Initialization report determination method and device, storage medium and electronic equipment

Abstract

The invention discloses an initialization report determination method and device, a storage medium and electronic equipment, and the method comprises the steps: obtaining a to-be-diagnosed image and each candidate report, and determining a disease tag of the to-be-diagnosed image from disease tags of each candidate report according to the to-be-diagnosed image, and enabling the disease tag to serve as a target disease tag. And determining an initialized report from the candidate reports according to the target disease tag. And determining an initialized report according to each candidate report of which the disease tag is the target disease tag. The initialization report in the specification is determined according to the disease label, so that the disease corresponding to the initialization report in the specification is similar to the disease reflected by the image to be diagnosed, and the medical report which needs to be written for the image to be diagnosed by a doctor and the initialization report determined in the specification have relatively high consistency; a doctor needs to make a small change for the initialized report, and the medical report is written based on the initialized report, so that the writing speed can be greatly improved.

Description

Technical Field

[0001] The present specification relates to the field of computer technology, and in particular to a method, device, storage medium and electronic device for determining an initialization report. Background Art

[0002] In the diagnosis and treatment scenario, doctors need to write medical imaging reports for patients based on their images. To improve writing efficiency, doctors currently usually select a suitable report template from the template library based on the patient's actual symptoms as the initial report, and then modify or supplement the initial report to obtain a personalized medical report for the patient.

[0003] In the existing template library, there are usually report templates for different parts of the human body or different diseases, and the content of these report templates is relatively fixed. Due to the diversity of patient symptoms, when doctors write medical reports based on the report templates in the existing template library, they often need to make many modifications or content supplements, which has limited improvement in writing efficiency.

[0004] Therefore, the present specification provides an initialization report determination method. Summary of the invention

[0005] The present specification provides an initialization report determination method, device, storage medium and electronic device to at least partially solve the above-mentioned problems existing in the prior art.

[0006] This manual adopts the following technical solutions: This specification provides an initialization report determination method, including: Obtain the images to be diagnosed and the candidate reports; According to the image to be diagnosed and the candidate reports, determining the disease label of the image to be diagnosed from the disease labels of the candidate reports as the target disease label; An initialization report is determined from the candidate reports according to the target disease label.

[0007] Optionally, the method for determining each candidate report is as follows: Determine the medical images corresponding to each historical report; Clustering the medical images according to the similarities between the medical images to obtain target clusters; The candidate reports are determined according to the cluster centers of the target clusters.

[0008] Optionally, according to the image to be diagnosed and the candidate reports, determining the disease label of the image to be diagnosed from the disease labels of the candidate reports as the target disease label specifically includes: Determine the medical images corresponding to each candidate report and the disease labels of each candidate report; Determining at least one related image of the image to be diagnosed according to the similarity between the medical image corresponding to each candidate report and the image to be diagnosed; A target disease label is determined according to the disease label corresponding to the at least one related image.

[0009] Optionally, determining a target disease label according to the disease label corresponding to the at least one related image specifically includes: Obtain various disease labels corresponding to each relevant image; For each type of disease label, determine the number of related images corresponding to the disease label as the statistical frequency of the disease label; Determine the number of relevant images as the total number, determine the ratio of the statistical frequency of the disease label of this type to the total number, and obtain the relevant probability of the disease label of this type; According to the relevant probabilities of various disease labels, the target disease label corresponding to the image to be diagnosed is determined.

[0010] Optionally, determining an initialization report from the candidate reports according to the target disease label specifically includes: Taking at least one candidate report corresponding to the target disease label as a similarity report; An initialization report is determined based on the similarity report.

[0011] Optionally, determining an initialization report according to the similarity report specifically includes: respectively determining the similarity between the image features of the image to be diagnosed and the text features of the similar report; According to each similarity, an initialization report is determined in the similarity report.

[0012] Optionally, the candidate report includes diagnostic conclusions of multiple diagnostic items; According to the image to be diagnosed and the candidate reports, determining the disease label of the image to be diagnosed from the disease labels of the candidate reports as the target disease label specifically includes: For each diagnostic item, determine the disease labels corresponding to the diagnostic item according to the disease labels of the candidate reports; Determine the disease label of the image to be diagnosed in the diagnostic item according to each disease label corresponding to the diagnostic item; According to the disease labels of the to-be-diagnosed image in each diagnostic item, a target disease label is determined.

[0013] This specification provides an initialization report determination device, the device comprising: An acquisition module is used to obtain the images to be diagnosed and each candidate report; A prediction module, based on the image to be diagnosed and the candidate reports, determines the disease label of the image to be diagnosed from the disease labels of the candidate reports as a target disease label; The initialization report determination module determines the initialization report from the candidate reports according to the target disease label.

[0014] The present specification provides a computer-readable storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned initialization report determination method is implemented.

[0015] The present specification provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned initialization report determination method when executing the program.

[0016] At least one of the above technical solutions adopted in this specification can achieve the following beneficial effects: In the method for determining the initialization report provided in this specification, the image to be diagnosed and each candidate report are obtained, and the disease label of the image to be diagnosed is determined from the disease labels of each candidate report according to the image to be diagnosed as the target disease label. According to the target disease label, the initialization report is determined from the candidate reports.

[0017] Each candidate report in this method is pre-selected from each historical report included in the historical diagnosis record, so the initialization report in this manual is the historical report selected from the historical diagnosis record. At the same time, because the initialization report is determined based on the disease label, the disease type corresponding to the initialization report in this manual is similar to the disease type reflected by the image to be diagnosed. The medical report that the doctor needs to write for the image to be diagnosed has a high degree of consistency with the initialization report determined in this manual, and the doctor needs to make fewer changes to the initialization report. Writing a medical report based on the initialization report can greatly improve the writing speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of this specification and constitute a part of this specification. The illustrative embodiments and descriptions of this specification are used to explain this specification and do not constitute an improper limitation on this specification. In the drawings: Figure 1 A flowchart of a method for determining an initialization report in this specification; Figure 2 A schematic diagram of an initial cluster provided in an embodiment of this specification; Figure 3A schematic diagram of a target cluster corresponding to an initial cluster provided in an embodiment of this specification; Figure 4 A schematic diagram of a process for determining a related image provided in an embodiment of this specification; Figure 5 A method corresponding to Figure 4 Schematic diagram of the target disease label determination process; Figure 6 A schematic diagram of an initialization report determination principle provided in an embodiment of this specification; Figure 7 A schematic diagram of an initialization report determination device provided in this specification; Figure 8 The corresponding Figure 1 Schematic diagram of electronic equipment. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of this specification more clear, the technical solutions of this specification will be clearly and completely described below in combination with the specific embodiments of this specification and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0020] The technical solutions provided by the embodiments of this specification are described in detail below in conjunction with the accompanying drawings.

[0021] Figure 1 This is a flowchart of a method for determining an initialization report in this specification, which specifically includes the following steps: S100: Obtain the image to be diagnosed and each candidate report.

[0022] All steps in the initialization report determination method provided in this specification can be implemented by any electronic device with computing functions, such as a terminal, a server, etc. For ease of description, the initialization report determination method provided in this specification is described below with the server as the execution subject.

[0023] During each diagnosis process, when the doctor needs to write a medical report for the patient based on the medical images, the server obtains the image to be diagnosed so as to apply the initialization report determination method in this specification to determine the initialization report.

[0024] This specification does not limit the type of image to be diagnosed, and it can be any type of medical image, such as magnetic resonance imaging (MRI), computed tomography (CT), positron emission computed tomography (PET-CT) images, etc.

[0025] The historical diagnosis records contain a large number of historical reports, that is, medical reports written by doctors for historical patients during the historical diagnosis process. At the same time, the historical diagnosis records also include the diagnostic basis for doctors to write each medical report, such as medical images obtained by imaging examinations, test results obtained by laboratory examinations, etc.

[0026] The initialization report determined by the initialization report determination method in this specification is not a traditional disease template or part template with a fixed format. Instead, it is retrieved from the historical reports contained in the historical diagnosis records to retrieve historical reports that are similar to the image information and disease characteristics represented by the image to be diagnosed.

[0027] Therefore, in order to retrieve the initialization report from each historical report, after obtaining the image to be diagnosed, the server needs to obtain the historical report for retrieval, that is, obtain each candidate report. Each candidate report is a part of the historical report selected from each historical report.

[0028] Because the number of historical reports in the historical diagnosis records is relatively large, if the initialization report is retrieved within the scope of all historical reports included in the historical diagnosis records, the retrieval time will be long, and the number of reports of various diseases or parts of the human body in the historical reports may also be unbalanced. Therefore, before step S100, the server will screen out a part of the historical reports from the historical reports included in the historical diagnosis records according to the preset screening method as candidate reports for matching with the image to be diagnosed. The specific screening method of the preset screening method is not limited in this specification, and it can be a clustering-based method or a manual screening.

[0029] The process of determining the candidate reports has been completed before the server obtains the image to be diagnosed. In each diagnostic process, the server only needs to obtain the image to be diagnosed required for the current diagnostic process and the candidate reports that have been determined, and select the initialization report from each candidate report based on the image to be diagnosed. There is no need to repeat the process of determining the candidate reports.

[0030] S102: Determine a target disease label of the image to be diagnosed from disease labels of the candidate reports according to the image to be diagnosed and the candidate reports.

[0031] The core of a medical report is the disease or abnormality seen in a medical image. Usually, the contents of negative reports are similar. How to automatically, accurately and efficiently predict and diagnose a patient's disease or abnormality is the key to initializing a report. Therefore, in the initialization report determination method of this specification, each historical report in the historical diagnosis record has a disease label, and the disease type in the initialization report of the image to be diagnosed is predicted based on the disease label.

[0032] The disease label can be recorded in the Radiology Information System (RIS), or a keyword query, or manually labeled. A disease label in a historical report is used to indicate the type of disease diagnosed in the historical report. This specification does not limit the specific form of the disease label, which can be a disease name, disease number, etc.

[0033] The number of disease labels corresponding to a historical report is related to the actual diagnosis of the historical report, that is, a historical report may correspond to one or more disease labels. Usually, the specific number of disease labels can be determined by medical experts based on the specific imaging scan type and scan site, or by technicians summarizing and determining from the historical report big data.

[0034] Taking the disease name as a disease label as an example, for a historical report obtained through diagnosis based on chest CT, the disease labels that the historical report may have may be composed of one or more of the following disease labels: lung shadows, lung nodules, pneumonia, tuberculosis, pulmonary edema, bullae; tracheal diverticulum, bronchitis; mediastinal mass, mediastinal lymphadenopathy; pericardial effusion, cardiomegaly, aortic hypertension, pulmonary artery enlargement; fractures, bone destruction, bone tumors, etc.

[0035] The server determines the target disease label of the image to be diagnosed from the disease labels of each candidate report according to the image to be diagnosed and each candidate report. The target disease label consists of at least one disease label.

[0036] Specifically, the server first determines the medical images corresponding to each selected candidate report and the disease label of each candidate report in the historical diagnosis record. The medical image of the candidate report is the medical image based on which the doctor wrote the candidate report and is one of the diagnostic bases of the candidate report.

[0037] Secondly, the server determines the related images of the image to be diagnosed according to the similarity between the medical image corresponding to each candidate report and the image to be diagnosed.

[0038] The server may input the medical images corresponding to each candidate report into the first image encoder, respectively, to obtain the image features of the medical images corresponding to each candidate report. The image to be diagnosed is input into the first image encoder, to obtain the image features of the image to be diagnosed. The server determines the feature similarity between the image features of the medical images corresponding to each candidate report and the image features of the image to be diagnosed, respectively, as the similarity between the medical images corresponding to each candidate report and the image to be diagnosed. Among them, this specification does not limit the specific method for determining the feature similarity, which can be cosine similarity, Euclidean distance, etc.

[0039] Then, the server predicts the disease label of the image to be diagnosed based on the disease labels in the corresponding historical reports of the relevant images to determine the target disease label. The target disease label is the type of disease that the image to be diagnosed may represent, which is initially determined in this step.

[0040] Here, the server may determine the medical images corresponding to one or more candidate reports as relevant images according to the similarity between the medical images corresponding to the candidate reports and the image to be diagnosed.

[0041] If the number of related images determined by the server is one, the server may use the disease label corresponding to the one related image as the target disease label.

[0042] If the server determines that there are multiple related images, the server may use the disease labels corresponding to the multiple related images as the target disease label. Alternatively, the server may further screen the disease labels corresponding to the related images to determine the target disease label.

[0043] Because the related images are medical images that are consistent with the image to be diagnosed in visual features based on image similarity, in a medical image, visual features such as brightness, contrast, and texture contain the tissue structure and lesion morphology of the corresponding scanned part of the medical image. Therefore, the disease reflected by the related images of the image to be diagnosed is also similar to the disease reflected by the image to be diagnosed. The server can predict the disease label of the image to be diagnosed based on the disease label of the related images and determine the type of disease that the image to be diagnosed may show.

[0044] The contents of medical reports corresponding to similar disease types will also be similar. Therefore, through the target disease label in this step, it is possible to accurately screen out historical reports with a high degree of similarity to the medical reports that need to be written for the images to be diagnosed from a large number of historical reports contained in the historical diagnosis records.

[0045] S104: Determine an initialization report from the candidate reports according to the target disease label.

[0046] The target disease label is the disease label that the server predicts may correspond to the image to be diagnosed. The candidate report itself has a corresponding disease label. Therefore, the server can use the candidate report corresponding to the target disease label as a similar report among the candidate reports.

[0047] Because the target disease tag consists of at least one disease tag, the candidate report may also correspond to one or more disease tags. The server may determine similar reports in different ways according to the number of disease tags contained in the target disease tag.

[0048] When the target disease tag includes only one disease tag, the server determines, among the candidate reports corresponding to only one disease tag, a candidate report whose corresponding disease tag is the same as the target disease tag as a similar report.

[0049] When the target disease tag contains multiple disease tags, the server determines for each candidate report whether each disease tag contained in the target disease tag is the same as the disease tag corresponding to the candidate report. If so, the candidate report is regarded as a similar report. If not, the candidate report is not regarded as a similar report.

[0050] After the server determines the similarity report, it determines the initialization report according to the similarity report.

[0051] Specifically, the server may use all similar reports as initialization reports, or further screen the similar reports and select one or a small number of similar reports as initialization reports.

[0052] In some scenarios, if the number of similarity reports determined by the server is relatively small, even only one or two, the server may directly use all similarity reports as initialization reports.

[0053] In some scenarios, the number of historical reports contained in the historical diagnostic records is relatively large, which may be hundreds of thousands or millions, or even higher. Therefore, even after preliminary screening, matching the initialization report within the range of candidate reports can still determine a large number of similar reports. All similar reports cannot be directly used as initialization reports, and further screening is required to determine the initialization report. This manual does not limit the method of further screening here, and methods such as image-text feature matching or image-image feature matching can be used.

[0054] The disease label corresponding to the similar report determined in this embodiment is completely consistent with the target disease label, because the disease label corresponding to the similar report is the disease type diagnosed by the similar report in the historical diagnosis process, and the target disease label indicates the disease type that the image to be diagnosed may have as predicted by the server.

[0055] The type of disease exhibited by the patient during the diagnosis process is an important basis for the doctor to write a medical report. In this embodiment, the initialization report is determined from the similar report, and the similar report and the medical report that the doctor needs to write for the patient based on the image to be diagnosed have a high degree of consistency in content. Therefore, the doctor only needs to make a small amount of modification on the basis of the initialization report to obtain a medical report written for the image to be diagnosed, which reduces the doctor's workload in the process of writing the medical report and improves the doctor's writing speed.

[0056] Based on the above Figure 1 In the method for determining the initialization report shown, since each candidate report is pre-selected from each historical report included in the historical diagnosis record, the initialization report in this specification is the historical report selected from the historical diagnosis record. At the same time, because the initialization report is determined based on the target disease label of the image to be diagnosed, that is, for the image to be diagnosed in each diagnostic process, the application of this method can specifically determine the historical report that matches the type of disease that may be manifested by the image to be diagnosed in the diagnostic process, as the initialization report. Therefore, the medical report that the doctor needs to write for the image to be diagnosed has a high degree of consistency with the initialization report determined in this specification, and the doctor needs to make relatively small changes to the initialization report. Writing a medical report based on the initialization report can greatly improve the writing speed.

[0057] The target disease label determined in this specification may include multiple disease labels. Since the disease labels included in the target disease label are the types of diseases that may be manifested in the images to be diagnosed predicted in this specification, the method in this specification can achieve the prediction of multiple disease types at the same time based on the images to be diagnosed.

[0058] Since the lesions of different diseases differ greatly in images, the existing disease detection methods based on the images to be diagnosed can only detect a single type of disease, and cannot detect multiple diseases at the same time. That is, the existing disease detection methods cannot simultaneously predict the multiple different types of diseases that may be manifested in the images to be diagnosed. The existing disease detection methods can only determine the target disease label composed of one disease label, and cannot obtain the target disease label composed of multiple different disease labels in this method.

[0059] Therefore, through the method in this specification, it is possible to detect various types of diseases that a patient may have, determine a more comprehensive and accurate target disease label, and then, based on the target disease label, determine an initialization report that is more suitable for the image to be diagnosed among the candidate reports.

[0060] Before the above step S100, the historical diagnosis record contains historical reports determined in all the historical diagnosis processes. Usually, the number of historical reports contained in the historical diagnosis record is relatively large. Therefore, in order to increase the speed of determining the initialization report, the server preliminarily screens each historical report before obtaining the image to be diagnosed, and uses the screened historical reports as candidate reports.

[0061] This specification does not limit the method of preliminary screening, for example, screening by clustering, or manual screening, etc.

[0062] If candidate reports are determined by manual screening, some historical reports can be manually selected from the historical reports included in the historical diagnosis records according to specific selection conditions as candidate reports. The specific content of the specific selection conditions for manual selection is not limited in this specification and can be set as needed. For example, typical reports that are generally representative of various historical reports, rare reports that represent rare diseases, etc. can be selected as candidate reports.

[0063] If filtering is performed by clustering, the server may perform filtering by clustering once or multiple times according to the filtering granularity and the number of candidate reports that need to be retained.

[0064] In one or more embodiments of the present specification, the server further performs a preliminary screening of each historical report in the historical diagnosis record through a clustering.

[0065] Specifically, the server determines each historical report contained in the historical diagnosis record and the medical image corresponding to each historical report. According to the similarity between each medical image, each medical image is clustered to obtain each target cluster, and each candidate report is determined according to the historical report corresponding to the cluster center of each target cluster.

[0066] The number of candidate reports can be controlled by setting the number of target clusters. In this embodiment, when clustering to determine candidate reports, the number of candidate reports obtained by the server is the same as the number of target clusters that need to be aggregated in the first clustering process.

[0067] Before the above step S100, the server may further perform more fine-grained classification on each historical report in the historical diagnosis record through two clusterings, and screen out more candidate reports, so that the content of each candidate report is more diversified.

[0068] First, the server performs a first clustering of the medical images corresponding to each historical report according to the similarity between the medical images corresponding to each historical report, and obtains each initial clustering cluster.

[0069] Then, for each initial cluster, the server performs a second clustering on each medical image included in the initial cluster to obtain each target cluster corresponding to the initial cluster.

[0070] Finally, the server determines each candidate report according to the cluster center of each target cluster.

[0071] Specifically, the server determines the target clusters corresponding to each initial cluster, and uses the historical reports corresponding to the cluster center of each target cluster as candidate reports.

[0072] Taking CT images as an example, the medical images corresponding to each historical report can be divided into chest CT plain scan, abdominal CT plain scan, pelvic CT plain scan, coronary artery CT angiography, head CT plain scan, urinary system CT plain scan and other part categories according to different scan parts. For each part category, there are diseases that may be detected in this part category, so disease labels can be divided according to different parts, and the disease label of a disease category diagnosed by a scan part can be attributed to the category of the scan part.

[0073] After the first clustering, the medical images corresponding to each historical report may be clustered into initial clusters according to different scanned parts, and each initial cluster corresponds to a CT image of a scanned part. After the second clustering of each initial cluster, the medical images contained in each initial cluster can be divided again according to different disease labels to obtain the target clusters divided by the initial cluster. The medical images contained in each target cluster are medical images with the same disease label of the scanned part.

[0074] Figure 2 A schematic diagram of an initial cluster provided in the embodiments of this specification is shown in FIG. Figure 2 In the figure, different shapes represent medical images of different scanned parts. It can be seen that after one clustering, medical images of the same scanned part are clustered together to form an initial cluster. Figure 2 As shown in the figure, in each initial cluster, the corresponding position of the cluster center is marked with a shadow and Figure 2 The historical report corresponding to the medical image of the cluster center is given in .

[0075] by Figure 2 The initial clustering clusters corresponding to the chest CT scan shown in the figure are clustered again as an example. Figure 3 The second clustering is described. Figure 3 Schematic diagram of target clustering corresponding to an initial clustering cluster provided in the embodiments of this specification, Figure 3The circle marked with dotted shadows represents the cluster center of an initial cluster of the initial cluster during the first clustering, and the four circles marked with mesh shadows represent the cluster centers of the four target clusters obtained after the second clustering. Figure 3 Each target cluster in , each target cluster contains historical reports that are similar in disease type and have similar disease labels.

[0076] The first clustering in this embodiment is coarse clustering, and the second clustering is fine clustering. Figure 3 It can be seen that the number of target clusters determined by secondary clustering is greater than the number of target clusters determined by only one clustering in the above embodiment. Therefore, more candidate reports can be determined in this embodiment. Therefore, when screening candidate reports according to target disease labels, more fine-grained matching can be performed to obtain an initialization report that is more similar to the medical report to be written, further improving the speed of writing medical reports.

[0077] The above-mentioned methods of screening by manual screening and screening by clustering can be performed either alone or simultaneously. When performed simultaneously, in one embodiment, the server can use the candidate report determined by manual screening and the candidate report determined by cluster screening as the final candidate report for determining the initialization report.

[0078] In the above step S102, the server may select relevant images from the medical images corresponding to each historical report according to the similarity threshold or the first quantity threshold.

[0079] That is, the server may determine a preset similarity threshold, and select the medical images corresponding to the candidate reports whose similarity is greater than or equal to the similarity threshold as the relevant images. The server may also determine a preset first quantity threshold, sort the medical images corresponding to the candidate reports according to the similarity, and select the medical images corresponding to the first quantity threshold in order as the relevant images.

[0080] The specific number of related images determined by the server is related to the specific value of the similarity threshold or the first number threshold, so the corresponding similarity threshold or the first number threshold can be set according to the specific number of related images to be determined. For example, if three related images need to be determined, the first number threshold can be set to 3.

[0081] Figure 4 FIG. 1 is a schematic diagram of a process for determining a related image provided in an embodiment of this specification. Figure 4 Image1~Image8 represent the medical images corresponding to 8 different candidate reports, and the content in brackets is the disease label of each candidate report. Figure 4As shown in the figure, I0 represents the image features of the image to be diagnosed, and I1~I8 represent the image features of the medical images corresponding to each candidate report. i The image feature I0 represents the image to be diagnosed, and the image feature I of the medical image corresponding to the i-th historical report i The similarity between .

[0082] Assume that relevant images are selected according to a preset number threshold, and the first number threshold is set to 3. After the server sorts the medical images corresponding to each historical report according to similarity, it selects the first three medical images as relevant images. Figure 4 In the figure, the related images of Top1~Top3 are marked with shadows, which are Image5, Image7 and Image3 respectively.

[0083] Note that Figure 4 In the example of eight historical reports, the process of determining related images is explained at the principle level. In actual application scenarios, the number of historical reports is very large. Among the medical images corresponding to a large number of historical reports, related images with a high degree of similarity to the image to be diagnosed are usually included. In the diagnosis scenario based on medical images, doctors write medical reports based on the image information presented by the medical images. Therefore, through this embodiment, historical reports with a high degree of similarity to the medical reports that doctors need to write for the images to be diagnosed can be determined from the historical reports corresponding to the related images as the initialization reports.

[0084] In the above step S102, when the server determines that there are multiple related images, the server may use the disease label corresponding to each related image as the target disease label. Alternatively, the server may further screen the disease labels corresponding to each related image to determine the target disease label.

[0085] Still Figure 4 For example, when the server uses the disease labels corresponding to each related image as the target disease labels, the target disease labels determined by the server are "pulmonary nodule", "bone tumor" and "mediastinal mass", which are all the disease labels of the medical images corresponding to Image5, Image7 and Image3.

[0086] In another embodiment, the server further screens the disease tags corresponding to the relevant images to determine the target disease tag.

[0087] First, the server determines the number of each related image as the total number, and for each type of disease label in the candidate report corresponding to each related image, determines the number of related images corresponding to the disease label as the statistical frequency of the disease label.

[0088] Then, the server determines the ratio of the statistical frequency of the disease label of this type to the total number, and obtains the relevant probability of the disease label of this type. According to the relevant probability of each disease label, the target disease label is determined among the disease labels of each type.

[0089] Specifically, the server may select at least one disease label as the target disease label from various disease labels according to the relevant probability threshold or the second quantity threshold.

[0090] If the server determines the target disease label according to the relevant probability threshold, the server may determine a preset relevant probability threshold, and use the disease label corresponding to the relevant probability greater than or equal to the relevant probability threshold as the target disease label. If the server selects according to the second quantity threshold, the various disease labels may be sorted according to the relevant probability, and the number of disease labels corresponding to the second quantity threshold may be selected in order as the target disease label.

[0091] The specific value of the relevant probability threshold or the second quantity threshold can be set according to actual selection requirements. The specific number of disease labels included in the target disease label in this specification is related to the value of the relevant probability threshold or the second quantity threshold set when determining the target disease label.

[0092] Because related images are medical images related to the image to be diagnosed based on the similarity between the images, the types of diseases that may be manifested in the image to be diagnosed are similar to the types of diseases corresponding to the disease labels of each related image. The frequency of a disease label appearing in each disease label of the candidate report corresponding to each related image is the number of related images with this type of disease label.

[0093] The greater the statistical frequency of the disease label, the greater the proportion of related images with this type of disease label in each related image, and the greater the probability that the image to be diagnosed also has this type of disease label. Therefore, the correlation probability indicates the possibility that this type of disease label matches the type of disease represented by the image to be diagnosed. For a type of disease label, the more frequent the occurrence of this type of disease label in each disease label of the candidate report corresponding to each related image, the greater the correlation probability of this type of disease label.

[0094] Figure 5 A method corresponding to Figure 4 Schematic diagram of the target disease label determination process. Figure 5 The process of determining target disease labels is described. Figure 5 In the illustrated embodiment, disease labels are screened according to a relevant probability threshold, and the relevant probability threshold is set to 0.5.

[0095] According to Figure 4The server has determined that the related images of Top1 to Top3 are Image5, Image7, and Image3. Based on each related image, the server determines the candidate reported disease labels corresponding to each related image, among which the historically reported disease labels corresponding to Image5 are "pulmonary nodules" and "bone tumors", the candidate reported disease label corresponding to Image3 is "pulmonary nodules", and the candidate reported disease label corresponding to Image7 is "mediastinal mass".

[0096] The server determines that the historical reports corresponding to the relevant images contain three types of disease labels, namely "pulmonary nodules", "mediastinal masses" and "bone tumors". Then, for each type of disease label, the server determines that the correlation probability corresponding to "pulmonary nodules" is 2 / 3, the correlation probability corresponding to "mediastinal masses" is 1 / 3, and the correlation probability corresponding to "bone tumors" is 1 / 3. Among these three correlation probabilities, only the correlation probability corresponding to "pulmonary nodules" is greater than the correlation probability threshold of 0.5, so the server determines that "pulmonary nodules" is the target disease label.

[0097] In the above embodiment in which the disease labels of each relevant image are directly used as the target disease labels without screening, the target disease labels determined are "pulmonary nodules", "bone tumors" and "mediastinal masses". In this embodiment, after screening, the target disease label determined is "pulmonary nodules". By comparison, it can be seen that the method of determining the target disease label in this embodiment can determine the target disease label containing fewer disease labels, that is, the determined target disease label has a higher matching degree with the disease type of the image to be diagnosed.

[0098] In the above step S104, the server can determine the similarity between each similar report and the image to be diagnosed, and further screen according to the similarity to determine the initialization report. The matching degree can be determined by image-text feature matching or image-image feature matching.

[0099] In one or more embodiments of the present specification, the server further screens each similar report by matching image and text features.

[0100] Specifically, the server inputs the image to be diagnosed into the second image encoder to obtain the image features of the image to be diagnosed, and inputs each similar report into the text encoder to obtain the text features of each similar report.

[0101] Then, the server compares the similarity between the image features of the image to be diagnosed and the text features of each similar report, and determines the initialization report among the similar reports.

[0102] This specification does not limit the specific network structure of the second image encoder and the text encoder, but it must be ensured that the second image encoder and the text encoder have undergone comparative learning training to ensure that the features obtained by the two different encoders are mapped to the same feature space and are comparable.

[0103] In one or more embodiments of the present specification, the server may further filter each similar report by using image feature matching.

[0104] Specifically, the server inputs the image to be diagnosed into the third image encoder to obtain the image features of the image to be diagnosed, determines the medical images corresponding to each similar report, and inputs the medical images corresponding to each similar report into the third image encoder to obtain the image features of the medical images corresponding to each similar report.

[0105] Then, the server determines the initialization report in each similar report according to the similarity between the image features of the image to be diagnosed and the image features of the medical images corresponding to each similar report.

[0106] In the above two embodiments, this specification does not limit the specific number of the initialization reports finally determined, and the server selects one or more similar reports as the initialization reports as needed.

[0107] When there is one initialization report, the doctor can directly write a medical report based on the initialization report. When there are multiple initialization reports, the server can display the determined initialization reports to the doctor, so that the doctor can select an initialization report as needed to write a medical report, thereby giving the doctor room for choice and making the initialization report determination method of this specification more flexible.

[0108] Because the initialization report determined in this manual is based on the image to be diagnosed, the disease type that may be represented by the image to be diagnosed is predicted, and then the target disease label is selected from each candidate report based on the predicted disease type. Therefore, the initialization report determined using the method in this manual has a high similarity with the medical report to be written by the doctor, and the doctor only needs to make minor modifications to complete the writing of the current medical report.

[0109] Figure 6 This is a schematic diagram of the initialization report determination principle provided in the embodiments of this specification. Figure 6 In , the candidate reports are represented by rectangles, and the disease label of each candidate report is marked inside the rectangle. It can be seen that some candidate reports have one disease label, and some candidate reports have multiple disease labels. Figure 6The candidate reports are determined by clustering, and the spatial distribution of each candidate report reflects the similarity between the medical images corresponding to each candidate report. During clustering, medical images with at least one same disease label will be clustered in space after clustering. The higher the similarity of disease labels corresponding to different medical images, the closer the spatial distance of these medical images.

[0110] exist Figure 6 In the corresponding embodiment, the target disease label is "pulmonary nodules", and the disease label of similar reports is the target disease label. Figure 6 The three rectangles marked with shadows in the middle represent the three similar reports determined, wherein the similarity between the similar report and the image to be diagnosed is marked below each similar report.

[0111] If the similarity report with the highest similarity is selected as the initialization report, the similarity report with a similarity of 0.7 is used as the initialization report. Figure 6 Among the three similar reports, the rectangle marked with dark shades represents the similar reports determined as initialization reports, and the two rectangles marked with light shades represent the similar reports that are not selected as initialization reports.

[0112] In the above step S104, the server can also perform personalized screening on each candidate report according to the personalized information input by the doctor to obtain each preliminary report. Then, the server determines the initialization report according to the preliminary report with the disease label as the target disease label in each preliminary report.

[0113] The personalized information here can be set according to actual needs. It can be some basic information of the patient, such as the patient's gender, age, etc., or it can be the doctor's department, or it can be the examination site, etc.

[0114] Through this embodiment, a portion of candidate reports that have a low degree of match with the medical report that the doctor needs to write for the image to be diagnosed can be screened out based on the personalized information input by the doctor, so as to obtain a preliminary report. Within the scope of the preliminary report, the disease type is matched according to the predicted target disease label, and the preliminary report that is consistent with the disease type shown by the image to be diagnosed is screened as the final initialization report.

[0115] Because even for the same type of disease, the lesion morphology, size, and onset characteristics of patients of different genders or ages are generally different, so the relevant descriptions of the lesions in the medical report are also different. Through this embodiment, matching is first performed based on personalized information, and historical reports with consistent disease labels and personalized information can be matched, which have a high degree of matching with the medical report that the doctor needs to write based on the image to be diagnosed. The doctor needs to modify less content in the initialization report, and only needs to make a very small amount of modification based on the initialization report to quickly obtain the medical report written during the current diagnosis process.

[0116] For example, in an application scenario of this embodiment, a doctor is reviewing a medical image of a patient. Based on the patient's age of 72 and gender of female, the server selects the candidate report in which the patient is between 70 and 80 years old and the gender is female from each candidate report as the preliminary report based on the personalized information.

[0117] Then, the server determines the initialization report based on the target disease label of the image to be diagnosed, among the obtained preliminary reports, based on the preliminary report whose disease label is consistent with the target disease label. The initialization report determined in this way is a historical report that matches the age, gender, and disease type of the patient in the current diagnosis process.

[0118] In one or more embodiments of the present specification, in order to facilitate the doctor to more quickly find the content that needs to be modified in the initialization report, the server may perform fuzzy processing on the initialization report.

[0119] Specifically, the server identifies information related to the specific condition of the patient's illness in the initialization report determined in the above step S104, such as information indicating the location of the lesion and the shape of the lesion.

[0120] Then, the identified information related to the specific condition of the patient's illness is removed from the initialization report to obtain a fuzzy initialization report.

[0121] This specification does not limit the specific method of the removal operation. For example, the server can replace the information identified as related to the specific condition of the patient's condition with preset characters, or set the position of the information identified as related to the specific condition of the patient's condition to be filled in.

[0122] Through this embodiment, the doctor writes the medical report based on the fuzzified initialization report, so that the doctor can directly fill in the information related to the specific conditions of the patient's symptoms after the removal operation in the fuzzified initialization report to obtain a complete medical report.

[0123] In another embodiment, the server may also perform fuzzification processing on each historical report included in the historical diagnosis record, or perform fuzzification processing on each candidate report obtained after preliminary screening, or perform fuzzification processing on similar reports. The specific method of fuzzification processing is the same as that in the above embodiment.

[0124] Usually, the medical report needs to state the diagnosis conclusions of multiple different diagnostic items in separate paragraphs, and each diagnostic item corresponds to a structure or tissue location to be examined. Taking the medical report corresponding to the chest CT plain scan image as an example, the medical report may include the diagnosis conclusions of diagnostic items such as "lungs", "trachea", "mediastinum", "heart and blood vessels", "thoracic cavity and chest wall", and "bone".

[0125] In one or more embodiments of the present specification, the server may split the medical report into diagnostic conclusions corresponding to different diagnostic items, and determine the project templates of different diagnostic items for different diagnostic items according to the image to be diagnosed, and finally obtain the initialization report by combining multiple project templates, so as to achieve a more fine-grained matching of the initialization report and determine the initialization report that is more suitable for the image to be diagnosed.

[0126] In the above step S102, the server determines, for each diagnostic item, the disease labels corresponding to the diagnostic item from the disease labels of the candidate reports.

[0127] The disease label corresponding to a diagnostic item indicates various disease types that may be diagnosed from the diagnostic item. In this embodiment, the disease labels reported in the history and the diagnostic items that diagnose the disease types corresponding to the disease labels can be manually summarized in advance to obtain the corresponding relationship between each diagnostic item and the disease label.

[0128] For example, the disease labels corresponding to the diagnostic item "lung" may include "lung shadows", "lung nodules", "pneumonia", "tuberculosis", "pulmonary edema", and "bullets". The disease labels corresponding to the diagnostic item "trachea" may include "tracheal dilatation", "tracheal diverticulum", and "bronchitis". The disease labels corresponding to the diagnostic item "mediastinum" may include "mediastinal mass", "mediastinal lymphadenopathy". The disease labels corresponding to the diagnostic item "heart and blood vessels" may include "pericardial effusion", "cardiomegaly", "aortic hypertension", and "pulmonary artery thickening". The disease labels corresponding to the diagnostic item "thoracic cavity and chest wall" may include "pleural effusion", "pneumothorax", and "pleural thickening". The disease labels corresponding to the diagnostic item "bone" may include "fracture", "bone tumor", and "bone destruction".

[0129] Then, the server determines the disease tag of the to-be-diagnosed image in the diagnostic item according to each disease tag corresponding to the diagnostic item.

[0130] According to the above example, if the diagnosis item is "lung", the server will only predict the disease label of the image to be diagnosed in the diagnosis item among the disease labels corresponding to "lung", namely "lung shadow", "lung nodule", "pneumonia", "tuberculosis", "pulmonary edema" and "bulla".

[0131] The server determines the target disease label of the image to be diagnosed according to the disease label of each diagnostic item. The target disease label in this embodiment is used for disease label matching in step S104 to determine the initialization report, because in this embodiment, at the level of diagnostic items, a more fine-grained prediction of the target disease label is performed for each diagnostic item, so that a more accurate target disease label can be determined.

[0132] In the above step S104, the server determines the project template of the diagnostic item according to the candidate report whose project tag of the diagnostic item is the target disease tag. Finally, an initialization report is generated according to the project template of each diagnostic item.

[0133] Specifically, the server uses each candidate report having a disease label in the diagnostic project that is the same as the target disease label of the image to be diagnosed in the diagnostic project as a project similarity report for the diagnostic project.

[0134] The server uses the diagnostic conclusion corresponding to the diagnostic item in the item similarity report as the candidate conclusion of the diagnostic item. Then, the server determines the project template of the diagnostic item in each diagnostic conclusion based on the similarity between the image features of the image to be diagnosed and the text features of the candidate conclusion of the diagnostic item.

[0135] The server combines the project templates of each diagnostic project and generates an initialization report.

[0136] The specific process of determining the disease label of the diagnostic item and matching the item template for the diagnostic item is similar to the process of determining the target disease label and initializing the report within the range of disease labels corresponding to the entire candidate report. Please refer to the corresponding description of steps S102~S104 above.

[0137] The initialization report in this embodiment is obtained by reintegrating the project templates of different diagnostic items. The project templates contained in the initialization report obtained by applying this embodiment may come from multiple candidate reports. At the level of diagnostic items, in each candidate report, the candidate conclusion corresponding to each diagnostic item that matches the image to be diagnosed is selected, so as to obtain an initialization report that is more suitable for the type of disease represented by the image to be diagnosed, reduce the amount of modification when the doctor writes the medical report, and further improve the speed of the doctor writing the medical report.

[0138] The above is the method for determining the initialization report provided in this specification. Based on the same idea, this specification also provides a corresponding device for determining the initialization report, such as Figure 7 shown.

[0139] Figure 7 A schematic diagram of a device for determining an initialization report provided in this specification specifically includes: An acquisition module 200 is used to acquire the image to be diagnosed and each candidate report; A prediction module 202 is used to determine the disease label of the image to be diagnosed from the disease labels of the candidate reports according to the image to be diagnosed and the candidate reports, as a target disease label; The initialization report determination module 204 is used to determine an initialization report from the candidate reports according to the target disease label.

[0140] Optionally, the device further comprises an initial screening module 206; The medical images corresponding to the historical reports are determined, the medical images are clustered according to the similarities between the medical images to obtain target clusters, and the candidate reports are determined according to the cluster centers of the target clusters.

[0141] Optionally, the prediction module 202 is specifically used to determine the medical image corresponding to each candidate report and the disease label of each candidate report, determine at least one related image of the image to be diagnosed based on the similarity between the medical image corresponding to each candidate report and the image to be diagnosed, and determine the target disease label based on the disease label corresponding to the at least one related image.

[0142] Optionally, there are multiple related images, and the prediction module 202 is specifically used to obtain various types of disease labels corresponding to each related image, and for each type of disease label, determine the number of related images corresponding to the disease label as the statistical frequency of the disease label, determine the number of related images as the total number, determine the ratio of the statistical frequency of the disease label to the total number, and obtain the relevant probability of the disease label, and determine the target disease label corresponding to the image to be diagnosed according to the relevant probabilities of each type of disease label.

[0143] Optionally, the initialization report determination module 204 is specifically configured to use at least one candidate report corresponding to the target disease label as a similarity report, and determine an initialization report based on the similarity report.

[0144] Optionally, the initialization report determination module 204 is specifically configured to respectively determine similarities between image features of the image to be diagnosed and text features of the similar reports, and determine an initialization report in the similar reports based on the similarities.

[0145] Optionally, the candidate report contains diagnostic conclusions for multiple diagnostic items, and the prediction module 202 is specifically used to determine, for each diagnostic item, the disease labels corresponding to the diagnostic item according to the disease labels of the candidate reports, determine the disease label of the image to be diagnosed in the diagnostic item according to the disease labels corresponding to the diagnostic item, and determine the target disease label according to the disease labels of the image to be diagnosed in each diagnostic item.

[0146] This specification also provides a computer-readable storage medium, which stores a computer program, which can be used to execute the above Figure 1 The method for determining the initialization report provided.

[0147] This manual also provides Figure 8 The schematic structure diagram of the electronic device shown in FIG. Figure 8 As mentioned above, at the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, and may also include other hardware required for the business. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to achieve the above Figure 1 Of course, in addition to the software implementation, this specification does not exclude other implementations, such as logic devices or a combination of software and hardware, etc., that is, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0148] For the improvement of a technology, it can be clearly distinguished whether it is a hardware improvement (for example, improvement of the circuit structure of diodes, transistors, switches, etc.) or a software improvement (improvement of the method flow). However, with the development of technology, many improvements of the method flow today can be regarded as direct improvements of the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved method flow into the hardware circuit. Therefore, it cannot be said that the improvement of a method flow cannot be implemented with a hardware entity module. For example, a programmable logic device (PLD) (such as a field programmable gate array (FPGA)) is such an integrated circuit whose logical function is determined by the user's programming of the device. Designers can "integrate" a digital system on a PLD by programming themselves, without having to ask chip manufacturers to design and make dedicated integrated circuit chips. Moreover, nowadays, instead of manually making integrated circuit chips, this kind of programming is mostly implemented by "logic compiler" software, which is similar to the software compiler used when developing and writing programs, and the original code before compilation must also be written in a specific programming language, which is called hardware description language (HDL). There is not only one kind of HDL, but many kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also know that it is only necessary to program the method flow slightly in the above-mentioned hardware description languages ​​and program it into the integrated circuit, and then it is easy to obtain the hardware circuit that implements the logic method flow.

[0149] The controller may be implemented in any suitable manner, for example, the controller may take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (e.g., software or firmware) executable by the (micro)processor, a logic gate, a switch, an application-specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller, examples of which include but are not limited to the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320, and the memory controller may also be implemented as part of the control logic of the memory. It is also known to those skilled in the art that, in addition to implementing the controller in a purely computer-readable program code manner, the controller may be implemented in the form of a logic gate, a switch, an application-specific integrated circuit, a programmable logic controller, and an embedded microcontroller by logically programming the method steps. Therefore, such a controller may be considered as a hardware component, and the devices for implementing various functions included therein may also be considered as structures within the hardware component. Or even, the devices for implementing various functions may be considered as both software modules for implementing the method and structures within the hardware component.

[0150] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0151] For the convenience of description, the above device is described by dividing it into various units according to its functions. Of course, when implementing this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0152] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0153] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0154] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0155] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0156] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0157] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0158] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0159] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity 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, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0160] It should be understood by those skilled in the art that the embodiments of this specification may be provided as methods, systems or computer program products. Therefore, this specification may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0161] This specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0162] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system 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.

[0163] The above description is only an embodiment of this specification and is not intended to limit this specification. For those skilled in the art, this specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this specification should be included in the scope of the claims of this application.

Claims

1. A method for determining an initialization report, characterized in that: include: Obtain the images to be diagnosed and the candidate reports; According to the image to be diagnosed and the candidate reports, determining the disease label of the image to be diagnosed from the disease labels of the candidate reports as the target disease label; An initialization report is determined from the candidate reports according to the target disease label.

2. The method according to claim 1, characterized in that The method for determining each candidate report is as follows: Determine the medical images corresponding to each historical report; Clustering the medical images according to the similarities between the medical images to obtain target clusters; The candidate reports are determined according to the cluster centers of the target clusters.

3. The method according to claim 1, characterized in that According to the image to be diagnosed and the candidate reports, determining the disease label of the image to be diagnosed from the disease labels of the candidate reports as the target disease label specifically includes: Determine the medical images corresponding to each candidate report and the disease labels of each candidate report; Determining at least one related image of the image to be diagnosed according to the similarity between the medical image corresponding to each candidate report and the image to be diagnosed; A target disease label is determined according to the disease label corresponding to the at least one related image.

4. The method according to claim 3, characterized in that Determining a target disease label according to the disease label corresponding to the at least one related image specifically includes: Obtain various disease labels corresponding to each relevant image; For each type of disease label, determine the number of related images corresponding to the disease label as the statistical frequency of the disease label; Determine the number of relevant images as the total number, determine the ratio of the statistical frequency of the disease label of this type to the total number, and obtain the relevant probability of the disease label of this type; According to the relevant probabilities of various disease labels, the target disease label corresponding to the image to be diagnosed is determined.

5. The method according to claim 1, characterized in that According to the target disease label, determining an initialization report from the candidate reports specifically includes: Taking at least one candidate report corresponding to the target disease label as a similarity report; An initialization report is determined based on the similarity report.

6. The method according to claim 5, characterized in that According to the similarity report, an initialization report is determined, specifically including: respectively determining the similarity between the image features of the image to be diagnosed and the text features of the similar report; According to each similarity, an initialization report is determined in the similarity report.

7. The method according to claim 1, characterized in that The candidate report contains diagnostic conclusions of multiple diagnostic items; According to the image to be diagnosed and the candidate reports, determining the disease label of the image to be diagnosed from the disease labels of the candidate reports as the target disease label specifically includes: For each diagnostic item, determine the disease labels corresponding to the diagnostic item according to the disease labels of the candidate reports; Determine the disease label of the image to be diagnosed in the diagnostic item according to each disease label corresponding to the diagnostic item; According to the disease labels of the to-be-diagnosed image in each diagnostic item, a target disease label is determined.

8. An initialization report determination device, characterized in that: include: An acquisition module is used to obtain the images to be diagnosed and each candidate report; A prediction module, which determines the disease label of the image to be diagnosed from the disease labels of the candidate reports according to the image to be diagnosed and the candidate reports, as a target disease label; The initialization report determination module determines the initialization report from the candidate reports according to the target disease label.

9. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method described in any one of claims 1 to 7 is implemented.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method described in any one of claims 1 to 7 is implemented.

Images