Efficient image structured report construction method and system

Abstract

The invention provides an efficient image structured report construction method, which comprises the following steps of: according to disease types, acquiring a preset number of text reports of each disease; analyzing related data of image performance and diagnosis impressions in all the text reports, and generating a data table; counting the occurrence frequency of each disease text report data; if the occurrence frequency of the data is higher than a preset threshold value, displaying a control corresponding to the data on a disease type image structured report interface; and if the occurrence frequency of the data is lower than a preset threshold value, hiding the control corresponding to the data. The invention also discloses an efficient image structured report construction system. According to the method and the system, an existing image structured report system can be improved, so that the writing efficiency of the structured report is not lower than that of a traditional text report, and double benefits of improving the report quality and the writing efficiency are realized in the structured report.

Description

Technical Field

[0001] The present invention relates to the field of medical information, and more specifically, to an efficient method and system for constructing structured image reports. Background Art

[0002] Medical imaging structured reports are often constructed based on diagnostic guidelines or the clinical experience of imaging departments. Their contents are composed of single-choice controls, multi-choice controls, navigation maps, diagnostic logic, etc. The dimensions of the evaluation data of imaging structured reports are complete, but it also means that doctors have to spend more time checking these controls one by one to complete the imaging report. Therefore, it is generally believed that the quality of imaging structured reports is higher than that of traditional text reports, but the efficiency is lower than that of traditional text reports. In the imaging department, where the workload is increasingly heavy, if the efficiency problem of structured reports is not solved, it will seriously restrict the promotion and use of imaging structured reports.

[0003] In the traditional process of writing imaging text reports, an audit system is adopted, that is, after junior doctors complete the preliminary report, they submit it to senior doctors for review. Senior doctors will modify the errors in the preliminary report. There are different types of modifications to these errors, such as typos, diagnostic errors, missed diagnoses, supplementary descriptions, etc. The judgment of the type of modification is crucial for quantitatively judging the diagnostic level of a junior doctor and conducting targeted learning for these inadequacies. At present, it can only be judged manually, but it will consume a lot of manpower, so imaging departments generally do not carry out this work. Summary of the invention

[0004] In view of this, the main purpose of the present invention is to provide an efficient method and system for constructing structured image reports. This method can be used to improve the existing structured image report system, improve writing efficiency while ensuring report quality, and automatically determine the report modification type for the reviewing doctor's modifications to the preliminary report, thereby solving the problem of low efficiency in writing structured image reports in the prior art, which is not conducive to the promotion of structured reports.

[0005] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0006] On the one hand, the present invention provides an efficient method for constructing a structured image report, comprising: obtaining a preset number of text reports for each disease according to the type of disease; analyzing relevant data on image manifestations and diagnostic impressions in all the text reports to generate a data table; counting the frequency of occurrence of text report data for each disease; if the frequency of occurrence of the data is higher than a preset threshold, displaying the control corresponding to the data on the structured image report interface of the disease; if the frequency of occurrence of the data is lower than the preset threshold, hiding the control corresponding to the data.

[0007] Preferably, relevant data of imaging manifestations and diagnostic impressions in all the text reports are analyzed to generate a data table, and the steps include: based on the diagnostic impression, dividing the text report into normal report data and abnormal report data; judging the type of disease through the abnormal report data; obtaining the text description corresponding to each disease in the text report; identifying the key features of each disease, including: the location of the lesions, the number of lesions, the size of the lesions, the distribution of the lesions, the morphology of the lesions, the density of the lesions, the edge of the lesions, and secondary changes of the lesions; generating a data table: including the disease name, the location of the lesions, the number of lesions, the distribution of the lesions, the morphology of the lesions, the density of the lesions, the edge of the lesions, and secondary changes of the lesions in the text report.

[0008] Preferably, the method further comprises: associating relevant controls of the image structured report with their corresponding text descriptions; obtaining the checked controls and automatically outputting their text descriptions.

[0009] Preferably, the method further comprises: using a large language model to analyze all data related to imaging manifestations and diagnostic impressions in the text reports, and asking and analyzing questions of the data in batches through an API interface with the large language model.

[0010] Preferably, the method also includes: S1: obtaining the review results of the reviewing doctor on the normal report data, the abnormal report data, and the type of disease; S2: sending the review results to a large language model for training; S3: sending the relevant data of the imaging performance and the diagnostic impression in the text report to the trained large language model to judge the normal report data, the abnormal report data, and the type of disease; S4: executing S1-S3 a preset number of times, and outputting the final normal report data, the abnormal report data, and the type of disease.

[0011] Preferably, the method further comprises: defining an attribute of a control modification type in each image structured report; and automatically determining the modification type of the modified content based on the attribute of the modification type when the control content is modified.

[0012] On the other hand, the present invention also provides an efficient image structured report construction system, which includes: a report acquisition module, a data analysis module, a statistical module and a structured report construction module, wherein the report acquisition module is connected to the data analysis module, and is used to obtain a preset number of text reports for each disease according to the type of disease; the data analysis module is respectively connected to the report acquisition module and the statistical module, and is used to analyze the relevant data of the image performance and diagnostic impression in all the text reports to generate a data table; wherein the data analysis module uses a large language model to analyze the relevant data of the image performance and diagnostic impression in all the text reports, and through the API interface with the large language model, batch questions and analysis of the data; the statistical module is respectively connected to the data analysis module and the structured report construction module, and is used to count the frequency of occurrence of text report data for each disease; the structured report construction module is connected to the statistical module, and is used to display the control corresponding to the data on the image structured report interface of the disease type when the frequency of occurrence of the data is higher than the preset threshold; if the frequency of occurrence of the data is lower than the preset threshold, the control corresponding to the data is hidden.

[0013] Preferably, the data analysis module includes: a first classification unit, a second classification unit, an extraction unit, an identification unit and a table generation unit, wherein the first classification unit is connected to the second classification unit, and is used to divide the text report into normal report data and abnormal report data based on the diagnostic impression; the second classification unit is connected to the first classification unit, the extraction unit and the identification unit, and is used to determine the type of disease through the abnormal report data; the extraction unit is connected to the second classification unit and the table generation unit, and is used to obtain the text description corresponding to each disease in the text report; the identification unit is respectively connected to the second classification unit and the table generation unit, and is used to identify the key features of each disease, including: the location of the lesion, the number of lesions, the size of the lesions, the distribution of the lesions, the morphology of the lesions, the density of the lesions, the edge of the lesions, and the secondary changes of the lesions; the table generation unit is connected to the extraction unit and the identification unit, and is used to generate a data table: including the disease name, the location of the lesions, the number of lesions, the distribution of the lesions, the morphology of the lesions, the density of the lesions, the edge of the lesions, and the secondary changes of the lesions in the text report.

[0014] Preferably, the system also includes an output module, which is connected to the data analysis module and the structured report construction module, and is used to associate the relevant controls of the image structured report with their corresponding text descriptions, obtain the checked controls, and automatically output their text descriptions.

[0015] Preferably, the system also includes a definition module and a judgment module, wherein the definition module is connected to the structured report construction module and the judgment module, and is used to define the attributes of the control modification type in each image structured report; the judgment module is connected to the definition module and the structured report construction module, and is used to automatically judge the modification type of the modified content based on the attributes of the modification type when the control content is modified.

[0016] Technical effects of the present invention:

[0017] 1. The method of the present invention can improve the existing structured imaging report. By analyzing the data of historical text reports, the usage frequency of single disease lesion-related data can be obtained, so that the relevant controls of the structured imaging report corresponding to the text report can be displayed or hidden accordingly, thereby improving the writing efficiency of the structured imaging report. Due to the characteristics of the structured imaging report itself (structured data, wide data dimensions, and clear diagnostic logic), it has important value in improving diagnostic accuracy, optimizing clinical decision-making, and reducing the risk of medical disputes.

[0018] 2. Another method of the present invention improves the existing structured image report by associating the relevant controls of the structured image report with their corresponding text descriptions, obtaining the checked controls, and automatically outputting their text descriptions. The text descriptions corresponding to each disease in the image performance field in the text report can be identified. After the reporting doctor selects, writes, and other operations on the relevant controls of the structured image report, the text descriptions corresponding to these controls are automatically output, which is more in line with natural language descriptions.

[0019] 3. The method of the present invention can also define the attributes of the control modification type in each structured image report; when the control content is modified, the modification type of the modified content is automatically determined based on the attributes of the modification type, so as to realize the judgment of the modification type of the audit report, and realize automatic and objective quantitative evaluation of the diagnostic level of the initial reporting doctor in the single disease report. Subsequent statistics on frequently erroneous reports can be used for targeted learning, which is also of great value for the training of junior doctors. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0021] Figure 1 A flow chart of a method for constructing an efficient structured image report according to Embodiment 1 of the present invention is shown;

[0022] Figure 2A schematic diagram of a data table interface generated by taking head CT as an example in the method for constructing an efficient structured image report according to the first embodiment of the present invention is shown;

[0023] Figure 3 A schematic diagram of an interface for counting the frequency of occurrence of text report data of each disease using head CT as an example in an efficient image structured report construction method according to Embodiment 1 of the present invention is shown;

[0024] Figure 4 A schematic diagram of an image structured report interface constructed by taking head CT as an example in the method for constructing an efficient image structured report according to the first embodiment of the present invention is shown;

[0025] Figure 5 A schematic diagram of the structure of an efficient image structured report construction system according to the second embodiment of the present invention is shown;

[0026] Figure 6 A schematic diagram of the structure of an efficient image structured report construction system according to Embodiment 3 of the present invention is shown;

[0027] Figure 7 A schematic diagram of a data table interface generated by taking head CT as an example in a system for constructing an efficient structured image report according to Embodiment 3 of the present invention is shown;

[0028] Figure 8 A schematic diagram of an interface for counting the frequency of occurrence of text report data of each disease using head CT as an example in an efficient image structured report construction system according to Embodiment 3 of the present invention is shown;

[0029] Fig. 9 A schematic diagram of an image structured report interface constructed by taking head CT as an example in the efficient image structured report construction system according to the third embodiment of the present invention is shown;

[0030] Fig.10 A schematic diagram of the structure of an efficient image structured report construction system according to a fourth embodiment of the present invention is shown;

[0031] Fig.11 A schematic diagram of the structure of an efficient image structured report construction system according to Embodiment 5 of the present invention is shown. DETAILED DESCRIPTION

[0032] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments.

[0033] Embodiment 1

[0034] The present invention improves the existing image structured report system. The method describes how to construct an efficient structured report in the field of medical imaging, so that the writing efficiency of the structured report is not lower than that of the traditional text report, and the dual benefits of improving the report quality and improving the writing efficiency are achieved in the structured report; at the same time, in the field of image structured reports, automatic judgment of the type of modified content of the audit report compared with the preliminary report is achieved, and quantitative evaluation of the report writing quality of junior doctors is achieved.

[0035] Figure 1 FIG. 4 shows a flow chart of a method for constructing an efficient image structured report according to Embodiment 1 of the present invention; Figure 1 As shown, the method comprises the following steps:

[0036] Step S101, obtaining a preset number of text reports for each disease according to the disease type;

[0037] The templates for text reports and structured imaging reports are set according to the type of disease, such as CT, breast MR, etc. After the type of disease is confirmed, previous text reports of the disease are continuously collected. The more text reports are collected, the better the performance of the efficient structured report constructed after analysis will be. Regarding the preset quantity, it can be based on the time period, if it is not less than three months, or the quantity can be set directly, such as 5,000 copies.

[0038] Examples of conditions for text report collection:

[0039] a) Inspection time: past 6 months

[0040] b) Equipment type: CT / MR

[0041] c) Examination items: including keywords such as "head" and "breast", which are determined according to the research field.

[0042] The output of a text report search is:

[0043] a) Check number

[0044] b) Inspection items

[0045] c) Imaging performance

[0046] d) Diagnostic impression

[0047] Step S102, analyzing all the data related to the imaging manifestations and diagnostic impressions in the text reports to generate a data table, the steps of which include:

[0048] Based on the diagnostic impression, dividing the text report into normal report data and abnormal report data;

[0049] The diagnostic impressions in the obtained text reports are used through LLM to determine whether the patient is normal or abnormal, and to classify the patient into diseased and non-disease types. The purpose of this step is to classify the previous text reports into normal report data and abnormal report data.

[0050] Determine the type of disease through abnormal report data;

[0051] The abnormal report data is used through LLM to determine what types of diseases appear in the abnormal reports.

[0052] Get the text description of each disease in the text report;

[0053] Based on the disease type, the text description corresponding to each disease is obtained from the imaging manifestation field of the text report through LLM. The purpose is to determine what text description should be generated after the control in the efficient structured report is edited;

[0054] Identify the key features of each disease, including: location of lesions, number of lesions, size of lesions, distribution of lesions, morphology of lesions, density of lesions, margins of lesions, and secondary changes of lesions;

[0055] Based on the type of disease, the key characteristics of each disease are analyzed through LLM. The purpose is to obtain the dimensions from which the reporting doctor will evaluate the disease when evaluating a report, so as to determine the controls that should be included in an efficient structured report.

[0056] Generate data tables: including the disease name, lesion location, number of lesions, distribution of lesions, lesion morphology, lesion density, lesion margin, and secondary changes of lesions reported in the text.

[0057] Based on the above steps, we can get a data table. Generally, we can use an Excel table: Figure 2 FIG. 1 shows a schematic diagram of a data table interface generated by taking head CT as an example in the method for constructing an efficient structured image report according to the first embodiment of the present invention; Figure 2 As shown, the table contains statistics of parameters such as disease name, location of lesions, number of lesions, distribution of lesions, morphology of lesions, density of lesions, edge of lesions, and secondary changes of lesions;

[0058] The method further includes: associating relevant controls of the image structured report with their corresponding text descriptions; obtaining the checked controls and automatically outputting their text descriptions.

[0059] In addition, there is logic for forming free text between the controls of the structured imaging report, which can be recombined according to language rules based on the attributes of the lesion parameters and converted into a complete and fluent natural language. The natural language can then be sorted and integrated according to the order of the content in the lesion list to generate a text description.

[0060] The present invention associates the relevant controls of the image structured report with their corresponding text descriptions, obtains the checked controls, automatically outputs their text descriptions and the logic between the controls of the image structured report, and the text description formed can be selected from the two and set in the background.

[0061] The method further includes: using a large language model (LLM) to analyze all the data related to the imaging performance and diagnostic impression in the text report, and batch-questioning and analyzing the data through an API interface with the large language model.

[0062] The method further comprises:

[0063] S1: Obtain the review result of the reviewing doctor on the normal report data, the abnormal report data, and the disease type;

[0064] S2: sending the audit result to a large language model for training;

[0065] S3: Sending the relevant data of the imaging performance and the diagnostic impression in the text report to the trained large language model to judge the normal report data, the abnormal report data, and the type of disease;

[0066] S4: Execute S1-S3 for a preset number of times, and output the final normal report data, the abnormal report data, and the disease type.

[0067] Among them, when using a large language model for prediction, you can first use a small amount of data for LLM prediction, manually review, obtain the review results, and feed the review results back to the large language model for personalized training. After several cycles, the accuracy of LLM will also increase accordingly.

[0068] The method further includes: defining the properties of the control modification type in each image structured report; and automatically determining the modification type of the modified content based on the properties of the modification type when the control content is modified.

[0069] The use of LLM to analyze text reports mentioned in the present invention can also be implemented using technical means such as natural language processing (Natural Language Processing, NLP), but the implementation steps and costs are higher than the method of using LLM to analyze text proposed in the present invention.

[0070] Step S103, counting the frequency of occurrence of each disease text report data;

[0071] Figure 3 The figure shows a schematic diagram of an interface for counting the frequency of occurrence of text report data of each disease by taking head CT as an example in the method for constructing an efficient image structured report according to the first embodiment of the present invention; Figure 3 As shown, the frequency of each disease appearing in the abnormal report data, the rate of occurrence of the disease location, the frequency of occurrence of the disease's key attributes, and the text description method of each disease's key attributes are statistically analyzed.

[0072] Step S104, if the frequency of occurrence of the data is higher than a preset threshold, the control corresponding to the data is displayed on the structured report interface of the disease image; if the frequency of occurrence of the data is lower than the preset threshold, the control corresponding to the data is hidden.

[0073] The image structured report interface is simplified by setting the frequency of data occurrence. The construction of efficient image structured reports is based on the data processing and statistical results of text reports. The relevant thresholds are set according to the frequency of occurrence of diseases, parts, key features, etc. according to actual needs. Content above the threshold is displayed in the report interface; content below the threshold is hidden and not displayed in the report interface.

[0074] In head CT examinations, lacunar infarctions mostly occur in the bilateral basal ganglia, left basal ganglia, bilateral radiata corona, right basal ganglia, left radiata corona, basal ganglia, and bilateral frontal lobes. When constructing a structured report, we display these common sites in the report. Other brain sites are not displayed on the report page. They are either filled in manually through the edit box or selected by clicking to expand all sites. By setting thresholds to simplify the page of structured reports, the efficiency of writing structured reports is greatly improved, and it can also ensure that this method meets the writing requirements of most reports of this type. The setting of thresholds is not static, but rather the threshold is raised or lowered as the reporting doctor uses it to present reports of different complexities, so that it can better serve doctors and improve the efficiency of reporting.

[0075] By setting default values, the efficiency of report filling can be further improved. For the key features of the disease, default values ​​can also be set through statistical analysis results to reduce the number of doctor-checked actions and further improve the efficiency of report writing. For example, in the distribution of lacunar infarction in head CT examination, according to the statistical text report data of a medical institution, 98% of the distribution of the disease is multiple and only 2% is single. Therefore, when the doctor selects the lacunar infarction disease template, the option "multiple" for the distribution of the key features of the disease has been selected by default.

[0076] Figure 4FIG. 4 shows a schematic diagram of an image structured report interface constructed by taking head CT as an example in the method for constructing an efficient image structured report according to the first embodiment of the present invention; Figure 4 Shown is the construction of a structured interface for reporting demyelination in head CT images.

[0077] The method further includes: defining the properties of the control modification type in each image structured report; and automatically determining the modification type of the modified content based on the properties of the modification type when the control content is modified.

[0078] Since the characteristic of structured reports is that all contents are structured controls, such as single-select, multiple-select, lesion list, navigation map, etc., the properties of the control modification type can be defined to automatically determine the modification type after the control is modified.

[0079] For example, define the add and delete actions of the lesion list. If the lesion list in the preliminary report is deleted, it means that the lesion content filled in by the junior doctor in the preliminary report is wrong, and the senior review doctor does not agree with it, so it is deleted, which means misdiagnosis;

[0080] If the list of lesions is increased, it means that there are lesions that were not found in the initial report of the junior doctor, that is, missed diagnosis;

[0081] If the structured controls for supplementary performance are supplemented by senior reviewing doctors, it means that there is a supplementary description / supplementary diagnosis, etc.

[0082] After defining a structured report control using this method, the system can automatically determine the type of report modification by determining the modification of the corresponding control content.

[0083] The embodiments of the present invention can obtain the usage frequency of single disease lesion related data through data analysis of historical text reports, so that the related controls of the image structured report corresponding to the text report can be displayed or hidden accordingly, thereby improving the writing efficiency of the image structured report. Due to the characteristics of the image structured report itself (structured data, wide data dimension, clear diagnostic logic), it is of great value in improving diagnostic accuracy, optimizing clinical decision-making and reducing the risk of medical disputes. The embodiments of the present invention can associate the related controls of the image structured report with their corresponding text descriptions, obtain the checked controls, automatically output their text descriptions, and can display the image performance fields in the text report. The text description corresponding to each disease is identified, and after the reporting doctor selects, writes, and other operations on the relevant controls of the image structured report, the text descriptions corresponding to these controls are automatically output, which is more in line with natural language descriptions. The embodiments of the present invention can also define the attributes of the control modification type in each image structured report; when the control content is modified, the modification type of the modified content is automatically determined based on the attributes of the modification type, so as to realize the judgment of the modification type of the audit report, and realize automatic and objective quantitative evaluation of the diagnostic level of the initial reporting doctor in the single disease report. Targeted learning can be carried out for the subsequent statistics of frequently erroneous reports, which is also of great value for the training of junior doctors.

[0084] Embodiment 2

[0085] Figure 5 FIG. 4 shows a schematic diagram of the structure of a system for constructing an efficient structured image report according to Embodiment 2 of the present invention; Figure 5 As shown, the system includes: a report acquisition module 10, a data analysis module 20, a statistics module 30 and a structured report construction module 40, wherein:

[0086] The report acquisition module 10 is connected to the data analysis module 20 and is used to acquire a preset number of text reports for each disease according to the disease type;

[0087] The templates for text reports and structured imaging reports are set according to the type of disease, such as CT, breast MR, etc. After the type of disease is confirmed, previous text reports of the disease are continuously collected. The more text reports are collected, the better the performance of the efficient structured report constructed after analysis will be. Regarding the preset quantity, it can be based on the time period, if it is not less than three months, or the quantity can be set directly, such as 5,000 copies.

[0088] Examples of conditions for text report collection:

[0089] d) Inspection time: past 6 months

[0090] e) Equipment type: CT / MR

[0091] f) Examination items: including keywords such as "head" and "breast", which are determined according to the research field.

[0092] The output of a text report search is:

[0093] e) Check number

[0094] f) Inspection items

[0095] g) Imaging performance

[0096] h) Diagnostic Impression

[0097] The data analysis module 20 is connected to the report acquisition module 10 and the statistical module 30 respectively, and is used to analyze the relevant data of the imaging manifestations and diagnostic impressions in all the text reports and generate data tables;

[0098] The data analysis module 20 uses a large language model to analyze all the data related to the imaging performance and diagnostic impression in the text report, and through the API interface with the large language model, batches of data are asked and analyzed; the steps are:

[0099] S1: Obtain the review result of the reviewing doctor on the normal report data, the abnormal report data, and the disease type;

[0100] S2: sending the audit result to a large language model for training;

[0101] S3: Sending the relevant data of the imaging performance and the diagnostic impression in the text report to the trained large language model to judge the normal report data, the abnormal report data, and the type of disease;

[0102] S4: Execute S1-S3 for a preset number of times, and output the final normal report data, the abnormal report data, and the disease type.

[0103] Among them, when using a large language model for prediction, you can first use a small amount of data for LLM prediction, manually review, obtain the review results, and feed the review results back to the large language model for personalized training. After several cycles, the accuracy of LLM will also increase accordingly.

[0104] The use of LLM to analyze text reports mentioned in this embodiment can also be implemented using technical means such as natural language processing (Natural Language Processing, NLP), but the implementation steps and costs are higher than the method of using LLM to analyze text proposed in the present invention.

[0105] The statistical module 30 is connected to the data analysis module 20 and the structured report construction module 40 respectively, and is used to count the frequency of occurrence of each disease text report data;

[0106] The structured report construction module 40 is connected to the statistical module 30, and is used to display the controls corresponding to the data on the structured report interface of the disease image when the frequency of occurrence of the data is higher than the preset threshold; if the frequency of occurrence of the data is lower than the preset threshold, the controls corresponding to the data are hidden.

[0107] The embodiments of the present invention can obtain the usage frequency of single disease lesion-related data through data analysis of historical text reports, so that the relevant controls of the image structured report corresponding to the text report can be displayed or hidden accordingly, thereby improving the writing efficiency of the image structured report. Due to the characteristics of the image structured report itself (structured data, wide data dimensions, and clear diagnostic logic), it is of great value in improving diagnostic accuracy, optimizing clinical decision-making, and reducing the risk of medical disputes.

[0108] Embodiment 3

[0109] Figure 6 FIG. 4 shows a schematic diagram of the structure of an efficient image structured report construction system according to Embodiment 3 of the present invention; Figure 6 As shown, the data analysis module 20 further includes: a first classification unit 202, a second classification unit 204, an extraction unit 206, a recognition unit 208 and a table generation unit 210, wherein:

[0110] The first classification unit 202 is connected to the second classification unit 204 and is used to classify the text report into normal report data and abnormal report data based on the diagnostic impression;

[0111] The diagnostic impressions in the obtained text reports are used through LLM to determine whether the patient is normal or abnormal, and to classify the patient into diseased and non-disease types. The purpose of this step is to classify the previous text reports into normal report data and abnormal report data.

[0112] The second classification unit 204 is connected to the first classification unit 202, the extraction unit 206, and the identification unit 208, and is used to determine the type of disease through the abnormal report data;

[0113] The abnormal report data is used through LLM to determine what types of diseases appear in the abnormal reports.

[0114] The extraction unit 206 is connected to the second classification unit 204 and the table generation unit 210, and is used to obtain the text description corresponding to each disease in the text report;

[0115] Based on the disease type, the text description corresponding to each disease is obtained from the imaging manifestation field of the text report through LLM. The purpose is to determine what text description should be generated after the control in the efficient structured report is edited;

[0116] The identification unit 208 is connected to the second classification unit 204 and the table generation unit 210, respectively, and is used to identify key features of each disease, including: the location of the lesion, the number of lesions, the size of the lesions, the distribution of the lesions, the morphology of the lesions, the density of the lesions, the edge of the lesions, and the secondary changes of the lesions;

[0117] Based on the type of disease, the key characteristics of each disease are analyzed through LLM. The purpose is to obtain the dimensions from which the reporting doctor will evaluate the disease when evaluating a report, so as to determine the controls that should be included in an efficient structured report.

[0118] The table generating unit 210 is connected to the extracting unit 206 and the identifying unit 208, and is used to generate a data table including the disease name, the location of the lesion, the number of lesions, the distribution of the lesions, the morphology of the lesions, the density of the lesions, the edge of the lesions, and the secondary changes of the lesions reported in the text.

[0119] Based on the above steps, we can get a data table. Generally, we can use an Excel table: Figure 7 FIG. 4 is a schematic diagram showing a data table interface generated by taking head CT as an example in a system for constructing an efficient structured image report according to Embodiment 3 of the present invention; Figure 7 As shown, the table contains statistics of parameters such as disease name, location of lesions, number of lesions, distribution of lesions, morphology of lesions, density of lesions, edge of lesions, and secondary changes of lesions.

[0120] Figure 8 The schematic diagram of the interface for counting the frequency of occurrence of text report data of each disease using head CT as an example in the efficient image structured report construction system according to the third embodiment of the present invention is shown; Figure 8 As shown, the frequency of each disease in the abnormal report data, the rate of disease location, the frequency of disease key attributes, and the text description method of each disease key attribute are counted. By setting the frequency of data occurrence, the image structured report interface is simplified. The construction of efficient image structured reports is based on the data processing and statistical results of text reports. By setting the frequency of occurrence of diseases, locations, key features and other contents, relevant thresholds are set according to actual needs. Content above the threshold is displayed in the report interface; content below the threshold is hidden and not displayed in the report interface.

[0121] In head CT examinations, lacunar infarctions mostly occur in the bilateral basal ganglia, left basal ganglia, bilateral radiata corona, right basal ganglia, left radiata corona, basal ganglia, and bilateral frontal lobes. When constructing a structured report, we display these common sites in the report. Other brain sites are not displayed on the report page. They are either filled in manually through the edit box or selected by clicking to expand all sites. By setting thresholds to simplify the page of structured reports, the efficiency of writing structured reports is greatly improved, and it can also ensure that this method meets the writing requirements of most reports of this type. The setting of thresholds is not static, but rather the threshold is raised or lowered as the reporting doctor uses it to present reports of different complexities, so that it can better serve doctors and improve the efficiency of reporting.

[0122] By setting default values, the efficiency of report filling can be further improved. For the key features of the disease, default values ​​can also be set through statistical analysis results to reduce the number of doctor-checked actions and further improve the efficiency of report writing. For example, in the distribution of lacunar infarction in head CT examination, according to the statistical text report data of a medical institution, 98% of the distribution of the disease is multiple and only 2% is single. Therefore, when the doctor selects the lacunar infarction disease template, the option "multiple" for the distribution of the key features of the disease has been selected by default.

[0123] Fig. 9 FIG. 4 shows a schematic diagram of an image structured report interface constructed by taking head CT as an example in the efficient image structured report construction system according to the third embodiment of the present invention; Fig. 9 Shown is the construction of a structured interface for reporting demyelination in head CT images.

[0124] Embodiment 4

[0125] Fig.10 FIG. 4 shows a schematic diagram of the structure of an efficient image structured report construction system according to Embodiment 4 of the present invention; Fig.10 As shown, the system also includes an output module 50, which is connected to the data analysis module 20 and the structured report construction module 40, and is used to associate the relevant controls of the image structured report with their corresponding text descriptions, obtain the checked controls, and automatically output their text descriptions.

[0126] In addition, there is logic for forming free text between the controls of the structured imaging report, which can be recombined according to language rules based on the attributes of the lesion parameters and converted into a complete and fluent natural language. The natural language can then be sorted and integrated according to the order of the content in the lesion list to generate a text description.

[0127] The present invention associates the relevant controls of the image structured report with their corresponding text descriptions, obtains the checked controls, automatically outputs their text descriptions and the logic between the controls of the image structured report, and the text description formed can be selected from the two and set in the background.

[0128] The embodiments of the present invention can associate the relevant controls of the image structured report with their corresponding text descriptions, obtain the checked controls, and automatically output their text descriptions. The text descriptions corresponding to each disease in the image performance field in the text report can be identified. After the reporting doctor clicks, writes, and other operations on the relevant controls of the image structured report, the text descriptions corresponding to these controls are automatically output, which is more in line with natural language descriptions.

[0129] Embodiment 5

[0130] Fig.11 FIG. 4 shows a schematic diagram of a system for constructing an efficient image structured report according to Embodiment 5 of the present invention. Fig.11 As shown, the system further includes a definition module 60 and a judgment module 70, wherein:

[0131] The definition module 60 is connected to the structured report construction module 40 and the judgment module 70, and is used to define the properties of the control modification type in each image structured report;

[0132] The determination module 70 is connected to the definition module 60 and the structured report construction module 40, and is used to automatically determine the modification type of the modified content based on the attribute of the modification type when the content of the control is modified.

[0133] Since the characteristic of structured reports is that all contents are structured controls, such as single-select, multiple-select, lesion list, navigation map, etc., the properties of the control modification type can be defined to automatically determine the modification type after the control is modified.

[0134] For example, define the add and delete actions of the lesion list. If the lesion list in the preliminary report is deleted, it means that the lesion content filled in by the junior doctor in the preliminary report is wrong, and the senior review doctor does not agree with it, so it is deleted, which means misdiagnosis;

[0135] If the list of lesions is increased, it means that there are lesions that were not found in the initial report of the junior doctor, that is, missed diagnosis;

[0136] If the structured controls for supplementary performance are supplemented by senior reviewing doctors, it means that there is a supplementary description / supplementary diagnosis, etc.

[0137] After defining a structured report control using this method, the system can automatically determine the type of report modification by determining the modification of the corresponding control content.

[0138] The embodiments of the present invention can also define the attributes of the control modification type in each structured image report; when the control content is modified, the modification type of the modified content is automatically determined based on the attributes of the modification type, so as to realize the judgment of the modification type of the audit report, and realize automatic and objective quantitative evaluation of the diagnostic level of the preliminary reporting doctor in the single disease report. The subsequent statistics of frequently erroneous reports can be used for targeted learning, which is also of great value for the training of junior doctors.

[0139] From the above description, it can be seen that the above-mentioned embodiments of the present invention achieve the following technical effects: the embodiments of the present invention can obtain the usage frequency of single disease lesion related data by analyzing the data of historical text reports, so that the relevant controls of the image structured report corresponding to the text report can be displayed and hidden accordingly, thereby improving the writing efficiency of the image structured report. Due to the characteristics of the image structured report itself (structured data, wide data dimension, clear diagnostic logic), it is of great value to improve diagnostic accuracy, optimize clinical decision-making and reduce the risk of medical disputes; the embodiments of the present invention can associate the relevant controls of the image structured report with its corresponding text description, obtain the checked controls, and automatically output its text description. The text description corresponding to each disease in the image performance field in the text report can be identified, and after the reporting doctor clicks, writes, and other operations on the relevant controls of the image structured report, the text descriptions corresponding to these controls are automatically output, which is more in line with natural language descriptions. The embodiment of the present invention can also define the attributes of the control modification type in each image structured report; when the control content is modified, based on the attributes of the modification type, the modification type of the modified content is automatically judged, the judgment of the modification type of the audit report is realized, and the automatic and objective quantitative evaluation of the preliminary reporting doctor's diagnostic level in the single disease report is realized. Targeted learning can be carried out for the subsequent statistics of frequently erroneous reports, which is also of great value for the training of junior doctors.

[0140] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0141] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An efficient method for constructing a structured image report, characterized in that: include: Get a preset number of text reports for each disease by disease type; Analyze the relevant data of imaging manifestations and diagnostic impressions in all the text reports and generate data tables; Count the frequency of each disease text report data; If the frequency of occurrence of the data is higher than the preset threshold, the control corresponding to the data will be displayed on the structured report interface of the disease image; if the frequency of occurrence of the data is lower than the preset threshold, the control corresponding to the data will be hidden.

2. The method for constructing an efficient image structured report according to claim 1, characterized in that: Analyze all the data related to the imaging manifestations and diagnostic impressions in the text reports to generate a data table, the steps of which include: Based on the diagnostic impression, dividing the text report into normal report data and abnormal report data; Determine the type of disease through abnormal report data; Get the text description of each disease in the text report; Identify the key features of each disease, including: location of lesions, number of lesions, size of lesions, distribution of lesions, morphology of lesions, density of lesions, margins of lesions, and secondary changes of lesions; Generate data tables: including the disease name, location of lesions, number of lesions, distribution of lesions, morphology of lesions, density of lesions, margin of lesions, and secondary changes of lesions reported in text.

3. The method for constructing an efficient image structured report according to claim 2, characterized in that: The method further includes: Associating the relevant controls of the image structured report with their corresponding text descriptions; Get the checked control and automatically output its text description.

4. The method for constructing an efficient structured image report according to claim 2, characterized in that: The method also includes: using a large language model to analyze all the relevant data of imaging manifestations and diagnostic impressions in the text reports, and batch questioning and analyzing the data through an API interface with the large language model.

5. The method for constructing an efficient structured image report according to claim 4, characterized in that: The method further includes: S1: Obtain the review result of the reviewing doctor on the normal report data, the abnormal report data, and the disease type; S2: sending the audit result to a large language model for training; S3: Sending the relevant data of the imaging performance and the diagnostic impression in the text report to the trained large language model to judge the normal report data, the abnormal report data, and the type of disease; S4: Execute S1-S3 for a preset number of times, and output the final normal report data, the abnormal report data, and the disease type.

6. The method for constructing an efficient image structured report according to claim 1, characterized in that: The method further includes: Define the properties of the control modification type in each image structured report; When the control content is modified, the modification type of the modified content is automatically determined based on the attribute of the modification type.

7. An efficient image structured report construction system, characterized in that: The system includes: a report acquisition module, a data analysis module, a statistics module and a structured report building module, wherein: The report acquisition module is connected to the data analysis module and is used to acquire a preset number of text reports for each disease according to the disease type; The data analysis module is connected to the report acquisition module and the statistical module respectively, and is used to analyze the relevant data of the imaging performance and the diagnostic impression in all the text reports and generate a data table; wherein the data analysis module uses a large language model to analyze the relevant data of the imaging performance and the diagnostic impression in all the text reports, and through the API interface with the large language model, batches of data are asked questions and analyzed; The statistical module is connected to the data analysis module and the structured report construction module respectively, and is used to count the frequency of occurrence of each disease text report data; The structured report construction module is connected to the statistical module and is used to display the controls corresponding to the data on the structured report interface of the disease image when the frequency of occurrence of the data is higher than the preset threshold; if the frequency of occurrence of the data is lower than the preset threshold, the controls corresponding to the data are hidden.

8. The efficient image structured report construction system according to claim 7, characterized in that: The data analysis module includes: a first classification unit, a second classification unit, an extraction unit, a recognition unit and a table generation unit, wherein: The first classification unit is connected to the second classification unit and is used to classify the text report into normal report data and abnormal report data based on the diagnostic impression; The second classification unit is connected to the first classification unit, the extraction unit, and the identification unit, and is used to determine the type of disease through the abnormal report data; The extraction unit is connected to the second classification unit and the table generation unit, and is used to obtain the text description corresponding to each disease in the text report; The identification unit is connected to the second classification unit and the table generation unit, respectively, and is used to identify key features of each disease, including: the location of lesions, the number of lesions, the size of lesions, the distribution of lesions, the morphology of lesions, the density of lesions, the edge of lesions, and secondary changes of lesions; The table generation unit is connected to the extraction unit and the recognition unit, and is used to generate a data table: including the disease name, lesion location, number of lesions, distribution of lesions, morphology of lesions, lesion density, lesion edge, and secondary changes of lesions reported in the text.

9. The efficient image structured report construction system according to claim 7, characterized in that: The system also includes an output module, which is connected to the data analysis module and the structured report construction module, and is used to associate the relevant controls of the image structured report with their corresponding text descriptions, obtain the checked controls, and automatically output their text descriptions.

10. The efficient image structured report construction system according to claim 7, characterized in that: The system also includes a definition module and a judgment module, wherein: The definition module is connected to the structured report construction module and the judgment module, and is used to define the properties of the control modification type in each image structured report; The judgment module is connected to the definition module and the structured report construction module, and is used to automatically judge the modification type of the modified content based on the attribute of the modification type when the content of the control is modified.

Images