Diagnosis recommendation method and device, equipment, storage medium and program product

Abstract

The invention provides a diagnosis recommendation method and device, equipment, a storage medium and a program product. The diagnosis recommendation method comprises the steps that target specific diagnosis corresponding to a region where a patient is located is determined; wherein the target specific diagnosis refers to historical diagnosis with distinction degree from other regions in the region where the patient is located; determining a target historical medical record similar to the to-be-diagnosed medical record of the patient, and obtaining a target historical diagnosis corresponding to the target historical medical record; determining candidate diagnoses based on the target historical diagnoses and the target specific diagnoses; and based on the candidate diagnosis, determining a diagnosis recommendation result corresponding to the to-be-diagnosed medical record. The accuracy of diagnosis recommendation can be improved.

Description

Technical Field

[0001] The present application relates to the field of text processing technology, and in particular to a diagnosis recommendation method, device, equipment, storage medium and program product. Background Art

[0002] Auxiliary diagnosis reasoning refers to a solution that uses artificial intelligence algorithms to help doctors make faster and more accurate judgments on patients' conditions in clinical consultation scenarios. As patients' conditions become more and more diverse in clinical medicine, some misdiagnosis and missed diagnosis are inevitable in clinical consultation scenarios. Auxiliary diagnosis methods based on the combination of big data analysis and artificial intelligence technology can effectively assist doctors in clinical consultation, reduce misdiagnosis and missed diagnosis, and improve the accuracy and efficiency of diagnosis.

[0003] Current auxiliary diagnosis technologies usually only make diagnosis recommendations based on the patient's condition information, and the accuracy of diagnosis recommendations is low. Summary of the invention

[0004] The present application provides a diagnosis recommendation method, apparatus, device, storage medium and program product to improve the accuracy of diagnosis recommendation.

[0005] According to a first aspect of an embodiment of the present application, a diagnosis recommendation method is provided, comprising:

[0006] Determine a target-specific diagnosis corresponding to the patient's region; wherein the target-specific diagnosis refers to a historical diagnosis in the patient's region that is distinguishable from other regions;

[0007] Determine a target historical medical record that is similar to the patient's medical record to be diagnosed, and obtain a target historical diagnosis corresponding to the target historical medical record;

[0008] Determining a candidate diagnosis based on the target historical diagnosis and the target-specific diagnosis;

[0009] Based on the candidate diagnoses, a diagnosis recommendation result corresponding to the medical record to be diagnosed is determined.

[0010] According to a second aspect of an embodiment of the present application, a diagnosis recommendation device is provided, comprising:

[0011] A first processing unit is used to determine a target-specific diagnosis corresponding to the region where the patient is located; wherein the target-specific diagnosis refers to a historical diagnosis in the region where the patient is located that is distinguishable from other regions;

[0012] The second processing unit is used to determine a target historical medical record similar to the patient's medical record to be diagnosed, and obtain a target historical diagnosis corresponding to the target historical medical record;

[0013] a third processing unit, configured to determine a candidate diagnosis based on the target historical diagnosis and the target-specific diagnosis;

[0014] The recommendation unit is used to determine the diagnosis recommendation result corresponding to the medical record to be diagnosed based on the candidate diagnosis.

[0015] According to a third aspect of an embodiment of the present application, there is provided an electronic device, including a memory and a processor;

[0016] The memory is connected to the processor and is used to store programs;

[0017] The processor is used to implement the diagnosis recommendation method as described in the first aspect by running the program in the memory.

[0018] According to a fourth aspect of an embodiment of the present application, a storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the diagnosis recommendation method as described in the first aspect is implemented.

[0019] According to a fifth aspect of an embodiment of the present application, a computer program product is provided, comprising computer program instructions, which, when executed by a processor, cause the processor to execute the diagnosis recommendation method as described in the first aspect.

[0020] In this application, the target-specific diagnosis corresponding to the patient's region is determined, wherein the target-specific diagnosis refers to a historical diagnosis in the patient's region that is distinguishable from other regions, and the target historical medical records similar to the patient's medical records to be diagnosed are determined, and the target historical diagnosis corresponding to the target historical medical records is obtained. Based on the target historical diagnosis and the target-specific diagnosis, the candidate diagnosis is determined, and based on the candidate diagnosis, the diagnosis recommendation result corresponding to the medical record to be diagnosed is determined. When making a diagnosis recommendation, not only the target historical diagnosis corresponding to the target historical medical records similar to the patient's medical records to be diagnosed is considered, but also the target-specific diagnosis corresponding to the patient's region is considered, which can more accurately locate the patient's condition, which is of significant significance for some regional rare diseases and high-value diseases, and effectively helps clinicians to control the risks of regional rare diseases, high-value diseases, etc., and improves the accuracy of diagnosis recommendations compared to diagnosis recommendations based only on patient condition information. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0022] Figure 1 A schematic diagram of a diagnostic recommendation method provided in an embodiment of the present application;

[0023] Figure 2 A schematic diagram of a process of step 101 provided in an embodiment of the present application;

[0024] Figure 3 A schematic diagram of a process for determining a target historical medical record provided in an embodiment of the present application;

[0025] Figure 4 A schematic diagram of a process of step 104 provided in an embodiment of the present application;

[0026] Figure 5 A schematic diagram of a process of step 401 provided in an embodiment of the present application;

[0027] Figure 6 A schematic diagram of a process of step 503 provided in an embodiment of the present application;

[0028] Figure 7 A schematic diagram of the structure of a diagnosis recommendation device provided in an embodiment of the present application;

[0029] Figure 8 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0030] Auxiliary diagnosis reasoning refers to a solution that uses artificial intelligence algorithms to help doctors make faster and more accurate judgments on patients' conditions in clinical consultation scenarios. As patients' conditions become more and more diverse in clinical medicine, some misdiagnosis and missed diagnosis are inevitable in clinical consultation scenarios. Auxiliary diagnosis methods based on the combination of big data analysis and artificial intelligence technology can effectively assist doctors in clinical consultation, reduce misdiagnosis and missed diagnosis, and improve the accuracy and efficiency of diagnosis.

[0031] In medical scenarios, auxiliary diagnosis based on patient medical records is very important, which can greatly improve the accuracy and efficiency of diagnosis. In recent years, with the rapid development of artificial intelligence and machine learning technologies, rule-based algorithms, machine learning algorithms, and deep learning-based algorithms have been widely used in auxiliary diagnosis.

[0032] At present, there are mainly the following algorithms for auxiliary diagnosis:

[0033] Algorithm 1

[0034] Rule-based algorithms: Rule-based algorithms mainly rely on expert systems and decision trees. Expert systems use expert knowledge to simulate the decision-making process and perform diagnostic reasoning through a series of established logical rules. Decision trees navigate the decision path through a series of questions and ultimately reach a diagnostic conclusion. Decision trees are easy to understand and implement, but may lack flexibility and accuracy.

[0035] It is highly dependent on professional medical knowledge and requires experts to manually design disease diagnosis rules. It has poor adaptability to some rare or unseen diseases and requires continuous manual updating of the rules, which is time-consuming and laborious.

[0036] Algorithm 2

[0037] Machine learning-based algorithms: Traditional machine learning models such as random forests, support vector machines (SVM), and K-Nearest Neighbor (KNN) also play a role in medical diagnosis. Random forests build multiple decision trees and vote to get the final result. They can handle high-dimensional data and are often used for feature selection and disease prediction. Support vector machines are suitable for classification and regression problems, especially for small sample data sets. K-Nearest Neighbor is a simple and direct algorithm that performs classification or regression by measuring the distance between different feature values.

[0038] Relying on manually extracted features and the knowledge of domain experts, it has relatively stronger generalization capabilities than rule-based algorithms, but it still requires people with professional medical knowledge to construct features, and the optimization threshold is high. In addition, machine learning algorithms usually rely on a large amount of labeled data, which makes the collection and labeling process time-consuming and laborious.

[0039] Algorithm 3

[0040] Algorithms based on deep learning: Deep learning methods, especially convolutional neural networks (CNN) and recurrent neural networks (RNN), have achieved remarkable results in medical image analysis and natural language processing in recent years. Convolutional neural networks perform well in medical image processing (such as X-ray, CT (Computed Tomography), MRI (Magnetic Resonance Imaging) image analysis) and are used to automatically identify lesion areas in images. Recurrent neural networks (such as LSTM (Long-Short Term Memory)) are suitable for processing sequence data, such as patients' electronic health records (EHR), and are used to predict the development trend of diseases and the future health status of patients.

[0041] The internal structure and parameters of deep learning models are complex, and it is difficult to intuitively explain the model's learning process and prediction results. In addition, the effectiveness of deep learning models also depends on a large amount of task-specific training data. For fields with small sample sizes, the model's generalization ability may be limited.

[0042] In addition, current auxiliary diagnosis technologies usually only make diagnosis recommendations based on the patient's condition information, and the accuracy of diagnosis recommendations is low.

[0043] In order to improve the accuracy of diagnosis recommendations, the present application provides a diagnosis recommendation method, apparatus, device, storage medium and program product.

[0044] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0045] Exemplary Implementation Environment

[0046] The diagnostic recommendation method according to the embodiment of the present application can be executed by an electronic device such as a terminal device or a server. The terminal device can be a user device, a mobile device, a computing device, a wearable device, etc. The server can be an independent physical server, a server cluster composed of multiple physical servers, or a cloud server capable of cloud computing. The method can be implemented by a processor calling a computer-readable program instruction stored in a memory.

[0047] Exemplary Methods

[0048] See also Figure 1 In an exemplary embodiment, a diagnosis recommendation method is provided. Figure 1 As shown in the figure, the process of the recommended diagnostic method mainly includes:

[0049] Step 101, determine the target-specific diagnosis corresponding to the patient's region.

[0050] Among them, the target-specific diagnosis refers to the historical diagnosis in the patient's region that is distinguishable from other regions.

[0051] In some embodiments, other regions refer to regions in a preset region set other than the region where the patient is located.

[0052] In an exemplary embodiment, the preset region set may include all regions of the entire country, the preset region set may also include all regions of a province, and the preset region set may also include all regions of a city. It can be set according to needs, and this application does not limit this.

[0053] In an exemplary embodiment, the region may be divided into provinces, cities, or counties. It may be set according to needs, and this application does not limit this.

[0054] In some embodiments, Figure 2 As shown, step 101 includes:

[0055] Step 201 , performing the following operations for any historical diagnosis in the area where the patient is located: determining the occurrence frequency of any historical diagnosis in the area where the patient is located.

[0056] In the exemplary embodiment, TF-IDF is a commonly used weighted technology for information retrieval and data mining. TF (Term Frequency) is the word frequency, and IDF (Inverse Document Frequency) is the inverse document frequency index. The main idea of ​​TF-IDF is: if a word or phrase appears frequently in an article and rarely appears in other articles, it is considered that this word or phrase has good distinguishing ability.

[0057] In an exemplary embodiment, in an auxiliary diagnosis scenario, a region is analogized to an article, and all historical diagnoses in the region are analogized to all words in the article. Therefore, when the TF-IDF weight value of a historical diagnosis is high, it means that this historical diagnosis is significant in the current region and can be used as a unique diagnosis corresponding to the region.

[0058] In the exemplary embodiment, the region where the patient is located is represented by d, any historical diagnosis in the region where the patient is located is represented by t, and the frequency of occurrence of any historical diagnosis in the region where the patient is located is represented by TF(t, d). The specific calculation formula of TF(t, d) is as follows:

[0059]

[0060] That is, determining the frequency of occurrence of any historical diagnosis in the patient's area includes: dividing the number of occurrences of any historical diagnosis in the patient's area by the total number of all historical diagnoses in the patient's area, and determining the quotient as the frequency of occurrence of any historical diagnosis in the patient's area.

[0061] Step 202 : determining an inverse regional frequency corresponding to any historical diagnosis based on the total number of regions in the preset regional set and the number of regions in the preset regional set that contain any historical diagnosis.

[0062] In the exemplary embodiment, the region where the patient is located is represented by d, any historical diagnosis in the region where the patient is located is represented by t, the preset region set is represented by D, and the inverse region frequency corresponding to any historical diagnosis is represented by IDF(t,D). The specific calculation formula of IDF(t,D) is as follows:

[0063]

[0064] That is, step 202 includes: taking the sum of the number of regions in the preset region set that contain any historical diagnosis plus the first preset value as the divisor, taking the total number of regions in the preset region set as the dividend, and determining the quotient obtained as the target value; and taking the sum of the logarithm of the target value and the second preset value as the inverse region frequency corresponding to any historical diagnosis. The first preset value and the second preset value can be set as needed, for example, the first preset value is 1, and the second preset value is 1. As needed, the first preset value can be set to other values, and the second preset value can also be set to other values. The first preset value and the second preset value are not necessarily equal, and the present application does not limit the values ​​of the first preset value and the second preset value.

[0065] Step 203, based on the occurrence frequency and the inverse regional frequency, determine the correlation between any historical diagnosis and the region where the patient is located.

[0066] In the exemplary embodiment, the region where the patient is located is represented by d, any historical diagnosis in the region where the patient is located is represented by t, the preset region set is represented by D, the frequency of occurrence of any historical diagnosis in the region where the patient is located is represented by TF(t,d), the inverse regional frequency corresponding to any historical diagnosis is represented by IDF(t,D), the correlation between any historical diagnosis and the region where the patient is located is represented by TF-IDF(t,d,D), and the specific calculation formula of TF-IDF(t,d,D) is as follows:

[0067] TF―IDF(t,d,D)=TF(t,d)×IDF(t,D)

[0068] That is, step 203 includes: multiplying the occurrence frequency by the inverse regional frequency to obtain a product, and determining the correlation between any historical diagnosis and the region where the patient is located.

[0069] Step 204 , based on the correlations corresponding to the various historical diagnoses in the patient's area, determine a target-specific diagnosis corresponding to the patient's area from the various historical diagnoses in the patient's area.

[0070] In an exemplary embodiment, for any historical diagnosis in the area where the patient is located, after determining the correlation between any historical diagnosis and the area where the patient is located, the correlation corresponding to each historical diagnosis in the area where the patient is located is obtained.

[0071] In an exemplary embodiment, step 204 may include: from various historical diagnoses in the patient's area, determining a historical diagnosis with a correlation greater than a preset correlation threshold as a target-specific diagnosis corresponding to the patient's area; step 204 may also include: from various historical diagnoses in the patient's area, sorting the correlations from large to small, and determining the historical diagnoses corresponding to the top N correlations after sorting as the target-specific diagnosis corresponding to the patient's area, where N is a positive integer.

[0072] The following operations are performed for any historical diagnosis in the patient's region: determine the frequency of occurrence of any historical diagnosis in the patient's region, determine the inverse regional frequency corresponding to any historical diagnosis based on the total number of regions in the preset regional set and the number of regions in the preset regional set that contain any historical diagnosis, determine the correlation between any historical diagnosis and the patient's region based on the frequency of occurrence and the inverse regional frequency, determine the target-specific diagnosis corresponding to the patient's region from each historical diagnosis in the patient's region based on the correlation corresponding to each historical diagnosis in the patient's region, draw on the idea of ​​TF-IDF to determine the historical diagnosis in the patient's region that is distinguishable from other regions, which is of significant significance for some regional rare diseases and high-value diseases, and effectively helps clinicians control the risks of regional rare diseases and high-value diseases, and improves the accuracy of diagnosis recommendations compared to diagnosis recommendations based only on patient condition information.

[0073] In other embodiments, step 101 includes: obtaining a specific diagnosis corresponding to each region in a preset region set; and searching for a target specific diagnosis corresponding to the region where the patient is located based on the region where the patient is located.

[0074] The unique diagnosis corresponding to each region in the preset region set may be determined in advance and stored. Each time the diagnosis recommendation method in the present application is executed, the target unique diagnosis corresponding to the patient's region may be directly searched based on the patient's region, thereby improving the efficiency of diagnosis recommendation.

[0075] In an exemplary embodiment, the process of predetermining the unique diagnosis corresponding to each region in the preset region set can refer to the processing process from step 201 to step 204, which will not be described in detail here.

[0076] In other embodiments, step 101 includes: determining a high-frequency historical diagnosis set corresponding to each region in a preset region set; wherein the high-frequency historical diagnosis set corresponding to any region refers to a set of historical diagnoses whose occurrence frequency in any region is greater than an occurrence frequency threshold; and determining a historical diagnosis that is in the high-frequency historical diagnosis set corresponding to the patient's region and not in the high-frequency historical diagnosis sets corresponding to other regions as a target-specific diagnosis corresponding to the patient's region.

[0077] Step 102, determine a target historical medical record that is similar to the patient's medical record to be diagnosed, and obtain a target historical diagnosis corresponding to the target historical medical record.

[0078] In some embodiments, Figure 3 As shown, in step 102, determining the target historical medical record similar to the patient's medical record to be diagnosed includes:

[0079] Step 301, extracting key clinical information from the patient's medical record to be diagnosed.

[0080] In an exemplary embodiment, the medical record to be diagnosed generally includes information such as the chief complaint, current medical history, past medical history, auxiliary examinations, physical examination, age, and gender.

[0081] In an exemplary embodiment, key clinical information mainly includes symptoms, positive signs, positive examination results, positive test results, medication, surgery, etc.

[0082] In an exemplary embodiment, step 301 includes: based on the patient's medical records to be diagnosed, obtaining key clinical information in the patient's medical records to be diagnosed through an information extraction big model.

[0083] In an exemplary embodiment, the information extraction large model may be a large language model pre-trained based on a large amount of medical texts.

[0084] In an exemplary embodiment, model instructions can be written in combination with professional medical knowledge. A complete model instruction generally includes a task description and a pair of inputs and outputs. A small number of examples can also be added to help the model better understand the task requirements.

[0085] Step 302: Determine a first vector of key clinical information and a second vector corresponding to each historical medical record.

[0086] In an exemplary embodiment, each historical medical record may refer to each historical medical record in a historical medical record library.

[0087] In an exemplary embodiment, the second vector corresponding to each historical medical record can be calculated in advance and stored. Each time step 302 is executed, it is only necessary to calculate the first vector of the key clinical information and read the second vector corresponding to each stored historical medical record, which can improve the efficiency of determining the target historical medical record similar to the patient's medical record to be diagnosed.

[0088] In an exemplary embodiment, determining the first vector of key clinical information may be performing text semantic encoding using a deep learning model to obtain the first vector of key clinical information.

[0089] Step 303: Determine a first similarity between the first vector and each second vector.

[0090] In an exemplary embodiment, the first similarity may refer to semantic similarity, which may be calculated by cosine similarity or other similarity calculation methods, and the present application is not limited thereto.

[0091] Step 304: Based on each first similarity, determine a target historical medical record similar to the patient's medical record to be diagnosed from each historical medical record.

[0092] In an exemplary embodiment, it is possible to compare whether the key clinical information and each historical medical record are semantically similar through each first similarity, even if they use different vocabulary or expressions. Therefore, it is possible to determine from the various historical medical records a target historical medical record that is similar to the patient's medical record to be diagnosed, and the target historical medical record and the patient's medical record to be diagnosed have similar clinical manifestations.

[0093] In an exemplary embodiment, step 304 may include: from each historical medical record, determining the historical medical record with a first similarity greater than a similarity threshold as a target historical medical record similar to the patient's medical record to be diagnosed; step 304 may also include: from each historical medical record, sorting each first similarity from large to small, and determining the historical medical records corresponding to the first S first similarities after sorting as the target historical medical record similar to the patient's medical record to be diagnosed, where S is a positive integer. For example, S may be 50, and of course S may also be other values, which are not limited by the present application.

[0094] Extract key clinical information from the patient's medical record to be diagnosed, determine the first vector of the key clinical information and the second vector corresponding to each historical medical record, determine the first similarity between the first vector and each second vector, and based on each first similarity, determine the target historical medical record that is similar to the patient's medical record to be diagnosed from each historical medical record. The target historical medical record and the patient's medical record to be diagnosed have similar clinical manifestations, and the target historical diagnosis corresponding to the target historical medical record is more valuable for reference. Based on the target historical diagnosis and the target unique diagnosis, determine the candidate diagnosis, which can more accurately locate the patient's condition and improve the accuracy of diagnosis recommendations.

[0095] In other embodiments, determining the target historical medical record that is similar to the patient's medical record to be diagnosed in step 102 includes: determining a third vector of the patient's medical record to be diagnosed and a second vector corresponding to each historical medical record; determining a fourth similarity between the third vector and each second vector; and determining a target historical medical record that is similar to the patient's medical record to be diagnosed from each historical medical record based on each fourth similarity.

[0096] In other embodiments, determining the target historical medical records that are similar to the patient's medical records to be diagnosed in step 102 includes: extracting key clinical information from the patient's medical records to be diagnosed; splicing the key clinical information with the patient's medical records to be diagnosed to obtain a spliced ​​text; determining a fourth vector of the spliced ​​text and a second vector corresponding to each historical medical record; determining a fifth similarity between the fourth vector and each second vector; and determining a target historical medical record that is similar to the patient's medical record to be diagnosed from each historical medical record based on each fifth similarity.

[0097] Step 103, determining a candidate diagnosis based on the target historical diagnosis and the target-specific diagnosis.

[0098] In an exemplary embodiment, step 103 may include: merging the target historical diagnosis and the target specific diagnosis to obtain a candidate diagnosis set; wherein the candidate diagnosis set includes each candidate diagnosis.

[0099] Step 104: Determine the diagnosis recommendation result corresponding to the medical record to be diagnosed based on the candidate diagnosis.

[0100] In some embodiments, step 104 includes: determining the sixth similarity between the medical record to be diagnosed and each candidate diagnoses; based on each sixth similarity, determining a recommended diagnosis from each candidate diagnoses; sorting the recommended diagnoses from large to small according to the sixth similarity, and obtaining a diagnosis recommendation result corresponding to the medical record to be diagnosed.

[0101] In other embodiments, Figure 4 As shown, step 104 includes:

[0102] Step 401, determining the target disease knowledge fragment corresponding to the candidate diagnosis.

[0103] In some embodiments, step 401 includes: searching from various disease knowledge fragments based on keywords in the candidate diagnosis to obtain the target disease knowledge fragment corresponding to the candidate diagnosis.

[0104] In other embodiments, Figure 5 As shown, step 401 includes:

[0105] Step 501, obtaining various disease knowledge fragments related to the candidate diagnosis.

[0106] In an exemplary embodiment, step 501 may include: searching for various disease knowledge fragments related to the candidate diagnosis in disease knowledge bases from multiple sources.

[0107] In an exemplary embodiment, various disease knowledge fragments related to the candidate diagnosis may be searched through means such as keyword search and similarity calculation.

[0108] In an exemplary embodiment, the content of the disease knowledge segment may include a description of common symptoms of the disease, treatment methods, etc.

[0109] Step 502: Determine the second similarity between the key clinical information and each disease knowledge fragment.

[0110] In an exemplary embodiment, the second similarity may refer to semantic similarity, which may be calculated by cosine similarity or other similarity calculation methods, and the present application is not limited thereto.

[0111] Step 503: Based on each second similarity, determine the target disease knowledge segment corresponding to the candidate diagnosis from each disease knowledge segment.

[0112] Obtain various disease knowledge fragments related to the candidate diagnosis, determine the second similarities between the key clinical information and each disease knowledge fragment, and based on each second similarity, determine the target disease knowledge fragment corresponding to the candidate diagnosis from each disease knowledge fragment. This can accurately obtain the target disease knowledge fragment that is semantically similar to the candidate diagnosis, and based on the target disease knowledge fragment, assist in determining the diagnosis recommendation result corresponding to the medical record to be diagnosed, thereby improving the accuracy of the diagnosis recommendation.

[0113] In some embodiments, Figure 6 As shown, step 503 includes:

[0114] Step 601, sorting the second similarities from large to small, and determining the disease knowledge fragments corresponding to the first M second similarities after sorting as the initial disease knowledge fragments corresponding to the candidate diagnosis.

[0115] Wherein, M is a positive integer.

[0116] For example, M may be 3. Of course, M may also be other values, and this application is not limited to this.

[0117] Step 602: determine the initial disease knowledge fragments with a second similarity greater than a threshold as the precise disease knowledge fragments corresponding to the candidate diagnosis.

[0118] Step 603: divide the initial disease knowledge segments whose second similarity is less than or equal to the threshold into initial disease knowledge sub-segments.

[0119] In an exemplary embodiment, step 603 may include: dividing the initial disease knowledge segment with the second similarity less than or equal to the threshold into each initial disease knowledge sub-segment according to the preset text length. For example, the preset text length may be one sentence, so that the segmentation at the sentence level is realized, and one initial disease knowledge sub-segment is one sentence; the preset text length may be two sentences, and one initial disease knowledge sub-segment includes two sentences; the preset text length may also be 100 characters, and one initial disease knowledge sub-segment includes 100 characters; of course, there are other situations, which will not be listed one by one, and the present application does not limit this.

[0120] Step 604 , determining a third similarity between the key clinical information and each initial disease knowledge sub-segment.

[0121] In an exemplary embodiment, the third similarity may refer to semantic similarity, which may be calculated by cosine similarity or other similarity calculation methods, and the present application is not limited thereto.

[0122] Step 605 : Based on each third similarity, determine the precise disease knowledge segment corresponding to the candidate diagnosis from each initial disease knowledge sub-segment.

[0123] In an exemplary embodiment, step 605 may include: from each initial disease knowledge sub-segment, determining the initial disease knowledge sub-segment with a third similarity greater than a preset similarity as the precise disease knowledge segment corresponding to the candidate diagnosis; step 605 may also include: from each initial disease knowledge sub-segment, sorting each third similarity from large to small, and determining the initial disease knowledge sub-segments corresponding to the first T third similarities after sorting as the precise disease knowledge segment corresponding to the candidate diagnosis, where T is a positive integer. For example, T may be 3, and of course T may also be other values, which are not limited by the present application.

[0124] Step 606 , based on the precise disease knowledge fragments corresponding to the candidate diagnosis, determine the target disease knowledge fragments corresponding to the candidate diagnosis.

[0125] In an exemplary embodiment, step 606 may include: directly determining the precise disease knowledge fragment corresponding to the candidate diagnosis as the target disease knowledge fragment corresponding to the candidate diagnosis; step 606 may also include: determining the seventh similarity between the candidate diagnosis and each precise disease knowledge fragment corresponding to the candidate diagnosis; based on each seventh similarity, determining the target disease knowledge fragment corresponding to the candidate diagnosis from each precise disease knowledge fragment corresponding to the candidate diagnosis.

[0126] In an exemplary embodiment, based on each seventh similarity, determining the target disease knowledge fragment corresponding to the candidate diagnosis from each precise disease knowledge fragment corresponding to the candidate diagnosis may include: sorting each seventh similarity from large to small from each precise disease knowledge fragment corresponding to the candidate diagnosis, and determining the precise disease knowledge fragments corresponding to the first L seventh similarities after sorting as the target disease knowledge fragment corresponding to the candidate diagnosis, where L is a positive integer. For example, L may be 3, and of course L may also be other values, which are not limited by the present application.

[0127] For example, the second similarities are sorted from large to small, and the disease knowledge segments corresponding to the first three second similarities after sorting are determined as the initial disease knowledge segments corresponding to the candidate diagnosis. If the second similarities corresponding to the three initial disease knowledge segments are all greater than the threshold, the three initial disease knowledge segments are directly determined as the precise disease knowledge segments corresponding to the candidate diagnosis.

[0128] If the second similarities corresponding to two of the three initial disease knowledge fragments are greater than the threshold, these two initial disease knowledge fragments are directly determined as the precise disease knowledge fragments corresponding to the candidate diagnosis. If the second similarity corresponding to one initial disease knowledge fragment is less than or equal to the threshold, this one initial disease knowledge fragment is divided into initial disease knowledge sub-fragments, and the third similarities of each initial disease knowledge sub-fragment are sorted from large to small. The initial disease knowledge sub-fragments corresponding to the first three third similarities after sorting are determined as the precise disease knowledge fragments corresponding to the candidate diagnosis, and a total of 2+1*3=5 precise disease knowledge fragments are obtained.

[0129] If the second similarity corresponding to one of the three initial disease knowledge fragments is greater than the threshold, this initial disease knowledge fragment is directly determined as the precise disease knowledge fragment corresponding to the candidate diagnosis. If the second similarities corresponding to two initial disease knowledge fragments are less than or equal to the threshold, each initial disease knowledge fragment obtains three precise disease knowledge fragments, and a total of 1+2*3=7 precise disease knowledge fragments are obtained.

[0130] If the second similarities corresponding to three of the three initial disease knowledge fragments are less than or equal to the threshold, each initial disease knowledge fragment obtains three precise disease knowledge fragments, and a total of 3*3=9 precise disease knowledge fragments are obtained.

[0131] Therefore, the number of precise disease knowledge fragments corresponding to one candidate diagnosis may be 3, 5, 7, or 9. Among these precise disease knowledge fragments, the seventh similarities are sorted from large to small, and the precise disease knowledge fragments corresponding to the first three seventh similarities after sorting are determined as the target disease knowledge fragments corresponding to the candidate diagnosis. Therefore, the number of precise disease knowledge fragments corresponding to each candidate diagnosis is 3.

[0132] In some embodiments, the diagnosis recommendation method further includes: recording the knowledge source and index to which the target disease knowledge fragment corresponding to the candidate diagnosis belongs, so as to facilitate subsequent tracing and increase credibility.

[0133] Sort the second similarities from large to small, determine the disease knowledge fragments corresponding to the first M second similarities after sorting as the initial disease knowledge fragments corresponding to the candidate diagnosis, determine the initial disease knowledge fragments with second similarities greater than the threshold as the precise disease knowledge fragments corresponding to the candidate diagnosis, divide the initial disease knowledge fragments with second similarities less than or equal to the threshold into initial disease knowledge sub-fragments, determine the third similarities between the key clinical information and the initial disease knowledge sub-fragments, determine the precise disease knowledge fragments corresponding to the candidate diagnosis from the initial disease knowledge sub-fragments based on the third similarities, determine the target disease knowledge fragments corresponding to the candidate diagnosis based on the precise disease knowledge fragments corresponding to the candidate diagnosis, and realize multiple rounds of precise knowledge positioning from the document to the fragment and then to the sub-fragment level. In each round of knowledge positioning, perform semantic similarity matching and re-sorting in combination with the key clinical information in the patient's medical record to be diagnosed, and finally obtain the target disease knowledge fragments with a higher matching degree to assist in diagnostic reasoning.

[0134] In other embodiments, step 503 includes: sorting the second similarities from large to small, and determining the disease knowledge fragments corresponding to the first M second similarities after sorting as the target disease knowledge fragments corresponding to the candidate diagnosis. Wherein, M is a positive integer. For example, M can be 3, and of course M can also be other values, which are not limited by the present application.

[0135] Step 402: extract key clinical information from the patient's medical record to be diagnosed, and divide the key clinical information into various clinical information.

[0136] In an exemplary embodiment, key clinical information is divided into individual clinical information. There may be multiple symptoms in the key clinical information, each symptom can be used as a clinical information, there may be multiple positive test results, each positive test result can be used as a clinical information, and so on. One clinical information is the data with the smallest granularity.

[0137] Step 403: Determine the correlation between each clinical information and the candidate diagnosis.

[0138] In some embodiments, step 403 includes: inputting each clinical information and candidate diagnosis into a correlation determination model to obtain the correlation between each clinical information and candidate diagnosis.

[0139] The correlation determination model is obtained based on sample data training, and the sample data includes clinical information of each sample, diagnosis of each sample, and the correlation between the clinical information of the sample and the diagnosis of the sample.

[0140] The clinical information of each sample is obtained by dividing the key clinical information of each sample in the medical record of each sample. The correlation between the clinical information of the sample and the diagnosis of the sample is determined based on the number of co-occurrences of the clinical information of the sample and the diagnosis of the sample.

[0141] In an exemplary embodiment, a large number of online medical records are obtained as sample medical records, and sample diagnoses corresponding to the sample medical records are obtained, sample key clinical information in the sample medical records is extracted, and the sample key clinical information is divided to obtain individual sample clinical information, and each sample clinical information is associated with the sample diagnosis corresponding to the sample medical record, and the number of co-occurrences is counted. After counting the number of sample medical records, the association between the complete sample clinical information and the sample diagnosis is obtained. For example, the number of co-occurrences of the sample clinical information and the sample diagnosis can be directly determined as the correlation between the sample clinical information and the sample diagnosis; or the correlation between the sample clinical information and the sample diagnosis whose co-occurrences are greater than the co-occurrence threshold is determined to have correlation, and the correlation between the sample clinical information and the sample diagnosis whose co-occurrences are less than or equal to the co-occurrence threshold is determined to have no correlation.

[0142] In the exemplary embodiment, the correlation determination model is constructed based on the large language model, and the correlation between each sample clinical information, each sample diagnosis, and the correlation between the sample clinical information and the sample diagnosis is fine-tuned to make the correlation determination model have a better correlation effect on clinical information and diseases. Moreover, a large amount of online medical record data is used to automatically construct a large language model instruction fine-tuning data set, and high-quality training data is constructed without relying on manually constructed data, which effectively improves the task effect and optimizes the effect of the task of associating clinical information and diseases.

[0143] Input each clinical information and candidate diagnosis into the correlation determination model to obtain the correlation between each clinical information and candidate diagnosis. Use the large model to explicitly construct the correlation between each clinical information and each candidate diagnosis, simulate the doctor's thinking process of inferring the patient's possible disease based on the patient's condition during the clinical consultation, and explicitly construct the correlation between the patient's clinical information and the disease, so as to more effectively generate the subsequent diagnostic reasoning thinking process. Before using the model for diagnostic reasoning, explicitly construct the correlation between each clinical information and each candidate diagnosis, simulate the doctor's thinking process of inferring the patient's possible disease based on the patient's condition during the clinical consultation, effectively reduce the difficulty of the model's diagnostic reasoning task and ensure the accuracy of the diagnostic reasoning process.

[0144] Theoretically, large language models usually implicitly have the ability to construct clinical information and diagnostic correlations, that is, they can output diagnostic reasoning results end-to-end. However, if such information can be explicitly constructed in advance and used as input to the large model, it can often significantly reduce the difficulty of the task to be handled by the model, thereby improving the task effect. In this task, if a symptom or sign conforms to the common clinical manifestations of the disease, or a certain examination can be used to diagnose or exclude a certain disease, and a certain drug or surgery can be used for the clinical treatment of a certain disease, both can be considered to be related.

[0145] In other embodiments, step 403 includes: directly based on each clinical information and candidate diagnosis, obtaining the correlation between each clinical information and candidate diagnosis through a pre-trained large language model.

[0146] Step 404, based on the candidate diagnoses, the target disease knowledge fragments, the correlation between each clinical information and the candidate diagnoses, and the patient's medical records to be diagnosed, determine the diagnosis recommendation results corresponding to the medical records to be diagnosed.

[0147] Based on the target historical diagnosis and target-specific diagnosis, candidate diagnoses are determined. When making diagnosis recommendations, not only the target historical diagnosis corresponding to the target historical medical records that are similar to the patient's medical records to be diagnosed, but also the target-specific diagnosis corresponding to the patient's region is considered. This can more accurately locate the patient's condition, which is of significant significance for some regional rare diseases and high-value diseases. It effectively helps clinicians control the risks of regional rare diseases and high-value diseases, and improves the accuracy of diagnosis recommendations compared to diagnosis recommendations based only on patient condition information.

[0148] The target disease knowledge fragment can assist in strictly following professional medical knowledge in the diagnosis recommendation process and avoid medical logic errors. It can also provide knowledge traceability and facilitate the review and study of relevant literature and materials, greatly improving the professionalism and credibility of the output diagnosis recommendation results.

[0149] The explicit correlation between each clinical information and candidate diagnosis can effectively ensure that the correct disease knowledge is referenced during the reasoning process, effectively avoiding the logical error problem of irrelevance between symptoms and diseases.

[0150] Finally, by integrating the information from the above multiple sources, the diagnostic recommendation results corresponding to the medical records to be diagnosed are determined, which improves the accuracy and credibility of the diagnostic recommendations.

[0151] In some embodiments, step 404 includes: based on the candidate diagnoses, target disease knowledge fragments, the correlation between various clinical information and candidate diagnoses, and the patient's medical records to be diagnosed, obtaining diagnosis recommendation results corresponding to the medical records to be diagnosed through a diagnosis recommendation big model.

[0152] The diagnosis recommendation result includes a disease reasoning chain and at least one recommended diagnosis sorted from most likely to least likely.

[0153] When making diagnostic recommendations, the target-specific diagnosis corresponding to the patient's region is taken into consideration, which can more accurately locate the patient's condition. This is of significant significance for some regional rare diseases and high-value diseases, and can effectively help clinicians control the risks of regional rare diseases and high-value diseases. Compared with making diagnostic recommendations based only on patient condition information, the accuracy of diagnostic recommendations is improved.

[0154] The target disease knowledge fragments can assist the model to strictly follow professional medical knowledge in the process of generating disease reasoning thinking chains, avoiding medical logic errors. At the same time, they can also provide knowledge traceability, facilitate the review and study of relevant literature and materials, and greatly improve the professionalism and credibility of the output results.

[0155] The explicit correlation between each clinical information and candidate diagnosis can effectively ensure that the correct disease knowledge is referenced during the reasoning process, effectively avoiding the logical error problem of irrelevance between symptoms and diseases.

[0156] Ultimately, by integrating information from the above multiple sources and using a large diagnosis recommendation model to generate a disease reasoning chain and at least one recommended diagnosis ranked from most likely to least likely, the accuracy and credibility of the diagnosis recommendation are improved.

[0157] In an exemplary embodiment, in the process of outputting the disease reasoning thinking chain, the referenced target disease knowledge fragments will be indexed (labeled with knowledge serial numbers), and at the end, the most appropriate up to 5 recommended diagnoses (can be less than 5) will be given according to the probability.

[0158] For example, a sample output of a diagnosis recommendation result is as follows:

[0159] "The patient complained of recurrent headaches for more than a month. His medical history described a feeling of a hooded head, dizziness, and neck soreness and discomfort. He slept well at night and his symptoms recurred. Tension headaches are of various natures and can be bilateral or unilateral. They are mainly dull pains around the head, like a band pressing down, without pulsation. There is also a feeling of heaviness or tightness on the top of the head. There is also tightness and stiffness in the occipital region, which is particularly obvious when turning the neck. It may be accompanied by dizziness and other symptoms [1]. In addition, the patient also showed neck soreness and discomfort, which is consistent with the general symptoms of vertebral artery type cervical spondylosis. Vertebral artery type cervical spondylosis can also cause headaches, which are mostly manifested as migraines, mainly in the posterior occipital and temporal regions, and are often throbbing or stabbing [2]. Based on the above information, the patient's symptoms are consistent with the clinical manifestations of tension headache and vertebral artery type cervical spondylosis. Therefore, the diagnosis of tension headache and vertebral artery type cervical spondylosis is considered."

[0160] By adopting a multi-step reasoning strategy, the complex diagnostic reasoning task is split into more specific and simpler subtasks, and then the region-specific information is mined in combination with the patient's region. At the same time, the disease knowledge is accurately positioned to help generate a professional medical reasoning process and give a recommended diagnosis. Finally, the diagnostic reasoning thinking process is integrated to improve the accuracy of diagnostic recommendations. It can assist clinicians in making judgments faster and more accurately, reduce misdiagnosis and missed diagnosis in clinical consultation scenarios, effectively improve the disease diagnosis effect in clinical medical consultation scenarios, and provide corresponding reference knowledge to improve the credibility of disease diagnosis.

[0161] In other embodiments, step 404 includes: determining the correlation between the patient's medical records to be diagnosed and each candidate diagnoses based on the patient's medical records to be diagnosed, candidate diagnoses, and the correlation between each clinical information and candidate diagnoses; determining a recommended diagnosis from each candidate diagnoses based on the correlation between the patient's medical records to be diagnosed and each candidate diagnoses and the target disease knowledge fragment; and generating a diagnosis recommendation result corresponding to the medical records to be diagnosed based on the recommended diagnosis.

[0162] In summary, in this application, the target-specific diagnosis corresponding to the patient's region is determined, wherein the target-specific diagnosis refers to a historical diagnosis in the patient's region that is distinguishable from other regions, and the target historical medical records similar to the patient's medical records to be diagnosed are determined, and the target historical diagnosis corresponding to the target historical medical records is obtained. Based on the target historical diagnosis and the target-specific diagnosis, the candidate diagnosis is determined, and based on the candidate diagnosis, the diagnosis recommendation result corresponding to the medical record to be diagnosed is determined. When making a diagnosis recommendation, not only the target historical diagnosis corresponding to the target historical medical records similar to the patient's medical records to be diagnosed is considered, but also the target-specific diagnosis corresponding to the patient's region is considered, which can more accurately locate the patient's condition, which is of significant significance for some regional rare diseases and high-value diseases, and effectively helps clinicians to control the risks of regional rare diseases, high-value diseases, etc., and improves the accuracy of diagnosis recommendations compared to diagnosis recommendations based only on patient condition information.

[0163] Exemplary Devices

[0164] Accordingly, the present application also provides a diagnosis recommendation device, such as Figure 7 As shown, the recommended diagnostic devices include:

[0165] The first processing unit 701 is used to determine a target-specific diagnosis corresponding to the region where the patient is located; wherein the target-specific diagnosis refers to a historical diagnosis in the region where the patient is located that is distinguishable from other regions;

[0166] The second processing unit 702 is used to determine a target historical medical record similar to the patient's medical record to be diagnosed, and obtain a target historical diagnosis corresponding to the target historical medical record;

[0167] A third processing unit 703 is used to determine a candidate diagnosis based on the target historical diagnosis and the target specific diagnosis;

[0168] The recommendation unit 704 is used to determine the diagnosis recommendation result corresponding to the medical record to be diagnosed based on the candidate diagnosis.

[0169] Optionally, the other regions refer to regions in a preset region set other than the region where the patient is located;

[0170] The first processing unit 701 is specifically configured to:

[0171] For any historical diagnosis in the patient's region, do the following:

[0172] Determining the frequency of occurrence of any one of the historical diagnoses in the region where the patient is located;

[0173] Determine the inverse regional frequency corresponding to any one of the historical diagnoses based on the total number of regions in the preset regional set and the number of regions in the preset regional set that contain any one of the historical diagnoses;

[0174] Determine the correlation between any one of the historical diagnoses and the region where the patient is located based on the occurrence frequency and the inverse region frequency;

[0175] Based on the respective corresponding correlations of the historical diagnoses in the area where the patient is located, a target-specific diagnosis corresponding to the area where the patient is located is determined from the historical diagnoses in the area where the patient is located.

[0176] Optionally, the second processing unit 702 is specifically configured to:

[0177] Extract key clinical information from the patient's medical records to be diagnosed;

[0178] Determine a first vector of the key clinical information and a second vector corresponding to each historical medical record;

[0179] determining a first similarity between the first vector and each of the second vectors;

[0180] Based on each of the first similarities, a target historical medical record similar to the patient's medical record to be diagnosed is determined from each of the historical medical records.

[0181] Optionally, the recommendation unit 704 includes:

[0182] A first processing subunit, configured to determine a target disease knowledge fragment corresponding to the candidate diagnosis;

[0183] A division subunit is used to extract key clinical information from the patient's medical record to be diagnosed, and divide the key clinical information into individual clinical information;

[0184] A second processing subunit, configured to determine the degree of association between each clinical information and the candidate diagnosis;

[0185] The third processing subunit is used to determine the diagnosis recommendation result corresponding to the medical record to be diagnosed based on the candidate diagnosis, the target disease knowledge fragment, the correlation between each clinical information and the candidate diagnosis, and the patient's medical record to be diagnosed.

[0186] Optionally, the first processing subunit is specifically configured to:

[0187] Obtaining disease knowledge fragments related to the candidate diagnosis;

[0188] Determining a second similarity between the key clinical information and each disease knowledge fragment;

[0189] Based on each of the second similarities, a target disease knowledge segment corresponding to the candidate diagnosis is determined from each of the disease knowledge segments.

[0190] Optionally, the first processing subunit is specifically configured to:

[0191] Sort the second similarities from large to small, and determine the disease knowledge fragments corresponding to the first M second similarities after sorting as the initial disease knowledge fragments corresponding to the candidate diagnosis; wherein M is a positive integer;

[0192] Determine the initial disease knowledge fragments whose second similarity is greater than a threshold as the precise disease knowledge fragments corresponding to the candidate diagnosis;

[0193] Divide the initial disease knowledge segments whose second similarity is less than or equal to the threshold into individual initial disease knowledge sub-segments; determine the third similarity between the key clinical information and each initial disease knowledge sub-segment; and determine the precise disease knowledge segment corresponding to the candidate diagnosis from each initial disease knowledge sub-segment based on each third similarity;

[0194] Based on the precise disease knowledge fragment corresponding to the candidate diagnosis, the target disease knowledge fragment corresponding to the candidate diagnosis is determined.

[0195] Optionally, the second processing subunit is specifically configured to:

[0196] Inputting each clinical information and the candidate diagnosis into a correlation determination model to obtain the correlation between each clinical information and the candidate diagnosis;

[0197] The correlation determination model is obtained by training based on sample data, wherein the sample data includes clinical information of each sample, diagnosis of each sample, and the correlation between the clinical information of the sample and the diagnosis of the sample;

[0198] The clinical information of each sample is obtained by dividing the key clinical information of the sample in the medical record of each sample; the correlation between the sample clinical information and the sample diagnosis is determined based on the number of co-occurrences of the sample clinical information and the sample diagnosis.

[0199] Optionally, the third processing subunit is specifically configured to:

[0200] Based on the candidate diagnosis, the target disease knowledge fragment, the correlation between each clinical information and the candidate diagnosis, and the patient's medical record to be diagnosed, a diagnosis recommendation result corresponding to the medical record to be diagnosed is obtained through a diagnosis recommendation big model;

[0201] The diagnosis recommendation result includes a disease reasoning chain and at least one recommended diagnosis sorted from most likely to least likely.

[0202] The diagnosis recommendation device provided in this embodiment belongs to the same application concept as the diagnosis recommendation method provided in the above embodiments of this application, and can execute the diagnosis recommendation method provided in any of the above embodiments of this application, and has the corresponding functional modules and beneficial effects of executing the diagnosis recommendation method. For technical details not fully described in this embodiment, please refer to the specific processing content of the diagnosis recommendation method provided in the above embodiments of this application, and will not be repeated here.

[0203] The functions implemented by the above-mentioned first processing unit 701, second processing unit 702, third processing unit 703 and recommendation unit 704 may be implemented by the same or different processors respectively, which is not limited in the embodiment of the present application.

[0204] It should be understood that the units in the above devices can be implemented in the form of a processor calling software. For example, the device includes a processor, the processor is connected to a memory, and instructions are stored in the memory. The processor calls the instructions stored in the memory to implement any of the above methods or realize the functions of each unit of the device, wherein the processor can be a general-purpose processor, such as a CPU or a microprocessor, etc., and the memory can be a memory in the device or a memory outside the device. Alternatively, the units in the device can be implemented in the form of hardware circuits, and the functions of some or all units can be realized by designing the hardware circuits. The hardware circuit can be understood as one or more processors; for example, in one implementation, the hardware circuit is an ASIC, and the functions of some or all of the above units are realized by designing the logical relationship of the components in the circuit; for another example, in another implementation, the hardware circuit can be implemented by PLD, taking FPGA as an example, which can include a large number of logic gate circuits, and the connection relationship between the logic gate circuits is configured by the configuration file, so as to realize the functions of some or all of the above units. All units of the above devices can be implemented in the form of a processor calling software, or in the form of hardware circuits, or in part by a processor calling software, and the remaining part is implemented in the form of hardware circuits.

[0205] In an embodiment of the present application, a processor is a circuit with the ability to process signals. In one implementation, the processor may be a circuit with the ability to read and run instructions, such as a CPU, a microprocessor, a GPU, or a DSP; in another implementation, the processor may implement certain functions through the logical relationship of a hardware circuit, and the logical relationship of the hardware circuit is fixed or reconfigurable, such as a hardware circuit implemented by an ASIC or PLD, such as an FPGA. In a reconfigurable hardware circuit, the process of the processor loading a configuration document to implement the hardware circuit configuration can be understood as the process of the processor loading instructions to implement the functions of some or all of the above units. In addition, it can also be a hardware circuit designed for artificial intelligence, which can be understood as an ASIC, such as an NPU, TPU, DPU, etc.

[0206] It can be seen that each unit in the above device can be one or more processors (or processing circuits) configured to implement the above method, such as: CPU, GPU, NPU, TPU, DPU, microprocessor, DSP, ASIC, FPGA, or a combination of at least two of these processor forms.

[0207] In addition, all or part of the units in the above device can be integrated together, or can be implemented independently. In one implementation, these units are integrated together and implemented in the form of a SOC. The SOC may include at least one processor for implementing any of the above methods or implementing the functions of each unit of the device. The type of the at least one processor may be different, for example, including a CPU and an FPGA, a CPU and an artificial intelligence processor, a CPU and a GPU, etc.

[0208] Exemplary Electronic Devices

[0209] An embodiment of the present application provides an electronic device, see Figure 8 As shown, the device includes:

[0210] Memory 200 and processor 210;

[0211] The memory 200 is connected to the processor 210 and is used to store programs;

[0212] The processor 210 is used to implement the diagnosis recommendation method disclosed in any of the above embodiments by running the program stored in the memory 200.

[0213] Specifically, the electronic device may further include: a bus, a communication interface 220 , an input device 230 and an output device 240 .

[0214] The processor 210, the memory 200, the communication interface 220, the input device 230 and the output device 240 are connected to each other via a bus.

[0215] A bus may include a pathway that transfers information between components of a computer system.

[0216] The processor 210 may be a general-purpose processor, such as a general-purpose central processing unit (CPU), a microprocessor, etc., or an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the scheme of the present invention. It may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0217] The processor 210 may include a main processor, and may also include a baseband chip, a modem, and the like.

[0218] The memory 200 stores a program for executing the technical solution of the present invention, and may also store an operating system and other key services. Specifically, the program may include a program code, and the program code includes a computer operation instruction. More specifically, the memory 200 may include a read-only memory (ROM), other types of static storage devices that can store static information and instructions, a random access memory (RAM), other types of dynamic storage devices that can store information and instructions, a disk storage, a flash, and the like.

[0219] The input device 230 may include a device for receiving data and information input by a user, such as a keyboard, a mouse, a camera, a scanner, a light pen, a voice input device, a touch screen, a pedometer, or a gravity sensor.

[0220] Output device 240 may include devices that allow information to be output to a user, such as a display screen, printer, speaker, etc.

[0221] The communication interface 220 may include any transceiver or the like to communicate with other devices or communication networks, such as Ethernet, Radio Access Network (RAN), Wireless Local Area Network (WLAN), etc.

[0222] The processor 210 executes the program stored in the memory 200 and calls other devices, which can be used to implement each step of any diagnosis recommendation method provided in the above embodiments of the present application.

[0223] Exemplary computer program products and storage media

[0224] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the diagnostic recommendation method according to various embodiments of the present application described in any of the above embodiments of this specification.

[0225] The computer program product may be written in any combination of one or more programming languages ​​to write program codes for performing the operations of the embodiments of the present application, including object-oriented programming languages, such as Java, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0226] In addition, the embodiment of the present application may also be a storage medium on which a computer program is stored. The computer program is executed by a processor to execute the steps of the diagnosis recommendation method according to various embodiments of the present application described in any of the above embodiments of this specification, and specifically the following steps may be implemented:

[0227] Step 101, determine the target-specific diagnosis corresponding to the patient's region.

[0228] Among them, the target-specific diagnosis refers to the historical diagnosis in the patient's region that is distinguishable from other regions.

[0229] Step 102, determining a target historical medical record that is similar to the patient's medical record to be diagnosed, and obtaining a target historical diagnosis corresponding to the target historical medical record.

[0230] Step 103, determining a candidate diagnosis based on the target historical diagnosis and the target-specific diagnosis.

[0231] Step 104: Determine the diagnosis recommendation result corresponding to the medical record to be diagnosed based on the candidate diagnosis.

[0232] For the aforementioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the order of the actions described, because according to the present application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0233] It should be noted that each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the embodiments can be referred to each other. For the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0234] The steps in the methods of each embodiment of the present application can be adjusted in order, combined and deleted according to actual needs, and the technical features recorded in each embodiment can be replaced or combined.

[0235] The modules and sub-modules in the devices and terminals in the various embodiments of the present application can be combined, divided and deleted according to actual needs.

[0236] In the several embodiments provided in the present application, it should be understood that the disclosed terminals, devices and methods can be implemented in other ways. For example, the terminal embodiments described above are only schematic, for example, the division of modules or submodules is only a logical function division, and there may be other division methods in actual implementation, for example, multiple submodules or modules can be combined or integrated into another module, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or modules, which can be electrical, mechanical or other forms.

[0237] The modules or submodules described as separate components may or may not be physically separated, and the components of the modules or submodules may or may not be physical modules or submodules, that is, they may be located in one place, or they may be distributed on multiple network modules or submodules. Some or all of the modules or submodules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0238] In addition, each functional module or submodule in each embodiment of the present application may be integrated into one processing module, or each module or submodule may exist physically separately, or two or more modules or submodules may be integrated into one module. The above-mentioned integrated modules or submodules may be implemented in the form of hardware or in the form of software functional modules or submodules.

[0239] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0240] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly by hardware, software units executed by a processor, or a combination of the two. The software units may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0241] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0242] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A diagnosis recommendation method, characterized in that: include: Determine a target-specific diagnosis corresponding to the patient's region; wherein the target-specific diagnosis refers to a historical diagnosis in the patient's region that is distinguishable from other regions; Determine a target historical medical record that is similar to the patient's medical record to be diagnosed, and obtain a target historical diagnosis corresponding to the target historical medical record; Determining a candidate diagnosis based on the target historical diagnosis and the target-specific diagnosis; Based on the candidate diagnoses, a diagnosis recommendation result corresponding to the medical record to be diagnosed is determined.

2. The diagnosis recommendation method according to claim 1, characterized in that: The other regions refer to the regions in the preset region set except the region where the patient is located; Determining the target-specific diagnosis corresponding to the patient's region includes: For any historical diagnosis in the patient's region, do the following: Determining the frequency of occurrence of any one of the historical diagnoses in the region where the patient is located; Determine the inverse regional frequency corresponding to any one of the historical diagnoses based on the total number of regions in the preset regional set and the number of regions in the preset regional set that contain any one of the historical diagnoses; Determine the correlation between any one of the historical diagnoses and the region where the patient is located based on the occurrence frequency and the inverse region frequency; Based on the respective correlations of the historical diagnoses in the area where the patient is located, a target-specific diagnosis corresponding to the area where the patient is located is determined from the historical diagnoses in the area where the patient is located.

3. The diagnosis recommendation method according to claim 1, characterized in that: The target historical medical records that are similar to the patient's medical records to be diagnosed are determined to include: Extract key clinical information from the patient's medical records to be diagnosed; Determine a first vector of the key clinical information and a second vector corresponding to each historical medical record; determining a first similarity between the first vector and each of the second vectors; Based on each of the first similarities, a target historical medical record similar to the patient's medical record to be diagnosed is determined from each of the historical medical records.

4. The diagnosis recommendation method according to claim 1, characterized in that: The step of determining the diagnosis recommendation result corresponding to the medical record to be diagnosed based on the candidate diagnosis includes: Determining a target disease knowledge fragment corresponding to the candidate diagnosis; Extracting key clinical information from the patient's medical record to be diagnosed, and dividing the key clinical information into individual clinical information; Determining the correlation between each clinical information and the candidate diagnosis; Based on the candidate diagnosis, the target disease knowledge fragment, the correlation between each clinical information and the candidate diagnosis, and the patient's medical record to be diagnosed, a diagnosis recommendation result corresponding to the medical record to be diagnosed is determined.

5. The diagnosis recommendation method according to claim 4, characterized in that: The step of determining the target disease knowledge fragment corresponding to the candidate diagnosis includes: Obtaining disease knowledge fragments related to the candidate diagnosis; Determining a second similarity between the key clinical information and each disease knowledge fragment; Based on each of the second similarities, a target disease knowledge segment corresponding to the candidate diagnosis is determined from each of the disease knowledge segments.

6. The diagnosis recommendation method according to claim 5, characterized in that: The determining, based on each of the second similarities, a target disease knowledge segment corresponding to the candidate diagnosis from each of the disease knowledge segments comprises: Sort the second similarities from large to small, and determine the disease knowledge fragments corresponding to the first M second similarities after sorting as the initial disease knowledge fragments corresponding to the candidate diagnosis; wherein M is a positive integer; Determine the initial disease knowledge fragments whose second similarity is greater than a threshold as the precise disease knowledge fragments corresponding to the candidate diagnosis; Divide the initial disease knowledge segments whose second similarity is less than or equal to the threshold into individual initial disease knowledge sub-segments; determine the third similarity between the key clinical information and each initial disease knowledge sub-segment; and determine the precise disease knowledge segment corresponding to the candidate diagnosis from each initial disease knowledge sub-segment based on each third similarity; Based on the precise disease knowledge fragment corresponding to the candidate diagnosis, the target disease knowledge fragment corresponding to the candidate diagnosis is determined.

7. The diagnosis recommendation method according to claim 4, characterized in that: Determining the degree of association between each clinical information and the candidate diagnosis includes: Inputting each clinical information and the candidate diagnosis into a correlation determination model to obtain the correlation between each clinical information and the candidate diagnosis; The correlation determination model is obtained by training based on sample data, wherein the sample data includes clinical information of each sample, diagnosis of each sample, and the correlation between the clinical information of the sample and the diagnosis of the sample; The clinical information of each sample is obtained by dividing the key clinical information of the sample in the medical record of each sample; the correlation between the sample clinical information and the sample diagnosis is determined based on the number of co-occurrences of the sample clinical information and the sample diagnosis.

8. The diagnosis recommendation method according to claim 4, characterized in that: The step of determining the diagnosis recommendation result corresponding to the medical record to be diagnosed based on the candidate diagnosis, the target disease knowledge fragment, the correlation between each clinical information and the candidate diagnosis, and the medical record to be diagnosed of the patient includes: Based on the candidate diagnosis, the target disease knowledge fragment, the correlation between each clinical information and the candidate diagnosis, and the patient's medical record to be diagnosed, a diagnosis recommendation result corresponding to the medical record to be diagnosed is obtained through a diagnosis recommendation big model; The diagnosis recommendation result includes a disease reasoning chain and at least one recommended diagnosis sorted from most likely to least likely.

9. A diagnosis recommendation device, characterized in that: include: A first processing unit is used to determine a target-specific diagnosis corresponding to the region where the patient is located; wherein the target-specific diagnosis refers to a historical diagnosis in the region where the patient is located that is distinguishable from other regions; The second processing unit is used to determine a target historical medical record similar to the patient's medical record to be diagnosed, and obtain a target historical diagnosis corresponding to the target historical medical record; a third processing unit, configured to determine a candidate diagnosis based on the target historical diagnosis and the target-specific diagnosis; The recommendation unit is used to determine the diagnosis recommendation result corresponding to the medical record to be diagnosed based on the candidate diagnosis.

10. An electronic device, characterized in that: including memory and processor; The memory is connected to the processor and is used to store programs; The processor is used to implement the diagnosis recommendation method according to any one of claims 1 to 8 by running the program in the memory.

11. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, the diagnosis recommendation method according to any one of claims 1 to 8 is implemented.

12. A computer program product, characterized in that The method comprises computer program instructions, which, when executed by a processor, enable the processor to perform the diagnosis recommendation method according to any one of claims 1 to 8.

Images