generated," problem number 2 is "aging of a structure within the device").

The output of the embedding layer 440-1 may include an embedding vector 450-1. An embedded vector refers to a vector that aggregates and characterizes information of the first medical sub-data set, the patient feature class, and the first statement set. For example, (6,1,0.3,2,0.5) indicates that the disease is an intestinal disease (assuming that 6 indicates an intestinal disease), the probability of occurrence of "serious image artifacts" in the case of diagnosing the intestinal disease using the target medical device is 30%, and the probability of occurrence of "aging of a structure in the device" is 50%.

The embedding layer 440-n processes the first medical sub-data set 410-n, the patient feature class 420-n, and the first Chen Shuji-430-n to obtain the specific content of the embedded vector 450-n, which is the same as the content of the embedding layer 440-1 processing the first medical sub-data set 410-1, the patient feature class 420-1, and the first Chen Shuji-430-1 to obtain the embedded vector 450-1, which is not described herein.

In some embodiments, the sub-information determination layer 460 may process the plurality of embedded vectors and the first set of statements to obtain a plurality of first probability feature sub-information. In some embodiments, the sub-information determination layer 460 may be a variety of possible neural network models. For example, convolutional neural networks (Convolutional Neural Networks, CNN), deep neural networks (Deep Neural Networks, DNN), and the like, or combinations thereof.

As shown in fig. 4, the inputs of the sub-information determination layer 460-1 may include the embedded vector 450-1 and the first Chen Shuji-1, and the outputs may include the first probability feature sub-information 470-1. For example, the first statement set {1,2,3} indicates that { the generated image is severely noisy with severe white noise, the generated image has severe white noise, and there is aging of a structure within the device }, and the output first probability feature sub-information (0.3,0.4,0.5) indicates that the target medical device generates an image with a probability of 30% that the image artifact is severely severe, the generated image has a probability of 40% that there is aging of a structure within the target medical device, and the probability of 50%.

The sub-information determining layer 460-n processes the embedded vector 450-n and the first Chen Shuji-430-n to obtain the specific content of the first probability feature sub-information 470-n, and the sub-information determining layer 460-1 processes the embedded vector 450-1 and the first Chen Shuji-430-1 to obtain the content of the first probability feature sub-information 470-1, which is not described herein again.

In some embodiments, the sub-information integration layer 480 may process the plurality of first probability feature sub-information to obtain first probability feature information. In some embodiments, the sub-information integration layer 480 may be a simple mathematical operation, for example, taking an average of each element of all the first probability feature sub-information to obtain the first probability feature information. For another example, if there are three pieces of first probability feature sub-information (0.2,0.3,0.4), (0.5, 0.6, 0.7), and (0.8,0.9,0.1), respectively, the first probability feature information is (0.5,0.6,0.4).

As shown in fig. 4, the inputs of the sub-information integration layer 480 may include first probability feature sub-information 470-1, first probability feature sub-information 470-n, and the outputs may include first probability feature information 490. For example, the first statement set {1,2,3} indicates that { the generated image has serious image artifacts, the generated image has serious white noise, and a certain structure inside the device has aging }, and the output first probability characteristic information (0.4,0.5,0.6) indicates that the probability of the generated image artifacts of the target medical device being serious is 40%, the probability of the generated image having serious white noise is 50%, and the probability of the certain structure inside the target medical device having aging is 60%.

In some embodiments, the output of the embedding layer 440 may be an input of the sub-information determination layer 460, the output of the sub-information determination layer 460 may be an input of the sub-information integration layer 480, and the first probabilistic characteristic information determination model may be derived based on joint training of the embedding layer 440, the sub-information determination layer 460, and the sub-information integration layer 480.

In some embodiments, the first sample data of the joint training includes a first medical sub-data set of samples, a first set of sample patient characteristics, and a first set of sample statements, the first label corresponding to the first sample data being first probability characteristic information of the samples. The first sample data may be obtained based on historical data, and the first tag may be determined by means of manual labeling or automatic labeling. Inputting the first medical sub-data set of the sample, the characteristic category of the sample patient and the first statement set of the sample into an initial embedding layer to obtain the embedding characteristics output by the initial embedding layer; the embedded features are used as training sample data to be input into an initial sub-information determining layer, and first probability feature sub-information output by the initial sub-information determining layer is obtained; and inputting the first probability characteristic sub-information serving as training sample data into an initial sub-information integration layer to obtain the first probability characteristic information output by the initial sub-information integration layer. And constructing a loss function based on the first probability characteristic information output by the sample first probability characteristic information and the sub-information integration layer, and synchronously updating parameters of the embedded layer, the sub-information determination layer and the sub-information integration layer. And obtaining a trained embedded layer, a sub-information determining layer and a sub-information integrating layer through parameter updating.

In some embodiments, multiple embedding layers 440 (e.g., embedding layer 440-1, embedding layer 440-n, etc.) may be implemented based on parameter sharing, and multiple sub-information determination layers 460 (e.g., sub-information determination layers 460-1, sub-information determination layers 460-n) may be implemented based on parameter sharing.

Fig. 5 is an exemplary diagram of a second probabilistic information determination model as shown in some embodiments of the present description.

The second probabilistic characteristic information determination model refers to a model for determining the second probabilistic characteristic information. In some embodiments, the determination module may process the first medical sub-data set based on the second probabilistic characteristic information determination model, determining the second probabilistic characteristic information. In some embodiments, the input of the second probabilistic information determination model comprises a first medical sub-data set, a second medical sub-data set, and a second statement set, and the output comprises second probabilistic information.

For more on the first medical sub-data set, the second medical sub-data set and the second set of statements see fig. 3 and its associated description, for more on the second probability characteristic information see fig. 2 and its associated description.

In some embodiments, the output of the second probabilistic information determination model further comprises a second confidence. The second confidence level refers to a degree of confidence in the authenticity of the second probabilistic feature information. For example, the second statement set {4,5,6} indicates { image generation time unevenness, image generation speed is slow, there is aging of a certain structure inside the apparatus }, the second probability feature information is (0.6,0.2,0.4), the second confidence is (0.8,0.2,0.7), the probability of the target medical apparatus image generation time unevenness is 60% of the reliability is 0.8, the probability of the target medical apparatus image generation speed is slow is 20% of the reliability is 0.2, and the probability of the aging of a certain structure inside the target medical apparatus is 40% of the reliability is 0.7.

In some embodiments, the second confidence may be determined based on the size of each second medical sub-data set, the number of statements of the second Chen Shuji, and the degree of discretization (e.g., variance, standard deviation, etc.) of the second probabilistic characteristic sub-information. The larger the scale of the second medical sub-data set, the larger the number of statements of the second Chen Shuji, the lower the degree of dispersion of the second probability sub-information, the higher the second confidence; conversely, the lower the second confidence.

In some embodiments of the present disclosure, the second confidence corresponding to the second probability feature information is output, so that the degree of reality of the second probability feature information can be further determined, and the reality and the rationality of the target medical device fault probability determination result are ensured.

As shown in fig. 5, the second probabilistic information determination model may include an embedding layer 540, a sub-information determination layer 560, and a sub-information integration layer 580.

In some embodiments, the embedding layer 540 may process the plurality of first medical sub-data sets, the plurality of second medical sub-data sets, and the plurality of second statement sets to obtain a plurality of embedding vectors. In some embodiments, the embedding layer 540 may be a variety of viable neural network models. For example, convolutional neural networks (Convolutional Neural Networks, CNN), deep neural networks (Deep Neural Networks, DNN), and the like, or combinations thereof.

As shown in fig. 5, the inputs to the embedded layer 540-1 may include a first medical sub-data set 510-1 (e.g., image data generated by a target medical device for an intestinal disorder), a second medical sub-data set 520-1 (e.g., image data generated by a patient receiving the same type of device as the target medical device for an intestinal disorder during the same period), and a second Chen Shuji 530-1 (e.g., {1,2} indicates that problem No. 1 is "image generation speed is slow", problem No. 2 is "there is aging of a structure inside the device").

The output of the embedding layer 540-1 may include an embedding vector 550-1. An embedded vector refers to a vector that summarizes and characterizes information of a first medical sub-data set, a second medical sub-data set, and a second statement set. For example, (6,1,0.3,2,0.5) indicates that the disease is an intestinal disease (assuming that 6 indicates an intestinal disease), the probability of having a problem No. 1 that the image generation speed is slow is 30% and the probability of having a problem No. 2 that a certain structure inside the device is aged is 50% when diagnosis and treatment are performed on the intestinal disease using the target medical device.

The embedding layer 540-n processes the first medical sub-data set 510-n, the second medical sub-data set 520-n, and the second Chen Shuji-530-n to obtain the specific content of the embedded vector 550-n, which is the same as the content of the embedding layer 540-1 processing the first medical sub-data set 510-1, the second medical sub-data set 520-1, and the second Chen Shuji-530-1 to obtain the embedded vector 550-1, which is not described herein again.

In some embodiments, the sub-information determination layer 560 may process the plurality of embedded vectors and the second set of statements to obtain a plurality of second probabilistic characteristic sub-information. In some embodiments, the sub-information determination layer 560 may be a variety of possible neural network models. For example, convolutional neural networks (Convolutional Neural Networks, CNN), deep neural networks (Deep Neural Networks, DNN), and the like, or combinations thereof.

As shown in fig. 5, the inputs of the sub-information determination layer 560-1 may include the embedded vector 550-1 and the second Chen Shuji-530-1, and the outputs may include the second probability feature sub-information 570-1. For example, the second statement set {4,5,6} indicates { image generation time unevenness, image generation speed is slow, there is aging in a structure inside the apparatus }, the output second probability characteristic sub-information (0.3,0.4,0.5) indicates that the probability of image generation time unevenness of the target medical apparatus is 30%, the probability of image generation speed of the target medical apparatus is slow is 40%, and the probability of aging in a structure inside the target medical apparatus is 50%.

The sub-information determining layer 560-n processes the embedded vector 550-n and the second Chen Shuji-530-n to obtain the specific content of the second probability feature sub-information 570-n, which is the same as the content of the embedded vector 550-1 and the second Chen Shuji-1 obtained by the sub-information determining layer 560-1, and is not described herein.

In some embodiments, the sub-information integration layer 580 may process the plurality of second probability feature sub-information to obtain second probability feature information. In some embodiments, the sub-information integration layer 580 may be a simple mathematical operation, for example, taking an average of each element of all the second probability feature sub-information to obtain the second probability feature information. For another example, if there are three pieces of second probability feature sub-information (0.2,0.3,0.4), (0.5, 0.6, 0.7), and (0.8,0.9,0.1), respectively, the second probability feature information is (0.5,0.6,0.4).

As shown in fig. 5, the inputs of the sub-information integration layer 580 may include second probability feature sub-information 570-1, second probability feature sub-information 570-n, and the outputs may include second probability feature information 590. For example, the second statement set {4,5,6} indicates { image generation time unevenness, image generation speed is slow, there is aging in a structure inside the apparatus }, the output second probability characteristic information (0.4,0.5,0.6) indicates that the probability of image generation data unevenness of the target medical apparatus is 40%, the probability of image generation speed of the target medical apparatus is slow is 50%, and the probability of aging in a structure inside the target medical apparatus is 60%.

In some embodiments, the output of the embedding layer 540 may be an input of the sub-information determination layer 560, the output of the sub-information determination layer 560 may be an input of the sub-information integration layer 580, and the second probabilistic characteristic information determination model may be derived based on joint training of the embedding layer 540, the sub-information determination layer 560, and the sub-information integration layer 580.

In some embodiments, the second sample data of the joint training includes a first medical sub-data set of samples, a second medical sub-data set of samples, and a second Chen Shuji of samples, the second label corresponding to the second sample data being sample second probability characteristic information. The second sample data may be obtained based on historical data, and the second tag may be determined by means of manual labeling or automatic labeling. Inputting the first medical sub-data set of the sample, the second medical sub-data set of the sample and the second statement set of the sample into an initial embedding layer to obtain the embedding characteristics output by the initial embedding layer; the embedded features are used as training sample data to be input into an initial sub-information determining layer, and second probability feature sub-information output by the initial sub-information determining layer is obtained; and inputting the second probability characteristic sub-information serving as training sample data into an initial sub-information integration layer to obtain the second probability characteristic information output by the initial sub-information integration layer. And constructing a loss function based on the second probability characteristic information output by the sample second probability characteristic information and the sub-information integration layer, and synchronously updating parameters of the embedded layer, the sub-information determination layer and the sub-information integration layer. And obtaining a trained embedded layer, a sub-information determining layer and a sub-information integrating layer through parameter updating.

In some embodiments, multiple embedding layers 540 (e.g., embedding layer 540-1, embedding layer 540-n, etc.) may be implemented based on parameter sharing, and multiple sub-information determination layers 560 (e.g., sub-information determination layers 560-1, sub-information determination layers 560-n) may be implemented based on parameter sharing.

In some embodiments of the present disclosure, the first medical sub-data set, the patient feature class, and the first statement set are processed by the first probabilistic feature information determining model, the first probabilistic feature information is obtained, the first medical sub-data set, the second medical sub-data set, and the second statement set are processed by the second probabilistic feature information determining model, the second probabilistic feature information is obtained, the first probabilistic feature information and the second probabilistic feature information can be quickly and accurately obtained, and the first probabilistic feature information and the second probabilistic feature information are determined based on multiple factors, so that scientificity and rationality of an output result can be ensured.

In some embodiments, the early warning module may determine a failure probability of the target medical device based on the multi-dimensional probability feature information, and send an early warning to the user in response to the failure probability meeting a preset condition.

In some embodiments, the early warning module may analyze the multidimensional probability feature information by modeling or using various data analysis algorithms, such as regression analysis, discriminant analysis, and the like, determine a failure probability of the target medical device, and send early warning to the user in response to the failure probability meeting a preset condition.

In some embodiments, the early warning module may analyze the multidimensional probability feature information through a fault probability determination model, determine a fault probability of the target medical device, and send early warning to the user in response to the fault probability meeting a preset condition.

The failure probability determination model refers to a model for determining the probability of failure of the target medical device. In some embodiments, the input of the failure probability determination model includes first probability feature information, second probability feature information, and the output includes failure probabilities.

In some embodiments, the input of the failure probability determination model further includes a number of first medical sub-data sets. The first medical sub-dataset may be input to a first probabilistic feature information determination model shown in fig. 4, and output as first probabilistic feature information and a first confidence, which may be determined based on the number of first medical sub-datasets. For more content on the first probabilistic information see fig. 2 and its associated description, for more content on the first medical sub-data set see fig. 3 and its associated description, and for more content on the first probabilistic information determination model and the first confidence level see fig. 4 and its associated description.

In some embodiments of the present disclosure, the greater the number of the first medical sub-data sets, the greater the sample breadth, and the higher the accuracy of the model output result, so the accuracy of predicting the failure probability of the target medical device may be effectively improved by using the number of the first medical sub-data sets as the input of the failure probability determination model.

As shown in fig. 6, the failure probability determination model may include a first embedding layer 640-1, a second embedding layer 640-2, and a failure probability determination layer 670.

In some embodiments, the first embedding layer 640-1 may process the first probabilistic feature information 610 to obtain a first feature vector 650. In some embodiments, the first embedding layer 640-1 may be a variety of possible neural network models. For example, convolutional neural networks (Convolutional Neural Networks, CNN), deep neural networks (Deep Neural Networks, DNN), and the like, or combinations thereof.

As shown in fig. 6, the input of the embedding layer 640-1 may include first probability feature information 610 and the output may include a first feature vector 650.

In some embodiments, the first feature vector may be presented in a preset form. For example, the first feature vector may include a plurality of elements, each corresponding to a preset feature, reflecting the specifics of the corresponding feature. For example, the first feature vector 650 may be represented as (a, b, c.), where a represents the probability that the target medical device generated image has a blur problem, b represents the probability that the target medical device generated image has an artifact, and c represents the probability that a component of the target medical device has aging. For another example, the first feature vector 650 (0.2,0.3,0.4) indicates that the target medical device generated image has a probability of 20% of a blur problem, the target medical device generated image has a probability of 30% of an artifact, and a probability of 40% of a component of the target medical device having aging.

In some embodiments, the first feature vector may also be determined by the first embedding layer according to an actual situation, where at least a part of elements of the first feature vector may reflect relevant information corresponding to the first probability feature information. For example, the first feature vector 650 may reflect, in addition to the probability of the target medical device generating a problem and having a fault in the first probability feature information, the number of statements in the first set of statements having a probability greater than a certain threshold (e.g., 0.5). For another example, the first feature vector 650 (0.2,0.3,0.4, 600) indicates that the target medical device generates an image with a probability of 20% of blur problems, the target medical device generates an image with a probability of 30% of artifacts, the target medical device has a probability of 40% of aging of a component, and the number of statements with a probability of greater than 0.5 in the first set of statements is 600. I.e. the number of elements (dimensions) of the first feature vector can be preset, the actual meaning of the individual elements being adaptively determined by the first embedding layer. Thus, the determined first feature vector may have better integrity with the first embedded layer and enable deep mining of the input data.

The second embedding layer 640-2 processes the second probabilistic feature information 620, determines that a second feature vector 660 (for example, (0.2,0.3,0.4, 600) represents that the probability of generating an image by the target medical device is 20%, the probability of generating an image by the target medical device is 30% that the probability of generating an image by the target medical device is slower, the probability of aging a certain component of the target medical device is 40%, and the number of statements with a second statement concentration probability greater than 0.5 is 600), determines that the specific content of the second feature vector 660 is the same as that of the first embedding layer 640-1 for the first probabilistic feature information 610, determines that the content of the first feature vector 650 is the same, and the presentation forms of the second feature vector and the first feature vector are the same as those of the determination, which will not be repeated herein.

In some embodiments, the failure probability determination layer 670 may process the first feature vector 650, the second feature vector 660, and the number 630 of first medical sub-data sets to obtain the failure probability 680. In some embodiments, the failure probability determination layer 670 may be a variety of viable neural network models. For example, convolutional neural networks (Convolutional Neural Networks, CNN), deep neural networks (Deep Neural Networks, DNN), and the like, or combinations thereof.

As shown in fig. 6, the inputs of the failure probability determination layer 670 may include a first feature vector 650, a second feature vector 660, and a number 630 of first medical sub-data sets, and the outputs may include a failure probability 680. The fault probability refers to the probability that the target medical equipment has faults affecting the normal operation of the target medical equipment, the fault probability is an overall probability value (e.g. 70%) determined based on overall evaluation of factors (e.g. first probability characteristic information, second probability characteristic information and the like) of the target medical equipment, and the early warning module can judge whether to send early warning to a user or not based on the fault probability.

In some embodiments, the outputs of the first and second embedding layers 640-1, 640-2 may be inputs to the failure probability determination layer 670, and the failure probability determination model may be derived based on joint training of the first and second embedding layers 640-1, 640-2 and the failure probability determination layer 670.

In some embodiments, the third sample data of the joint training includes sample first probability feature information, sample second probability feature information, and the number of sample first medical sub-data sets, and a third label corresponding to the third sample data is a sample failure probability, where the third label may be represented by 0 and 1, 0 represents that no failure exists, and 1 represents that a failure exists. The third sample data may be obtained based on historical data, and the third tag may be determined by means of manual labeling or automatic labeling. Inputting the first probability characteristic information of the sample into an initial first embedding layer to obtain a first characteristic vector output by the initial first embedding layer; inputting the sample second probability characteristic information into an initial second embedding layer to obtain a second characteristic vector output by the initial second embedding layer; and inputting the first feature vector, the second feature vector serving as training sample data and the number of the first medical sub-data sets of the samples into an initial fault probability determining layer to obtain the fault probability output by the initial fault probability determining layer. And constructing a loss function based on the sample fault probability and the fault probability output by the fault probability determining layer, and synchronously updating parameters of the first embedded layer, the second embedded layer and the fault probability determining layer. And obtaining the trained first embedded layer, the trained second embedded layer and the trained fault probability determining layer through parameter updating.

The preset condition refers to a judgment condition for determining whether or not the target medical device has a failure. In some embodiments, the preset condition may be determined manually.

In some embodiments, the preset condition includes the probability of failure being greater than a probability threshold. The probability threshold may be determined based on the number of statements of the first set of statements and the second set of statements, the greater the number of statements of the first set of statements and the second set of statements, the lower the probability threshold, and vice versa, the higher the probability threshold. The probability threshold may also be determined based on a probability that the relevant statement of the second medical data authenticity in the second set of statements is true.

In some embodiments, the determination module may determine the authenticity of the second medical data generated by a hospital based on historical relevant data of the hospital (e.g., rate of misdiagnosis, number of visits per month, number of heals per month, etc.). In some embodiments, the determination module may analyze the historical relevant data of the hospital by modeling or employing various data analysis algorithms, such as regression analysis, discriminant analysis, etc., to determine the authenticity of the second medical data generated by the hospital. For more on the second medical data, the second set of statements see fig. 3 and its related description.

In some embodiments, the preset condition further comprises the first confidence level being greater than a confidence level threshold. In some embodiments, the preset condition further comprises the second confidence level being greater than a confidence level threshold. In some embodiments, the preset condition further comprises that both the first confidence level and the second confidence level are greater than a confidence level threshold. The confidence threshold may be determined based on historical data or based on expert advice. For more on the first confidence level, the second confidence level see fig. 3 and its associated description.

In some embodiments of the present disclosure, the preset condition is determined based on the number of statements in the first statement set and the second statement set, the probability that the relevant statement corresponding to the second medical data authenticity in the second probability feature information is true, the first confidence and/or the second confidence, and multiple factors may be considered to determine the preset condition, so as to ensure the reasonability of the preset condition.

While the basic concepts have been described above, it will be apparent to those skilled in the art that the foregoing detailed disclosure is by way of example only and is not intended to be limiting. Although not explicitly described herein, various modifications, improvements, and adaptations to the present disclosure may occur to one skilled in the art. Such modifications, improvements, and modifications are intended to be suggested within this specification, and therefore, such modifications, improvements, and modifications are intended to be included within the spirit and scope of the exemplary embodiments of the present invention.

Meanwhile, the specification uses specific words to describe the embodiments of the specification. Such as "one embodiment" _、 "one embodiment" _、 And/or "some embodiments" means a particular feature, structure, or characteristic described in connection with at least one embodiment of the present disclosure. Thus, it should be emphasized and should be appreciated that two or more references to "an embodiment" or "one embodiment" or "an alternative embodiment" in various positions in this specification are not necessarily referring to the same embodiment. Furthermore, certain features, structures, or characteristics of one or more embodiments of the present description may be combined as suitable.

Furthermore, the order in which the elements and sequences are processed, the use of numerical letters, or other designations in the description are not intended to limit the order in which the processes and methods of the description are performed unless explicitly recited in the claims. While certain presently useful inventive embodiments have been discussed in the foregoing disclosure, by way of various examples, it is to be understood that such details are merely illustrative and that the appended claims are not limited to the disclosed embodiments, but, on the contrary, are intended to cover all modifications and equivalent arrangements included within the spirit and scope of the embodiments of the present disclosure. For example, while the system components described above may be implemented by hardware devices, they may also be implemented solely by software solutions, such as installing the described system on an existing server or mobile device.

Likewise, it should be noted that in order to simplify the presentation disclosed in this specification and thereby aid in understanding one or more inventive embodiments, various features are sometimes grouped together in a single embodiment, figure, or description thereof. This method of disclosure, however, is not intended to imply that more features than are presented in the claims are required for the present description. Indeed, less than all of the features of a single embodiment disclosed above.

In some embodiments, numbers describing the components, number of attributes are used, it being understood that such numbers being used in the description of embodiments are modified in some examples by the modifier "about," approximately, "or" substantially. Unless otherwise indicated, "about," "approximately," or "substantially" indicate that the number allows for a 20% variation. Accordingly, in some embodiments, numerical parameters set forth in the specification and claims are approximations that may vary depending upon the desired properties sought to be obtained by the individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and employ a method for preserving the general number of digits. Although the numerical ranges and parameters set forth herein are approximations that may be employed in some embodiments to confirm the breadth of the range, in particular embodiments, the setting of such numerical values is as precise as possible.

Each patent, patent application publication, and other material, such as articles, books, specifications, publications, documents, etc., referred to in this specification is incorporated herein by reference in its entirety. Except for application history documents that are inconsistent or conflicting with the content of this specification, documents that are currently or later attached to this specification in which the broadest scope of the claims to this specification is limited are also. It is noted that, if the description, definition, and/or use of a term in an attached material in this specification does not conform to or conflict with what is described in this specification, the description, definition, and/or use of the term in this specification controls.

Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments of this specification. Other variations are possible within the scope of this description. Thus, by way of example, and not limitation, alternative configurations of embodiments of the present specification may be considered as consistent with the teachings of the present specification. Accordingly, the embodiments of the present specification are not limited to only the embodiments explicitly described and depicted in the present specification.

Claims

1. A machine learning-based medical data analysis processing method, the method comprising:

Acquiring first medical data related to target medical equipment, wherein the first medical data comprises diagnosis and treatment data generated by the target medical equipment in a preset time period;

determining multi-dimensional probability feature information of the target medical device based on the first medical data, the multi-dimensional probability feature information including first probability feature information and second probability feature information;

and responding to the multi-dimensional probability characteristic information meeting a preset condition, and sending an early warning to a user.

2. The method of claim 1, wherein the determining multi-dimensional probability characteristic information of the target medical device based on the first medical data comprises:

performing feature classification on the first medical data to obtain a plurality of first medical sub-data sets;

determining a plurality of first probability feature sub-information based on the plurality of first medical sub-data sets;

determining a plurality of second probabilistic feature sub-information based on the plurality of first medical sub-data sets; the method comprises the steps of,

determining the first probability feature information and the second probability feature information based on the plurality of first probability feature sub-information and the plurality of second probability feature sub-information.

3. The method of claim 2, wherein the determining a plurality of second probabilistic characteristic sub-information based on the plurality of first medical sub-data sets comprises:

Acquiring second medical data and a second set of statements, the second set of statements comprising relevant statements on the authenticity of the second medical data;

determining a plurality of second medical sub-data sets based on the second medical data, the plurality of second medical sub-data sets corresponding to the plurality of first medical sub-data sets;

the plurality of second probabilistic characteristic sub-information is determined based on the plurality of first medical sub-data sets, the plurality of second medical sub-data sets, and the second set of statements.

4. The method of claim 1, wherein the issuing an alert to a user in response to the multi-dimensional probability feature information meeting a preset condition comprises:

determining a fault probability of the target medical device based on the multi-dimensional probability feature information;

and responding to the fault probability meeting the preset condition, and sending the early warning to the user.

5. A machine learning based medical data analysis processing system, the system comprising:

the acquisition module is used for acquiring first medical data related to target medical equipment, wherein the first medical data comprise diagnosis and treatment data generated by the target medical equipment in a preset time period;

A determining module configured to determine, based on the first medical data, multi-dimensional probabilistic feature information of the target medical device, the multi-dimensional probabilistic feature information including first probabilistic feature information and second probabilistic feature information;

and the early warning module is used for sending early warning to a user in response to the multi-dimensional probability characteristic information meeting a preset condition.

6. The system of claim 5, wherein the determination module is further to:

7. The system of claim 6, wherein the determination module is further configured to:

8. The system of claim 5, wherein the pre-warning module is further configured to:

and responding to the fault probability meeting a preset condition, and sending the early warning to the user.

9. A machine learning based medical data analysis processing apparatus, the apparatus comprising at least one processor and at least one memory;

the at least one memory is configured to store computer instructions;

the at least one processor is configured to execute at least some of the computer instructions to implement the method of any one of claims 1-4.

10. A computer readable storage medium storing computer instructions which, when executed by a processor, implement a method as claimed in any one of claims 1 to 4.