CN117116472A - Medical diagnostic apparatus, electronic device, and storage medium - Google Patents

Abstract

The present disclosure provides a medical diagnostic apparatus, an electronic device, and a storage medium, the medical diagnostic apparatus including: the first acquisition module is used for acquiring first abnormal characteristic information of a target structure in the medical image, wherein the first abnormal characteristic information characterizes whether the target structure is abnormal or not; the second acquisition module is used for acquiring a disease type list corresponding to the target structure, wherein the disease type list comprises a plurality of diseases occurring in the target structure; and the diagnosis module is used for obtaining first diagnosis information based on the association degree between the first abnormal characteristic information and the plurality of diseases. With this, the first diagnosis information can be determined based on the correspondence between the abnormality characteristic information and the disease, and the interpretability of the output result can be improved. Meanwhile, the user can clearly output the judgment basis of the result so as to verify the credibility of the output result.

Description

Medical diagnostic apparatus, electronic device, and storage medium

Technical Field

The present disclosure relates to the field of medical diagnosis, and in particular, to a medical diagnosis apparatus, an electronic device, and a storage medium.

Background

Generally, medical image recognition technology or medical big data analysis can utilize the technical advantages of artificial intelligence in detection, classification and deep learning to help doctors to acquire information more quickly and perform quantitative analysis, and becomes an auxiliary tool for helping doctors to complete diagnosis and treatment work.

The machine learning algorithms such as convolutional neural networks and the like are outstanding in the fields of medical image recognition, intelligent auxiliary diagnosis based on medical knowledge and the like by virtue of the advantages of automatic feature learning, high precision, strong expandability and the like. In the related art, a medical image identification method is used for carrying out lesion level diagnosis of a target tissue by extracting characteristics of first order statistics, gray level co-occurrence matrix, gray level run-length matrix, gray level area size matrix, neighborhood gray level difference matrix and the like of a nuclear magnetic resonance image of the target tissue and utilizing a machine learning algorithm such as BP neural network and the like; the method reduces the calculation difficulty of the algorithm by screening a limited number of features, but has poor interpretability of the features and the algorithm, and the output result is difficult to visualize, so that when a doctor faces the output result, the judgment basis of the output result is difficult to be known, and the doctor is prevented from verifying and quantitatively analyzing the output result. The pulmonary lesion benign and malignant risk layering auxiliary diagnosis system in the other related technology comprises the steps of nodule detection, doctor guidance, semantic annotation generation, sample generation, online learning and the like; in the diagnosis process, a doctor is required to mark the concerned point, and the prediction probability of the lung nodule outline in the concerned region and each pixel point in the outline is calculated, so that the specific outline and probability of the lung nodule can be identified by the method, but the identification process requires manual intervention, and the diagnosis efficiency is low. Meanwhile, the method is used for identifying and diagnosing the same modal data, and has difficulty in applying different modal data such as texts and numerical values.

Disclosure of Invention

In view of the above, the present disclosure provides a medical diagnostic apparatus, an electronic device, and a storage medium, so as to solve the problem of poor interpretability of output artificial intelligence output results in cross-modal data collaborative applications and related technologies.

In a first aspect, there is provided a medical diagnostic apparatus comprising: the first acquisition module is used for acquiring first abnormal characteristic information of a target structure in the medical image, wherein the first abnormal characteristic information characterizes whether the target structure is abnormal or not; the second acquisition module is used for acquiring a disease type list corresponding to the target structure, wherein the disease type list comprises a plurality of diseases occurring in the target structure; and the diagnosis module is used for obtaining first diagnosis information based on the association degree between the first abnormal characteristic information and the plurality of diseases.

In some embodiments, obtaining the first diagnostic information based on the degree of association between the first abnormality characteristic information and the plurality of diseases includes: traversing a plurality of diseases in a list of disease types; calculating a first association degree between the first abnormal characteristic information and each disease according to expert opinion information corresponding to the target structure; calculating a second association degree between the first abnormal characteristic information and each disease according to the historical diagnosis information corresponding to the target structure; based on the first degree of association and the second degree of association between the first abnormality characteristic information and each disease, first diagnosis information is obtained.

In some embodiments, deriving the first diagnostic information based on the first degree of association and the second degree of association between the first abnormality characteristic information and each disease comprises: obtaining second diagnosis information based on the first association degree between the first abnormal characteristic information and each disease, wherein the second diagnosis information comprises at least one disease and the probability of diagnosing each disease in the at least one disease; obtaining third diagnosis information based on the second association degree between the first abnormality characteristic information and each disease, the third diagnosis information including at least one disease and a probability of diagnosing each disease in the at least one disease; the first diagnostic information is obtained based on the probability of each disease in the second diagnostic information and the third diagnostic information.

In some embodiments, deriving the second diagnostic information based on the first degree of association between the first abnormality characteristic information and each disease comprises: arranging a plurality of diseases in the disease type list according to the sequence from the high degree of association to the low degree of association to obtain a disease queue; based on the disease queue, second diagnostic information is generated.

In some embodiments, the expert opinion information includes original anomalous sample data and expert diagnostic results corresponding to the original anomalous sample data, the expert diagnostic results being obtained by delta film; wherein, according to the expert opinion information that the goal structure corresponds, calculate the first degree of association between each disease and the first unusual characteristic information, include: obtaining second abnormal characteristic information based on the original abnormal sample data; based on the second abnormal characteristic information and the expert diagnosis result, obtaining the correlation between each disease in the disease type list and the patient sign; based on the correlation between each disease and the patient sign and the first abnormality characteristic information, a first degree of correlation between the first abnormality characteristic information and each disease is obtained using an attention mechanism.

In some embodiments, the historical diagnostic information includes third abnormal characteristic information and a historical diagnostic result corresponding to the third abnormal characteristic information, and calculating a second degree of association between the first abnormal characteristic information and each disease from the historical diagnostic information corresponding to the target structure includes: calculating a frequent item set between the third abnormal characteristic information and the historical diagnosis result; based on the frequent item set, obtaining a correlation between each disease in the list of disease types and patient signs; and obtaining a second degree of association between the first abnormal characteristic information and each disease based on the correlation between each disease and the patient sign and the first abnormal characteristic information by using an attention mechanism.

In some embodiments, the diagnostic module is further to: inputting the first abnormal characteristic information into a prediction model to obtain fourth diagnosis information; obtaining fifth diagnosis information based on the first diagnosis information and the fourth diagnosis information; wherein the predictive model comprises a random forest model.

In some embodiments, acquiring first abnormal feature information of a target structure in a medical image includes: extracting image abnormal feature information based on the medical image, the image abnormal feature information including: morphological information of the target structure, gray information of an image of the target structure, size information of the target structure and spatial position information of the target structure; and/or acquiring a medical record text corresponding to the medical image, and extracting text abnormal characteristic information based on the medical record text; and/or acquiring a numerical inspection report corresponding to the medical image, and extracting numerical abnormality characteristic information based on the numerical inspection report; the first abnormal feature information is determined based on the image abnormal feature information, the text abnormal feature information, and the numerical abnormal feature information.

In some embodiments, the medical diagnostic apparatus is applied to middle ear lesion diagnosis, the first abnormality feature information is extracted from a middle ear image, comprising: middle ear morphology information, gray scale information of middle ear images, middle ear size information, and middle ear spatial position information.

In some embodiments, the medical diagnostic apparatus is applied to liver disease diagnosis, the first abnormality characteristic information is extracted from a liver image, comprising: liver morphology information, gray scale information of liver images, liver size information, and liver spatial position information.

In a second aspect, there is provided an electronic device comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform a medical diagnostic method via execution of the executable instructions, the medical diagnostic method comprising: acquiring first abnormal characteristic information of a target structure in the medical image, wherein the first abnormal characteristic information characterizes whether the target structure is abnormal or not; acquiring a disease type list corresponding to a target structure, wherein the disease type list comprises a plurality of diseases occurring in the target structure; based on the degree of association between the first abnormality characteristic information and the plurality of diseases, first diagnosis information is obtained.

In a third aspect, there is provided a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a medical diagnostic method comprising: acquiring first abnormal characteristic information of a target structure in the medical image, wherein the first abnormal characteristic information characterizes whether the target structure is abnormal or not; acquiring a disease type list corresponding to a target structure, wherein the disease type list comprises a plurality of diseases occurring in the target structure; based on the degree of association between the first abnormality characteristic information and the plurality of diseases, first diagnosis information is obtained.

The medical diagnosis device provided by the embodiment of the disclosure can acquire the first abnormal characteristic information of the target structure in the medical image and the disease type list corresponding to the target structure, and obtain the first diagnosis information according to the association degree between the first abnormal characteristic information and a plurality of diseases. The first diagnosis information is obtained based on each structural feature in the first abnormal feature information, so that the output result of the device can clearly determine the corresponding relation between the abnormal feature information and the diseases, and has good interpretability; meanwhile, when the user faces the output result, the user can know the judging basis of the disease according to the structural characteristics of the target structure so as to verify the credibility of the output result.

Drawings

Fig. 1 is a schematic view of a scenario in which a medical diagnostic apparatus according to an embodiment of the present disclosure is applied.

Fig. 2 is a schematic structural view of a medical diagnostic apparatus according to an embodiment of the present disclosure.

Fig. 3 is a schematic flow chart of obtaining first diagnosis information based on the association degree between the first abnormal characteristic information and the plurality of diseases according to the embodiment of the disclosure.

Fig. 4 is a schematic flow chart of calculating a first association degree between the first abnormal feature information and each disease according to expert opinion information corresponding to a target structure according to an embodiment of the present disclosure.

Fig. 5 is a schematic flow chart of calculating a second association degree between the first abnormal feature information and each disease according to the historical diagnosis information corresponding to the target structure according to the embodiment of the disclosure.

Fig. 6 is a schematic flow chart of obtaining first diagnosis information based on the first association degree and the second association degree between the first abnormal characteristic information and each disease according to the embodiment of the disclosure.

Fig. 7 is a flowchart illustrating steps performed by a diagnostic module according to an embodiment of the disclosure.

Fig. 8 is a flowchart illustrating a process of acquiring first abnormal feature information of a target structure in a medical image according to an embodiment of the disclosure.

Fig. 9 is a schematic structural diagram of an electronic device according to an embodiment of the disclosure.

Detailed Description

Example embodiments will now be described more fully with reference to the accompanying drawings. However, the exemplary embodiments may be embodied in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

Furthermore, the drawings are merely schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus a repetitive description thereof will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software or in one or more hardware modules or integrated circuits or in different networks and/or processor devices and/or microcontroller devices.

As described above, the medical image recognition often adopts a deep learning algorithm such as a convolutional neural network. Because of the self algorithm characteristics, the output result is often poor in interpretation. Therefore, it is difficult for a doctor to relate the medical image and the output result of the algorithm, so that the doctor is difficult to know the judgment basis when facing the output result, and further the doctor is prevented from analyzing and verifying the output result.

In view of this, the present disclosure provides a medical diagnostic apparatus comprising: the first acquisition module is used for acquiring first abnormal characteristic information of a target structure in the medical image, wherein the first abnormal characteristic information characterizes whether the target structure is abnormal or not; the second acquisition module is used for acquiring a disease type list corresponding to the target structure, wherein the disease type list comprises a plurality of diseases occurring in the target structure; and the diagnosis module is used for obtaining first diagnosis information based on the association degree between the first abnormal characteristic information and the plurality of diseases. Therefore, the medical diagnosis device provided by the embodiment of the disclosure can obtain the output result through the abnormal characteristics in the target structure shown by the medical image and the association relationship between the diseases and the abnormal characteristics. The result is obtained based on the association relationship between the disease and the abnormal characteristics, and the association relationship between the abnormal characteristic information and the disease can be clearly determined, so that the method has good interpretability. Therefore, doctors can further verify and quantitatively analyze the output result according to the embodiment of the disclosure, diagnosis efficiency and accuracy are improved, and application value of clinical medical images is improved.

The embodiment of the disclosure provides a medical diagnosis device, electronic equipment and a storage medium. The medical diagnostic apparatus may be integrated in an electronic device, which may be a terminal or a server or the like.

Fig. 1 is a schematic diagram of a scenario in which an embodiment of the present disclosure is applicable. The scenario includes a medical diagnostic apparatus 100, the medical diagnostic apparatus 100 including a feature extraction module 101 and a prediction module 102.

In some embodiments, the feature extraction module 101 is configured to extract the affected part feature from the medical image; the prediction module 102 is configured to predict the affected part feature extracted from the medical image by the feature extraction module 101, so as to obtain diagnostic information.

In practical application, the feature extraction module 101 can extract abnormal feature information of a target structure shown in a medical image from the medical image, and transmit the abnormal feature information to the prediction module 102, where the prediction module 102 predicts based on the abnormal feature information of the target structure to obtain first diagnosis information to assist a doctor in analyzing and judging the medical image. The medical image may be taken by X-ray imaging, computed Tomography (CT), magnetic Resonance Imaging (MRI), ultrasound imaging, or the like.

In some embodiments, the medical diagnostic apparatus 100 further comprises an imaging apparatus for acquiring corresponding medical images. The imaging device may include a camera, an X-ray device, an electronic computed tomography device, a magnetic resonance imaging device, or the like, depending on the target structure to which the medical image corresponds.

The present exemplary embodiment will be described in detail below with reference to the accompanying drawings and examples.

Fig. 2 illustrates a schematic structural diagram of a medical diagnostic apparatus in an embodiment of the present disclosure, and as illustrated in fig. 2, the medical diagnostic apparatus 200 includes: a first acquisition module 201, a second acquisition module 202, and a diagnostic module 203.

The first obtaining module 201 is configured to obtain first abnormal feature information of a target structure in a medical image.

The target structure is a part which is shown in the medical image and needs to be checked and diagnosed. The medical image can show the shape, size, gray information, spatial position and other characteristics of the target structure, so that corresponding first abnormal characteristic information can be extracted based on the medical image to represent whether the target structure is abnormal or not, namely whether the characteristics of the current target structure are different from the normal structural characteristics of the target structure or not.

In some embodiments, the image recognition model may perform data preprocessing on the medical image and clean up the data to extract the first abnormal feature information therein. The abnormal characteristic information comprises a plurality of components, each component represents whether one structural characteristic of the target structure is abnormal or not, and binary coding is carried out by 0 and 1.

In some embodiments, the doctor may further determine the medical image to obtain whether the target structure shown by the medical image is abnormal, and input the determination result into the medical diagnostic apparatus 200, and the first acquisition module 201 generates corresponding first abnormal feature information.

The second obtaining module 202 is configured to obtain a disease type list corresponding to the target structure.

In some embodiments, the second acquisition module 202 can acquire a corresponding list of disease types from the target structure, where the list of disease types includes disease categories that occur in the target structure, such as inflammation, cancer, tissue damage, and the like. More specifically, a plurality of multiple, representative specific disease names occurring at the target structure may also be included in the list of disease types. For example, if the target structure is a middle ear structure, then a list of corresponding disease types includes a plurality of diseases occurring in the middle ear structure, such as, for example, otitis media, otosclerosis, tympanic membrane perforation, and middle ear tumors. If the target structure is a liver structure, the list of corresponding disease types includes a plurality of diseases occurring in the liver structure, such as liver cancer, fatty liver, liver fibrosis, drug-induced liver injury, hepatic hemangioma, and the like.

The diagnosis module 203 is configured to obtain first diagnosis information based on the first abnormal characteristic information and the degree of association between the plurality of diseases.

In some embodiments, the first diagnostic information includes at least one disease, and a probability that the target structure shown in the medical image is afflicted with the disease. Wherein the number of diseases in the first diagnostic information may be the same as the number of diseases in the disease type list, and at this time, the first diagnostic information shows the disease probability of all diseases in the disease type list; or, setting preset conditions according to actual needs, so that the first diagnosis information only comprises diseases with the possibility of corresponding to the diseases higher than a preset threshold value; or, preset conditions are set according to actual needs, so that the first diagnosis information only comprises diseases with the top possibility rank corresponding to the diseases, flexibility of output results is enhanced, and part of calculation amount is reduced.

In some embodiments, because of the association between the lesions and the disease of the target structure morphology, i.e., the target structure afflicted with a disease, certain structural features of the morphology are more likely to be distinguishable from those of the normal target structure. Therefore, according to the association relationship, the degree of association between the first abnormal characteristic information and the plurality of diseases can be quantified, and the degree of association between the first abnormal characteristic information and the plurality of diseases can be obtained. And calculating the probability of each disease in the disease type list of the target structure shown in the medical image according to the association degree, and further obtaining first diagnosis information.

Therefore, the device of the embodiment can acquire the first abnormal characteristic information of the target structure and the disease type list corresponding to the target structure in the medical image, and obtain the first diagnosis information according to the association degree between the first abnormal characteristic information and the plurality of diseases. The output result of the embodiment is obtained based on each structural feature in the first abnormal feature information, so that the output result of the device according to the embodiment can clearly determine the corresponding relationship between the abnormal feature information and the disease, and has good interpretability; meanwhile, when facing the output result, doctors can know the judging basis of the diseases according to the structural characteristics of the target structure so as to verify the credibility of the output result; in addition, relevant technicians can further adjust device parameters according to the doctor verification result, and accuracy of output results of the device is further improved.

In some embodiments, the medical diagnostic apparatus is applied to middle ear lesion diagnosis, and the first abnormality feature information is extracted from a middle ear image. The first abnormal characteristic information includes at least 4 components, which are middle ear morphology information, middle ear image gray information, middle ear size information and middle ear spatial position information, respectively. Middle ear morphology information represents whether middle ear morphology is abnormal, middle ear image gray information represents whether middle ear image gray is abnormal, middle ear size information represents whether middle ear size is abnormal, and middle ear space position information represents whether middle ear space position is abnormal. Illustratively, middle ear morphology information may be 1 or 0, with 1 indicating that a middle ear morphology is abnormal and 0 indicating that a middle ear morphology is not abnormal.

In other embodiments, the medical diagnostic apparatus is applied to liver lesion diagnosis, and the first abnormality characteristic information is extracted from a liver image. The first abnormal characteristic information includes at least 4 components, which are liver morphology information, liver image gray scale information, liver size information, and liver spatial position information, respectively. Liver morphology information represents whether the liver morphology is abnormal, liver image gray information represents whether the liver image gray is abnormal, liver size information represents whether the liver size is abnormal, and liver space position information represents whether the liver space position is abnormal. Illustratively, the liver morphology information may be 1 or 0,1 indicating that the liver morphology is abnormal and 0 indicating that the middle ear morphology is not abnormal. It may be understood that the components included in the first abnormal feature information are only schematic, and other feature information may also be included in the first abnormal feature information, which may be specifically determined according to an actual application scenario.

In some embodiments, the diagnostic module 203 is configured to obtain the first diagnostic information based on the first abnormal characteristic information and the degree of association between the plurality of diseases; as shown in fig. 3, the first diagnosis information is obtained based on the degree of correlation between the first abnormality characteristic information and the plurality of diseases, and includes the following steps.

S301, traversing a plurality of diseases in the disease type list.

Specifically, traversing a plurality of diseases in the disease type list, and respectively carrying out the following calculation on the basis of the first abnormal characteristic information and each disease to respectively obtain the association degree corresponding to each disease. Therefore, the possibility of each disease corresponding to the first abnormal characteristic information can be determined, omission is avoided, and the comprehensiveness and the integrity of inspection are ensured.

S302, calculating a first association degree between the first abnormal characteristic information and each disease according to expert opinion information corresponding to the target structure.

Specifically, the degree of association between the target structural feature and the diseases can be quantified by collecting expert opinions, and then the first degree of association between the first abnormal feature information and the plurality of diseases can be obtained.

In some embodiments, the expert opinion information includes raw abnormal sample data and expert diagnostic results corresponding to the raw abnormal sample data. The original abnormal sample data comprises a plurality of images of the target structure, and the images respectively represent various structural features of the target structure to be normal and/or abnormal. And, each image corresponds to a disease that the target structure shown in the image may suffer from, i.e., an expert diagnosis result. Each image may correspond to a disease; alternatively, each image may correspond to a plurality of diseases.

In some embodiments, the expert diagnostic result corresponding to the raw abnormal sample data may be obtained by delta film (Delphi method). The Delphi method is a structured decision support method, adopts a form of poll investigation, adopts a mode of anonymously publishing comments according to a program of a system, repeatedly inquires, generalizes, modifies and technically processes the views of an expert about original abnormal sample data through multiple rounds of investigation, and finally gathers the views which are basically consistent with the expert to be used as a final diagnosis result. The specific steps are as follows.

Establishing an expert group, wherein the expert group comprises doctors or expert scholars in corresponding fields, and determining the number of expert people according to the actually required knowledge range and task amount;

providing all members of the panel with questions and associated requirements to be predicted, and attaching relevant data (i.e., raw abnormal sample data);

each member gives out own comments according to the received data, including whether the data belongs to abnormality, judging and indicating whether the data is normal or belongs to a certain type in multiple types of abnormality, and one or more diseases possibly suffering from the abnormal data;

summarizing the first opinion, comparing, and dividing the comparison result to each member to enable the member to modify own opinion according to the comparison result; alternatively, the comments of each member may be collated, and comments may be given to other experts than the panelist and then sent to each member so that they may modify their own comments after reference;

Collecting the opinions of all the members, summarizing, and distributing the opinion to the members again so as to make a second modification. The rounds are modified in this way until each member no longer changes his opinion. At this time, the opinions of the members are collected to obtain expert diagnosis results.

In some embodiments, as shown in fig. 4, calculating the first degree of association between the first abnormality characteristic information and each disease according to expert opinion information corresponding to the target structure further includes the following steps.

S3021, obtaining second abnormal characteristic information based on the original abnormal sample data.

Specifically, the second abnormal feature information may be extracted from the original abnormal sample data based on an image recognition method or manual recognition. The second abnormal characteristic information characterizes whether a plurality of structural characteristics of a target structure shown in each image in the original abnormal sample data are abnormal or not.

And S3022, obtaining the correlation between each disease in the disease type list and the patient sign based on the second abnormal characteristic information and the expert diagnosis result.

Specifically, through the steps, the expert diagnosis results and the second abnormal characteristic information corresponding to each image in the original abnormal sample data can be obtained. Thus, the correspondence, i.e. correlation, between each disease and patient sign can be obtained by statistical calculation.

S3023, obtaining a first degree of association between the first abnormality characteristic information and each disease based on the correlation between each disease and the patient' S sign and the first abnormality characteristic information using the attention mechanism.

Specifically, the attention mechanism can assign different weights to the information according to the importance degree of the information, so that the computing resources are distributed to more important information, the attention degree of other information is reduced, and the accuracy and the computing efficiency of an output result are improved.

In some embodiments, the second abnormality characteristic information may be grouped based on a correlation between each disease and patient sign, each group corresponding to one disease; respectively calculating the Similarity (Similarity) between each group of second abnormal characteristic information and the first abnormal characteristic information, and giving higher weight to the second abnormal characteristic information with higher Similarity; and weighting and summing expert diagnosis results corresponding to the second abnormal characteristic information according to the weight to finally obtain a first association degree between the first abnormal characteristic information and each disease.

According to the method and the device, the first abnormal characteristic information can be analyzed and judged according to the objective expression of expert opinion, and an attention mechanism is introduced, so that the accuracy and the calculation efficiency of an output result are improved. Therefore, the first degree of association obtained in the present embodiment is predicted based on the expert opinion information, so that the predicted result is closer to the clinical diagnosis result. At the same time, the results are based on the correlation between the disease and the patient's sign, and thus have good interpretability.

S303, calculating a second association degree between the first abnormal characteristic information and each disease according to the historical diagnosis information corresponding to the target structure.

Specifically, the degree of association between the target structural feature and the disease may be quantified according to the historical diagnostic information corresponding to the target structure, so as to obtain the first abnormal feature information and the second degree of association between the plurality of diseases.

In some embodiments, the historical diagnostic information includes third abnormality characteristic information and a historical diagnostic result corresponding to the third abnormality characteristic information. The third abnormal characteristic information and the historical diagnosis result are obtained by historical case data, and the historical case data comprises a plurality of images of a target structure and the historical diagnosis result corresponding to each image; the third abnormal feature information may be extracted from the images based on an image recognition method or manual recognition, and the third abnormal feature information characterizes whether a plurality of structural features of a target structure shown in each image in the historical case data are abnormal.

In some embodiments, as shown in fig. 5, the second degree of association between the first abnormality characteristic information and each disease is calculated according to the historical diagnosis information corresponding to the target structure, including the following steps.

S3031, a frequent item set between the third abnormal characteristic information and the historical diagnosis result is calculated.

In particular, a frequent item set is a set of variables that often occur together in a data set. The item set refers to a set of a plurality of items; the support of an item set is defined as the proportion of records in the data set that contain the item set; therefore, the item set with the support degree not less than the minimum support degree may be defined as a frequent item set. Thus, frequent item sets can represent variables that occur together frequently in a data set. The minimum support degree can be preset manually according to actual requirements, and can also be calculated by certain algorithms.

S3032, based on the frequent item sets, a correlation between each disease in the list of disease types and patient signs is obtained.

Specifically, the frequent item set may be statistically calculated or processed by an algorithm to obtain association rules between the disease and the patient's signs. The reliability of the association rule indicates the reliability of the association rule, and if the association rule is a→b, the reliability is: support ({ A, B })/support ({ A }), wherein A and B are each a class of item sets. And finally, determining the correlation between each disease in the disease type list and the patient sign based on the correlation rule that the credibility is not less than the minimum credibility. The minimum credibility can be preset manually according to actual requirements, and can also be calculated by certain algorithms.

In some embodiments, frequent term sets may be calculated by the Apriori algorithm, resulting in association rules. Firstly, finding out all frequency sets, wherein the frequency of occurrence of the frequency sets is at least the same as the predefined minimum support degree; then generating strong association rules from the frequency set, which must meet minimum support and minimum credibility; then using the above-mentioned frequency set to produce desired rule, producing all rules containing only the items of the set; and reserving rules with credibility not less than the minimum credibility in all the rules. In this way, the final association rule is obtained by a recursive method.

S3033, using the attention mechanism, obtaining a second degree of association between the first abnormal characteristic information and each disease based on the correlation between each disease and the patient' S sign and the first abnormal characteristic information.

As described above, the third abnormality characteristic information may be grouped based on the correlation between each disease and patient signs, each group corresponding to one disease; respectively calculating the similarity between each group of third abnormal characteristic information and the first abnormal characteristic information, and giving higher weight to the third abnormal characteristic information with higher similarity; and weighting and summing the historical diagnosis results corresponding to the third abnormal characteristic information according to the weight to finally obtain a second association degree between the first abnormal characteristic information and each disease.

According to the method and the device, the first abnormal characteristic information can be analyzed and judged according to objective rules of historical data, and an attention mechanism is introduced, so that accuracy and calculation efficiency of an output result are improved. Therefore, the first association degree obtained in the embodiment is predicted based on the historical data, so that the predicted result is closer to the statistical result. At the same time, the results are based on the correlation between the disease and the patient's sign, and thus have good interpretability.

S304, obtaining first diagnosis information based on the first association degree and the second association degree between the first abnormal characteristic information and each disease.

Specifically, the first association degree and the second association degree can be combined through preset weights, and first diagnosis information is obtained according to the combined weights. The preset weight can be determined based on experimental results and/or the reliability degree between expert opinions and historical data. And flexibly adjusting the weight between the first association degree and the second association degree so as to improve the accuracy of the output result.

In some embodiments, as shown in fig. 6, the first diagnosis information is obtained based on the first degree of association and the second degree of association between the first abnormality characteristic information and each disease, including the following steps.

S3041, obtaining second diagnosis information based on the first degree of association between the first abnormality characteristic information and each disease.

Specifically, the first association degree corresponding to each disease may be normalized, and the normalized data may be determined as the probability corresponding to each disease. The probability corresponding to the certain disease indicates the probability that the image corresponding to the first abnormal characteristic information is diagnosed as the disease. It is understood that the first abnormal characteristic information corresponds to a plurality of diseases and a plurality of probabilities of suffering from the diseases, respectively. Based on the probability, second diagnostic information may be further obtained.

In some embodiments, deriving the second diagnostic information based on the first degree of association between the first abnormality characteristic information and each disease comprises: arranging a plurality of diseases in the disease type list according to the sequence from the high degree of association to the low degree of association to obtain a disease queue; based on the disease queue, second diagnostic information is generated.

Illustratively, the first association degree corresponding to the plurality of diseases may be ranked, or the probability corresponding to the plurality of diseases may be ranked, so as to obtain a disease queue. After the disease queue is obtained, the disease included in the second diagnostic information may be determined based on the disease queue.

For example, the number of diseases included in the second diagnosis information may be predetermined according to actual demands, and a pre-set number of diseases in the disease queue may be selected as the second diagnosis information. For example, if only one diagnosis result is desired, the preset number is set to one, so that the disease ranked first in the disease queue is selected; if n diagnosis results are obtained, the preset number is set to be n, so that the first n diseases in the disease queue are selected, and a doctor is assisted to comprehensively judge and analyze according to a plurality of pieces of information.

Through the arrangement, the personalized output result can be flexibly output, and the use experience is improved.

And S3042, obtaining third diagnosis information based on the second association degree between the first abnormal characteristic information and each disease.

Specifically, the second association degree corresponding to each disease may be normalized, and the normalized data may be determined as the probability corresponding to each disease. Based on the probabilities, third diagnostic information may be further derived.

S3043, obtaining first diagnosis information based on the probability of each disease in the second diagnosis information and the third diagnosis information.

Specifically, a preset weight may be determined according to the reliability of the second diagnosis information and the third diagnosis information, and the probability weight of each disease in the second diagnosis information and the third diagnosis information may be summed up according to the preset weight. And sorting the plurality of diseases according to the final probability, and selecting a proper number of diseases as first diagnosis information.

By this arrangement, it is possible to integrate objective expression of expert opinion and objective rules of history data, and predict the first abnormality characteristic information from various angles. Expert opinion relies on priori knowledge of clinical experts, and clinical subjective experience has great influence on specific gravity; meanwhile, the historical data obtains a certain rule through learning the historical data, so that possible diagnosis results are predicted, and the historical data and the possible diagnosis results are both good and bad. Therefore, the embodiment flexibly combines the output results of the expert opinion and the historical data through the reliability degree of the expert opinion and the historical data to mutually compensate for each other, and finally the first diagnosis information is obtained. Therefore, the accuracy and the reliability of the output result can be improved.

In addition, the first abnormal characteristic information can be analyzed through some machine learning algorithms, so that the accuracy of the output result is further improved. As shown in fig. 7, the diagnostic module 203 is also used to implement the following steps.

S701, inputting the first abnormal characteristic information into a prediction model to obtain fourth diagnosis information.

In some embodiments, the expert opinion information and/or historical diagnostic information described above may be used as a training data set to train a predictive model. The prediction model can predict fourth diagnosis information corresponding to the first abnormal characteristic information based on the input first abnormal characteristic information, wherein the fourth diagnosis information comprises at least one disease, and the probability that a target structure shown by the first abnormal characteristic information is suffering from the disease.

In some embodiments, the predictive model may be a random forest model. The random forest model can be randomly sampled in the training data set to form n different sample data sets; constructing n different decision trees according to n different sample data sets; and finally, determining a final result according to the voting results of the n decision trees.

The random forest model can simply process high-dimensional data without feature selection, and can effectively prevent the model from being over-fitted. In addition, the accuracy of the output result is improved in a mode of voting by multiple decision trees.

S702, obtaining fifth diagnosis information based on the first diagnosis information and the fourth diagnosis information.

Although the prediction model has good accuracy, the prediction process is black box, and the interpretation is poor. Therefore, the fourth diagnosis information and the first diagnosis information can be organically combined to obtain the fifth diagnosis information, and the accuracy of the output result is further improved. At this time, the output result combines the advantages of the two parts of diagnosis information, has good accuracy and reliability, and has good interpretability.

The above embodiments describe the acquisition of the first abnormal feature information from the medical image, and the specific implementation of the acquisition of the first abnormal feature information from the medical image corresponding to the different target structures is described by way of example. In actual diagnosis, the inspector can present a relevant diagnosis report aiming at the medical image of the target structure. At the same time, the medical image may also be associated with textual information such as patient history, physical examination, pathology report, etc., and numerical information in the numerical examination report. Thus, the abnormal feature information of the target structure can be further extracted from the above information.

Specifically, as shown in fig. 8, the first abnormal feature information of the target structure in the medical image is acquired, which includes the following steps.

S801, extracting image anomaly characteristic information based on a medical image, the image anomaly characteristic information including: morphological information of the target structure, gray information of the target structure image, size information of the target structure, and spatial position information of the target structure.

S802, acquiring a medical record text corresponding to the medical image, and extracting text abnormal characteristic information based on the medical record text.

The medical history information may include medical history information of the patient, physical examination information, pathology report information, diagnostic reports of medical images, and the like. The text anomaly characteristic information may include: whether there is a past history, family history, smoking, drinking, etc. associated with the target structure. The text anomaly characteristic information may be represented by 1 or 0, where 1 indicates that the characteristic is anomalous and 0 indicates that the characteristic is not anomalous.

S803, acquiring a numerical inspection report corresponding to the medical image, and extracting numerical abnormality characteristic information based on the numerical inspection report.

The numerical examination report may include a patient's hormone, blood routine, tumor marker examination report, and the like. The numerical anomaly characteristic information may include: information on whether the hormone concentration is abnormal (within a reference range), whether the leukocyte concentration is abnormal, whether the tumor marker concentration is abnormal, and the like. The numerical anomaly characteristic information may be represented by either 1 or 0, with 1 indicating that the characteristic is anomalous and 0 indicating that the characteristic is not anomalous.

S804, determining first abnormal characteristic information based on the image abnormal characteristic information, the text abnormal characteristic information and the numerical value abnormal characteristic information.

Because of the association relationship between the abnormal features of the target structure, the abnormal features can be combined into a matrix or multidimensional feature parameters so as to obtain the multi-modal first abnormal feature information.

The information of the target structure acquired from multiple aspects can describe the specific condition of the target structure more accurately, so that more accurate results can be obtained based on the multi-mode first abnormal characteristic information. For example, the image abnormal characteristic information extracted from the medical image and the text abnormal characteristic information extracted from the diagnosis report of the medical image are obviously related, so that the image abnormal characteristic information and the text abnormal characteristic information can be mutually verified, and the accuracy of the result is further improved.

It should be noted that, when the medical diagnostic apparatus provided in the above embodiment is used for medical diagnosis, only the division of the above functional modules is used for illustration, and in practical application, the above functional allocation may be performed by different functional modules according to needs, that is, the internal structure of the apparatus is divided into different functional modules to perform all or part of the functions described above.

Those skilled in the art will appreciate that the various aspects of the present disclosure may be implemented as a system, method, or program product. Accordingly, various aspects of the disclosure may be embodied in the following forms, namely: an entirely hardware embodiment, an entirely software embodiment (including firmware, micro-code, etc.) or an embodiment combining hardware and software aspects may be referred to herein as a "circuit," module "or" system.

Based on the same inventive concept, an electronic device 900 is also provided in the embodiments of the present disclosure. The electronic device 900 shown in fig. 9 is merely an example and should not be construed as limiting the functionality and scope of application of the embodiments of the present disclosure.

As shown in fig. 9, the electronic device 900 is embodied in the form of a general purpose computing device. Components of electronic device 900 may include, but are not limited to: the at least one processing unit 910, the at least one storage unit 920, and a bus 930 connecting the different system components (including the storage unit 920 and the processing unit 910).

Wherein the storage unit stores program code that can be executed by the processing unit 910 such that the processing unit 910 performs the steps described in the present specification according to various exemplary embodiments of the present disclosure.

In some embodiments, the processing unit 910 may perform the following steps of a medical diagnostic method: acquiring first abnormal characteristic information of a target structure in a medical image, wherein the first abnormal characteristic information characterizes whether the target structure is abnormal or not; acquiring a disease type list corresponding to the target structure, wherein the disease type list comprises a plurality of diseases generated in the target structure; and obtaining first diagnosis information based on the association degree between the first abnormal characteristic information and the plurality of diseases.

The storage unit 920 may include readable media in the form of volatile storage units, such as Random Access Memory (RAM) 9201 and/or cache memory 9202, and may further include Read Only Memory (ROM) 9203.

The storage unit 920 may also include a program/utility 9204 having a set (at least one) of program modules 9205, such program modules 9205 include, but are not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of which may include an implementation of a network environment.

The bus 930 may be one or more of several types of bus structures including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

The electronic device 900 may also communicate with one or more external devices 940 (e.g., keyboard, pointing device, bluetooth device, etc.), one or more devices that enable a user to interact with the electronic device 900, and/or any devices (e.g., routers, modems, etc.) that enable the electronic device 900 to communicate with one or more other computing devices. Such communication may occur through an input/output (I/O) interface 950. Also, electronic device 900 may communicate with one or more networks such as a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network, such as the Internet, through network adapter 960. As shown, the network adapter 960 communicates with other modules of the electronic device 900 over the bus 930. It should be appreciated that although not shown, other hardware and/or software modules may be used in connection with electronic device 900, including, but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, data backup storage systems, and the like.

From the above description of embodiments, those skilled in the art will readily appreciate that the example embodiments described herein may be implemented in software, or may be implemented in software in combination with the necessary hardware. Thus, the technical solution according to the embodiments of the present disclosure may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (may be a CD-ROM, a U-disk, a mobile hard disk, etc.) or on a network, including several instructions to cause a computing device (may be a personal computer, a server, a terminal device, or a network device, etc.) to perform the method according to the embodiments of the present disclosure.

In an exemplary embodiment of the present disclosure, a computer-readable storage medium, which may be a readable signal medium or a readable storage medium, is also provided. On which a program product is stored which enables a medical diagnostic method. In some possible implementations, the various aspects of the disclosure may also be implemented in the form of a program product comprising program code for causing a terminal device to carry out the steps according to the various exemplary embodiments of the disclosure as described in the above method examples section of this specification, when the program product is run on the terminal device.

More specific examples of the computer readable storage medium in the present disclosure may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

In this disclosure, a computer readable storage medium may include a data signal propagated in baseband or as part of a carrier wave, with readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A readable signal medium may also be any readable medium that is not a readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

Alternatively, the program code embodied on a computer readable storage medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

In particular implementations, the program code for carrying out operations of the present disclosure may be written in any combination of one or more programming languages, including an object oriented programming language. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device, partly on a remote computing device, or entirely on the remote computing device or server. In the case of remote computing devices, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., connected via the Internet using an Internet service provider).

It should be noted that although in the above detailed description several modules or units of a device for action execution are mentioned, such a division is not mandatory. Indeed, the features and functionality of two or more modules or units described above may be embodied in one module or unit in accordance with embodiments of the present disclosure. Conversely, the features and functions of one module or unit described above may be further divided into a plurality of modules or units to be embodied.

Furthermore, although the steps of the methods in the present disclosure are depicted in a particular order in the drawings, this does not require or imply that the steps must be performed in that particular order or that all illustrated steps be performed in order to achieve desirable results. Additionally or alternatively, certain steps may be omitted, multiple steps combined into one step to perform, and/or one step decomposed into multiple steps to perform, etc.

Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This disclosure is intended to cover any adaptations, uses, or adaptations of the disclosure following the general principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

Claims (10)

1. A medical diagnostic apparatus, comprising:

the first acquisition module is used for acquiring first abnormal characteristic information of a target structure in the medical image, wherein the first abnormal characteristic information characterizes whether the target structure is abnormal or not;

The second acquisition module is used for acquiring a disease type list corresponding to the target structure, wherein the disease type list comprises a plurality of diseases generated in the target structure;

and the diagnosis module is used for obtaining first diagnosis information based on the association degree between the first abnormal characteristic information and the plurality of diseases.

2. The apparatus of claim 1, wherein the obtaining first diagnostic information based on the degree of association between the first abnormality characteristic information and the plurality of diseases comprises:

traversing a plurality of diseases in the list of disease types;

calculating a first association degree between the first abnormal characteristic information and each disease according to expert opinion information corresponding to the target structure;

calculating a second degree of association between the first abnormal characteristic information and each disease according to the historical diagnosis information corresponding to the target structure;

and obtaining the first diagnosis information based on the first association degree and the second association degree between the first abnormal characteristic information and each disease.

3. The apparatus according to claim 2, wherein the obtaining the first diagnostic information based on the first degree of association and the second degree of association between the first abnormality characteristic information and each of the diseases includes:

Obtaining second diagnosis information based on the first association degree between the first abnormal characteristic information and each disease, wherein the second diagnosis information comprises at least one disease and the probability of diagnosing each disease in the at least one disease;

obtaining third diagnosis information based on a second degree of association between the first abnormality characteristic information and each of the diseases, the third diagnosis information including at least one disease, and a probability of diagnosing each of the at least one disease;

and obtaining the first diagnosis information based on the probability of each disease in the second diagnosis information and the third diagnosis information.

4. The apparatus of claim 3, wherein the obtaining second diagnostic information based on the first degree of association between the first abnormality characteristic information and each of the diseases comprises:

arranging a plurality of diseases in the disease type list according to the sequence from the high degree of association to the low degree of association to obtain a disease queue;

the second diagnostic information is generated based on the disease queue.

5. The apparatus according to claim 2, wherein the expert opinion information includes original abnormal sample data and expert diagnosis results corresponding to the original abnormal sample data, the expert diagnosis results being obtained by delta film;

Wherein the calculating a first degree of association between the first abnormal characteristic information and each disease according to expert opinion information corresponding to the target structure includes:

obtaining second abnormal characteristic information based on the original abnormal sample data;

based on the second abnormal characteristic information and the expert diagnosis result, obtaining a correlation between each disease in the disease type list and patient signs;

and obtaining a first association degree between the first abnormal characteristic information and each disease based on the correlation between each disease and the patient sign and the first abnormal characteristic information by using an attention mechanism.

6. The apparatus of claim 2, wherein the historical diagnostic information includes third abnormality characteristic information and a historical diagnostic result corresponding to the third abnormality characteristic information, and wherein the calculating of the second degree of association between the first abnormality characteristic information and each of the diseases based on the historical diagnostic information corresponding to the target structure includes:

calculating a frequent item set between the third abnormal characteristic information and the historical diagnosis result;

based on the frequent item set, obtaining a correlation between each disease in the disease type list and patient signs;

And obtaining a second association degree between the first abnormal characteristic information and each disease based on the correlation between each disease and the patient sign and the first abnormal characteristic information by using an attention mechanism.

7. The apparatus of claim 1, wherein the diagnostic module is further to:

inputting the first abnormal characteristic information into a prediction model to obtain fourth diagnosis information;

obtaining fifth diagnosis information based on the first diagnosis information and the fourth diagnosis information;

wherein the predictive model comprises a random forest model.

8. The apparatus of claim 1, wherein the acquiring the first abnormal feature information of the target structure in the medical image comprises:

extracting image abnormal feature information based on the medical image, wherein the image abnormal feature information comprises: morphological information of the target structure, gray information of an image of the target structure, size information of the target structure and spatial position information of the target structure; and/or

Acquiring a medical record text corresponding to the medical image, and extracting text abnormal characteristic information based on the medical record text; and/or

Acquiring a numerical inspection report corresponding to the medical image, and extracting numerical abnormal characteristic information based on the numerical inspection report;

And determining the first abnormal characteristic information based on the image abnormal characteristic information, the text abnormal characteristic information and the numerical value abnormal characteristic information.

9. An electronic device, comprising:

a processor; and

a memory for storing executable instructions of the processor;

wherein the processor is configured to perform a medical diagnostic method via execution of the executable instructions, the medical diagnostic method comprising:

acquiring first abnormal characteristic information of a target structure in a medical image, wherein the first abnormal characteristic information characterizes whether the target structure is abnormal or not;

acquiring a disease type list corresponding to the target structure, wherein the disease type list comprises a plurality of diseases generated in the target structure;

and obtaining first diagnosis information based on the association degree between the first abnormal characteristic information and the plurality of diseases.

10. A computer readable storage medium having stored thereon a computer program, characterized in that the computer program, when executed by a processor, implements a medical diagnostic method comprising:

acquiring first abnormal characteristic information of a target structure in a medical image, wherein the first abnormal characteristic information characterizes whether the target structure is abnormal or not;

Acquiring a disease type list corresponding to the target structure, wherein the disease type list comprises a plurality of diseases generated in the target structure;

and obtaining first diagnosis information based on the association degree between the first abnormal characteristic information and the plurality of diseases.