Medical image multi-stage analysis system

Abstract

The invention relates to medical image intelligent analysis, in particular to a medical image multi-stage analysis system which is composed of an image diagnosis analysis module, a multi-stage comparison early warning module, a multi-stage image space alignment module, a multi-stage image correlation diagnosis module, a multi-stage comparison analysis module, a multi-stage comparison typical medical record knowledge base and an image clinical decision knowledge base. The system has the capabilities of multi-stage comparison reminding, interpretable multi-stage comparison report output, multi-stage comparison-based preliminary report updating, multi-stage comparison-based clinical decision support and the like, has the advantages of high interpretability, interactivity, expandability and the like, can realize accurate multi-stage comparison analysis and suggestion, and realizes high-order intelligent auxiliary diagnosis; according to the technical scheme provided by the invention, the defect in the prior art that multi-stage comparative analysis of interpretability, interactivity and expansibility is difficult to carry out on multi-stage images can be effectively overcome.

Description

Technical Field

[0001] The present invention relates to intelligent analysis of medical images, and in particular to a medical image multi-phase analysis system. Background Art

[0002] Medical imaging multi-phase analysis refers to the process of performing multiple imaging examinations on the same part at different time points and conducting a comprehensive analysis of the imaging data at these different time points. This method is particularly suitable for observing the dynamic changes of lesions, such as the growth rate of tumors, the effect response of treatment, etc. By comparing imaging data at different time points, the nature of lesions, development trends and treatment effects can be more accurately evaluated.

[0003] In modern medicine, the importance of multi-phase analysis of medical images is becoming increasingly prominent. It not only helps doctors detect abnormalities in the early stages of the disease and improves the accuracy of judgment, but also plays an irreplaceable role in the formulation of treatment plans, monitoring of treatment effects and tracking of disease progression. Therefore, the development of a system with multi-phase intelligent analysis capabilities for medical images has extremely important clinical application value.

[0004] In the prior art, intelligent analysis systems involving multiple images or multiple periods of images often extract high-level semantic features from multiple periods of images by designing a computer vision module with pre- or post-fusion features to output the results of multi-period analysis of medical images. This type of implicit modeling of the distribution between input images and output results does not have generalization capabilities on the one hand, and does not have the ability to perform interpretable and interactive multi-period comparative analysis on the other. Summary of the invention

[0005] In view of the above-mentioned shortcomings of the prior art, the present invention provides a medical image multi-phase analysis system, which can effectively overcome the defects of the prior art that it is difficult to perform interpretable, interactive and scalable multi-phase comparative analysis of multi-phase images.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A medical image multi-phase analysis system, comprising an image diagnosis and analysis module, a multi-phase contrast warning module, a multi-phase image space alignment module, a multi-phase image correlation diagnosis module, a multi-phase contrast analysis module, a multi-phase contrast typical medical record knowledge base and an image clinical decision knowledge base; The image diagnosis and analysis module generates a preliminary report based on the current medical image, and gives the specific coordinates of each lesion area in the current medical image in the preliminary report, and inputs the current medical image and the preliminary report together into the multi-period comparison warning module and the multi-period comparison analysis module; The multi-phase contrast warning module generates a multi-phase contrast reminder for reference by the clinician based on the current medical image and preliminary report output by the image diagnosis and analysis module, or the existing current medical image and preliminary report written by the corresponding clinician, and inputs the multi-phase contrast reminder into the multi-phase image space alignment module and the multi-phase contrast analysis module; The multi-phase image spatial alignment module maps each lesion area in the current medical image to the original medical image one by one according to the multi-phase comparison reminder and the preliminary report coordinate analysis using the spatial transformation matrix between the current medical image and the original medical image, generates a multi-phase image positioning coordinate transformation, completes the spatial alignment between the current medical image and the original medical image, realizes the longitudinal comparison of the same lesion area, and inputs the multi-phase image positioning coordinate transformation, the current medical image and the original medical image together into the multi-phase image association diagnosis module; The multi-phase image correlation diagnosis module performs longitudinal comparison of the same lesion area between the current medical image and the original medical image according to the multi-phase image positioning coordinate transformation, generates a multi-phase image comparison including the image findings of each lesion area in the current medical image and the original medical image, and inputs the multi-phase image comparison to the multi-phase comparison analysis module and the multi-phase comparison analysis module; The multi-period comparison analysis module integrates clinical medical records, current medical images, original medical images, a knowledge base of typical multi-period comparison cases, a knowledge base of clinical image decisions, and intermediate outputs of various reasoning links including preliminary reports, reminders of multi-period comparisons, and multi-period image comparisons to generate an explainable multi-period comparison report or a preliminary report update based on multi-period comparisons, and also generates clinical decision support based on multi-period comparisons that includes corresponding clinical decision recommendations; Wherein, when the input to the multi-phase comparison warning module is the current medical image and preliminary report output by the image diagnosis analysis module, the multi-phase comparison analysis module generates an interpretable multi-phase comparison report; When the multi-phase comparison warning module is input with the existing current medical images and the preliminary report written by the corresponding clinician, the multi-phase comparison analysis module generates a preliminary report update based on the multi-phase comparison.

[0007] Preferably, the multi-phase image spatial alignment module maps each lesion area in the current medical image to the original medical image one by one using the spatial transformation matrix between the current medical image and the original medical image according to the multi-phase comparison reminder and the preliminary report coordinate analysis, and generates a multi-phase image positioning coordinate transformation, including: Estimate the spatial transformation matrix from the current medical image to each original medical image ( f 1 , f 2 , …, f n ),in, f nFrom current medical image to original medical image n The space transformation matrix, n is the number of original medical images; According to the specific coordinates of each lesion area in the multi-phase comparison reminder and the preliminary report coordinate analysis, the following formula is used for spatial transformation to map each lesion area in the current medical image to the original medical image one by one, and generate the multi-phase image positioning coordinate transformation: ; in, x , y , ... represent the specific coordinates of a lesion area in the current medical image, f i From current medical image to original medical image i The space transformation matrix, , x i , y i , ... represent the original medical images i The specific coordinates of the same lesion area in the The current medical image is the current film, the original medical image i Front piece i , .

[0008] Preferably, the multi-phase image correlation diagnosis module performs a longitudinal comparison of the same lesion area of ​​the current medical image and the original medical image according to the multi-phase image positioning coordinate transformation, and generates a multi-phase image comparison including the image findings of each lesion area in the current medical image and the original medical image, including: According to the specific coordinates of the same lesion area in the current medical image and the original medical image, the corresponding attention map is constructed to extract the visual representation of the lesion area, which is then sent to the language decoder of the multimodal large model. Wherein, in the above mode, the input of the language decoder of the multimodal large model consists of [current film visual features, current film lesion area visual features, previous film 1 visual features, previous film 1 lesion area visual features, …, previous film n visual features, previous film n lesion area visual features, user instructions]; The user instruction is: "Please generate a multi-phase image comparison containing the image findings of each lesion area in the current medical image and the original medical image based on the input multi-phase image positioning coordinate transformation, the current medical image and the original medical image."

[0009] Preferably, the multi-period comparison analysis module integrates clinical data of medical records, current medical images, original medical images, multi-period comparison typical medical record knowledge base, image clinical decision knowledge base, and intermediate outputs of various reasoning links including preliminary reports, multi-period comparison reminders and multi-period image comparisons to generate an explainable multi-period comparison report or a preliminary report update based on multi-period comparison, and at the same time generates a clinical decision support based on multi-period comparison containing corresponding clinical decision suggestions, including: Utilize the retrieval enhancement technology RAG, that is, according to the visual features of the current case and the integration of the intermediate outputs of each reasoning link, extract the global semantic features of the current case through the feature integrator based on the "self-attention + mutual attention" mechanism, and retrieve the multi-phase comparison report and multi-phase comparison clinical decision recommendation results with high cosine similarity from the multi-phase comparison typical medical record knowledge base and the image clinical decision knowledge base, splice them behind the integration of the intermediate outputs of the above-mentioned reasoning links to form a unified "data link", and then input them into the multi-phase comparison analysis module to generate an explainable multi-phase comparison report or a preliminary report update based on multi-phase comparison, and at the same time generate a multi-phase comparison-based clinical decision support containing corresponding clinical decision recommendations; Among them, the intermediate output integration of each reasoning link specifically includes [current film visual features, current film lesion area visual features, previous film 1 visual features, previous film 1 lesion area visual features, …, previous film n visual features, previous film n lesion area visual features, medical record clinical data text features, preliminary report text features, multi-phase comparison reminder text features, multi-phase image comparison text features]; The update of the preliminary report based on the multi-period comparison is the revised interpretability multi-period comparison report.

[0010] Preferably, the interpretable multi-period comparison report and the preliminary report update based on the multi-period comparison include, on the one hand, an accurate multi-period comparison longitudinal diagnosis conclusion, and on the other hand, a detailed multi-period comparison longitudinal diagnosis basis; The explainable multi-phase comparison report and the preliminary report update based on the multi-phase comparison include the specific coordinates of each lesion area, support the display of the corresponding multi-phase comparison longitudinal diagnosis conclusion and the multi-phase comparison longitudinal diagnosis basis, thereby forming a complete explainable link.

[0011] Preferably, the image diagnosis and analysis module, the multi-phase contrast and warning module, the multi-phase image association diagnosis module and the multi-phase contrast and analysis module are all constructed based on a multimodal large model.

[0012] Compared with the prior art, the medical image multi-phase analysis system provided by the present invention is designed to be composed of an image diagnosis and analysis module, a multi-phase comparison warning module, a multi-phase image spatial alignment module, a multi-phase image correlation diagnosis module, a multi-phase comparison analysis module, a multi-phase comparison typical medical record knowledge base and an image clinical decision knowledge base. The system has the capabilities of multi-phase comparison reminders, explainable multi-phase comparison report output, preliminary report updates based on multi-phase comparisons and clinical decision support based on multi-phase comparisons, and has the advantages of high explainability, interactivity and scalability, and can achieve accurate multi-phase comparison analysis and suggestions, and realize high-order intelligent auxiliary diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0014] Figure 1 It is a system schematic diagram of the present invention; Figure 2 A schematic diagram of a process for generating a multi-period comparison reminder in the present invention; Figure 3 A schematic diagram of a process for generating an interpretable multi-period comparison report in the present invention; Figure 4 For the present invention Figure 2 and Figure 3 Schematic diagram of the mid-term multi-period comparison reminder; Figure 5 For the present invention Figure 3 Schematic diagram of the coordinate resolution of the preliminary report; Figure 6 For the present invention Figure 3 and Fig.10 Schematic diagram of coordinate transformation for multi-phase image positioning; Figure 7 For the present invention Figure 3 and Fig.10 Schematic diagram of multi-phase image comparison; Figure 8 For the present invention Figure 3 and Fig.10 Schematic diagram of the integration of intermediate outputs of each reasoning link in ; Fig. 9 For the present invention Figure 3 Schematic diagram of the multi-period comparative report on interpretability; Fig.10A schematic diagram of a process for updating the preliminary report based on multi-period comparison in the present invention, i.e., generating a revised interpretable multi-period comparison report; Fig.11 For the present invention Fig.10 Schematic diagram of coordinate analysis for comparison reminders in multiple periods; Fig.12 For the present invention Fig.10 Schematic diagram of the multi-period comparative report of the corrected interpretability. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0016] A medical image multi-phase analysis system, such as Figure 1 As shown, it includes an image diagnosis and analysis module, a multi-phase contrast warning module, a multi-phase image space alignment module, a multi-phase image association diagnosis module, a multi-phase contrast analysis module, a multi-phase contrast typical medical record knowledge base and an image clinical decision knowledge base; The image diagnosis and analysis module generates a preliminary report based on the current medical image, and gives the specific coordinates of each lesion area in the current medical image in the preliminary report, and inputs the current medical image and the preliminary report together into the multi-period comparison warning module and the multi-period comparison analysis module; The multi-phase comparison warning module generates a multi-phase comparison reminder (such as Figure 2 and Figure 4 As shown), and input the multi-phase comparison reminder into the multi-phase image space alignment module and the multi-phase comparison analysis module; The multi-phase image spatial alignment module is based on multi-phase comparison reminders and preliminary report coordinate analysis (such as Figure 5 As shown in the figure), the spatial transformation matrix between the current medical image and the original medical image is used to map each lesion area in the current medical image to the original medical image one by one, and a multi-phase image positioning coordinate transformation is generated (as shown in the figure). Figure 6 As shown in the figure), the spatial alignment between the current medical image and the original medical image is completed, the longitudinal comparison of the same lesion area is realized, and the multi-phase image positioning coordinate transformation, the current medical image and the original medical image are input into the multi-phase image correlation diagnosis module together; The multi-phase image correlation diagnosis module performs a longitudinal comparison of the same lesion area between the current medical image and the original medical image based on the multi-phase image positioning coordinate transformation, and generates a multi-phase image comparison containing the image findings of each lesion area in the current medical image and the original medical image (such as Figure 7 As shown), and input the multi-phase image comparison into the multi-phase comparison analysis module and the multi-phase comparison analysis module; The multi-period comparison analysis module integrates clinical data of medical records, current medical images, original medical images, multi-period comparison typical medical record knowledge base, image clinical decision knowledge base, and intermediate output integration of each reasoning link including preliminary report, multi-period comparison reminder and multi-period image comparison (such as Figure 8 ), generate interpretable multi-period comparison reports (such as Figure 3 and Fig. 9 ) or a preliminary report update based on a multi-period comparison (e.g. Fig.10 and Fig.12 As shown in Figure 2), a multi-phase comparison-based clinical decision support is generated, which includes corresponding clinical decision suggestions; Wherein, when the input to the multi-phase comparison warning module is the current medical image and preliminary report output by the image diagnosis analysis module, the multi-phase comparison analysis module generates an interpretable multi-phase comparison report; When the multi-phase comparison warning module is input with the existing current medical images and the preliminary report written by the corresponding clinician, the multi-phase comparison analysis module generates a preliminary report update based on the multi-phase comparison.

[0017] In the technical solution of the present application, the image diagnosis and analysis module, the multi-period contrast warning module, the multi-period image correlation diagnosis module and the multi-period contrast analysis module are all constructed based on a multimodal large model.

[0018] ① The multi-phase image spatial alignment module uses the spatial transformation matrix between the current medical image and the original medical image to map each lesion area in the current medical image to the original medical image one by one according to the multi-phase comparison reminder and preliminary report coordinate analysis, and generates a multi-phase image positioning coordinate transformation, including: Estimate the spatial transformation matrix from the current medical image to each original medical image ( f 1 , f 2 , …, f n ),in, f n From current medical image to original medical image n The space transformation matrix, n is the number of original medical images; According to the specific coordinates of each lesion area in the multi-phase comparison reminder and the preliminary report coordinate analysis, the following formula is used for spatial transformation to map each lesion area in the current medical image to the original medical image one by one, and generate the multi-phase image positioning coordinate transformation: ; in, x , y , ... represent the specific coordinates of a lesion area in the current medical image, f i From current medical image to original medical image i The space transformation matrix, , x i , y i , ... represent the original medical images i The specific coordinates of the same lesion area in the The current medical image is the current film, the original medical image i Front piece i , .

[0019] ② The multi-phase image correlation diagnosis module performs a longitudinal comparison of the same lesion area between the current medical image and the original medical image according to the multi-phase image positioning coordinate transformation, and generates a multi-phase image comparison containing the image findings of each lesion area in the current medical image and the original medical image, including: According to the specific coordinates of the same lesion area in the current medical image and the original medical image, the corresponding attention map is constructed to extract the visual representation of the lesion area, which is then sent to the language decoder of the multimodal large model. Wherein, in the above mode, the input of the language decoder of the multimodal large model consists of [current film visual features, current film lesion area visual features, previous film 1 visual features, previous film 1 lesion area visual features, …, previous film n visual features, previous film n lesion area visual features, user instructions]; The user instruction is: "Please generate a multi-phase image comparison containing the image findings of each lesion area in the current medical image and the original medical image based on the input multi-phase image positioning coordinate transformation, the current medical image and the original medical image."

[0020] ③ The multi-period comparison analysis module integrates clinical data of medical records, current medical images, original medical images, multi-period comparison typical medical record knowledge base, image clinical decision knowledge base, and intermediate outputs of each reasoning link including preliminary report, multi-period comparison reminder and multi-period image comparison to generate an explainable multi-period comparison report or a preliminary report update based on multi-period comparison, and generates clinical decision support based on multi-period comparison containing corresponding clinical decision recommendations, including: Utilize the retrieval enhancement technology RAG, that is, according to the visual features of the current case and the integration of the intermediate outputs of each reasoning link, extract the global semantic features of the current case through the feature integrator based on the "self-attention + mutual attention" mechanism, and retrieve the multi-phase comparison report and multi-phase comparison clinical decision recommendation results with high cosine similarity from the multi-phase comparison typical medical record knowledge base and the image clinical decision knowledge base, splice them behind the integration of the intermediate outputs of the above-mentioned reasoning links to form a unified "data link", and then input them into the multi-phase comparison analysis module to generate an explainable multi-phase comparison report or a preliminary report update based on multi-phase comparison, and at the same time generate a multi-phase comparison-based clinical decision support containing corresponding clinical decision recommendations; Among them, the intermediate output integration of each reasoning link specifically includes [current film visual features, current film lesion area visual features, previous film 1 visual features, previous film 1 lesion area visual features, …, previous film n visual features, previous film n lesion area visual features, medical record clinical data text features, preliminary report text features, multi-phase comparison reminder text features, multi-phase image comparison text features]; The update of the preliminary report based on the multi-period comparison is the revised interpretability multi-period comparison report.

[0021] In the technical solution of this application, the interpretable multi-period comparison report and the preliminary report update based on the multi-period comparison include accurate multi-period comparison longitudinal diagnostic conclusions (such as Fig. 9 and Fig.12 "1. Nodules in the left lower lobe of the lung [x1, y1, x2, y2]. Compared with chest CT scans before August 25, 2022, the possibility of inflammation is high and tumors cannot be ruled out. Please combine clinical data"; on the other hand, it contains detailed multi-phase comparison longitudinal diagnostic evidence (such as Fig. 9 and Fig.12 "1. Before August 25, 2022, multiple films showed no obvious abnormalities in the left lower lobe of the lung. The nodule shadows shown in the current film, combined with the lack of a history of malignant tumors, are likely to be caused by inflammation"); The explainable multi-phase comparison report and the preliminary report update based on the multi-phase comparison contain the specific coordinates of each lesion area, and support the display of the corresponding multi-phase comparison longitudinal diagnosis conclusion and multi-phase comparison longitudinal diagnosis basis, thus forming a complete explainable link.

[0022] In the technical solution of this application, the preliminary report update based on the multi-period comparison is the revised interpretable multi-period comparison report. Figure 2 The preliminary report written by the clinician stated: "Malignancy is considered to be highly likely"; Fig.12 The revised interpretable multi-phase comparison report shown in the figure points out: "Inflammation is considered to be highly likely", that is, through the longitudinal comparison of multiple phases of images, an effective revision of the preliminary report written by the clinician is achieved.

[0023] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A medical image multi-phase analysis system, characterized in that: It includes image diagnosis and analysis module, multi-phase contrast warning module, multi-phase image space alignment module, multi-phase image correlation diagnosis module, multi-phase contrast analysis module, multi-phase contrast typical medical record knowledge base and image clinical decision knowledge base; The image diagnosis and analysis module generates a preliminary report based on the current medical image, and gives the specific coordinates of each lesion area in the current medical image in the preliminary report, and inputs the current medical image and the preliminary report together into the multi-period comparison warning module and the multi-period comparison analysis module; The multi-phase contrast warning module generates a multi-phase contrast reminder for reference by the clinician based on the current medical image and preliminary report output by the image diagnosis and analysis module, or the existing current medical image and preliminary report written by the corresponding clinician, and inputs the multi-phase contrast reminder into the multi-phase image space alignment module and the multi-phase contrast analysis module; The multi-phase image spatial alignment module maps each lesion area in the current medical image to the original medical image one by one according to the multi-phase comparison reminder and the preliminary report coordinate analysis using the spatial transformation matrix between the current medical image and the original medical image, generates a multi-phase image positioning coordinate transformation, completes the spatial alignment between the current medical image and the original medical image, realizes the longitudinal comparison of the same lesion area, and inputs the multi-phase image positioning coordinate transformation, the current medical image and the original medical image together into the multi-phase image association diagnosis module; The multi-phase image correlation diagnosis module performs longitudinal comparison of the same lesion area between the current medical image and the original medical image according to the multi-phase image positioning coordinate transformation, generates a multi-phase image comparison including the image findings of each lesion area in the current medical image and the original medical image, and inputs the multi-phase image comparison to the multi-phase comparison analysis module and the multi-phase comparison analysis module; The multi-period comparison analysis module integrates clinical medical records, current medical images, original medical images, a knowledge base of typical multi-period comparison cases, a knowledge base of clinical image decisions, and intermediate outputs of various reasoning links including preliminary reports, reminders of multi-period comparisons, and multi-period image comparisons to generate an explainable multi-period comparison report or a preliminary report update based on multi-period comparisons, and also generates clinical decision support based on multi-period comparisons that includes corresponding clinical decision recommendations; Wherein, when the input to the multi-phase comparison warning module is the current medical image and preliminary report output by the image diagnosis analysis module, the multi-phase comparison analysis module generates an interpretable multi-phase comparison report; When the multi-phase comparison warning module is input with the existing current medical images and the preliminary report written by the corresponding clinician, the multi-phase comparison analysis module generates a preliminary report update based on the multi-phase comparison.

2. The medical image multi-phase analysis system according to claim 1, characterized in that: The multi-phase image spatial alignment module maps each lesion area in the current medical image to the original medical image one by one based on the multi-phase comparison reminder and the preliminary report coordinate analysis using the spatial transformation matrix between the current medical image and the original medical image, and generates a multi-phase image positioning coordinate transformation, including: Estimate the spatial transformation matrix from the current medical image to each original medical image ( f 1. f 2. … f n ),in, f n From current medical image to original medical image n The space transformation matrix, n is the number of original medical images; According to the specific coordinates of each lesion area in the multi-phase comparison reminder and the preliminary report coordinate analysis, the following formula is used for spatial transformation to map each lesion area in the current medical image to the original medical image one by one, and generate the multi-phase image positioning coordinate transformation: ； in, x , y , ... represent the specific coordinates of a lesion area in the current medical image, f i From current medical image to original medical image i The space transformation matrix, , x i , y i , ... represent the original medical images i The specific coordinates of the same lesion area in the The current medical image is the current film, the original medical image i Front piece i , .

3. The medical image multi-phase analysis system according to claim 2, characterized in that: The multi-phase image correlation diagnosis module performs a longitudinal comparison of the same lesion area of ​​the current medical image and the original medical image according to the multi-phase image positioning coordinate transformation, and generates a multi-phase image comparison containing the image findings of each lesion area in the current medical image and the original medical image, including: According to the specific coordinates of the same lesion area in the current medical image and the original medical image, the corresponding attention map is constructed to extract the visual representation of the lesion area, which is then sent to the language decoder of the multimodal large model. Wherein, in the above mode, the input of the language decoder of the multimodal large model consists of [current film visual features, current film lesion area visual features, previous film 1 visual features, previous film 1 lesion area visual features, …, previous film n visual features, previous film n lesion area visual features, user instructions]; The user instruction is: "Please generate a multi-phase image comparison containing the image findings of each lesion area in the current medical image and the original medical image based on the input multi-phase image positioning coordinate transformation, the current medical image and the original medical image." 4. The medical image multi-phase analysis system according to claim 3, characterized in that: The multi-period comparison analysis module integrates clinical data of medical records, current medical images, original medical images, multi-period comparison typical medical record knowledge base, image clinical decision knowledge base, and intermediate outputs of various reasoning links including preliminary reports, multi-period comparison reminders and multi-period image comparisons to generate an explainable multi-period comparison report or a preliminary report update based on multi-period comparison, and at the same time generates clinical decision support based on multi-period comparison containing corresponding clinical decision suggestions, including: Utilize the retrieval enhancement technology RAG, that is, according to the visual features of the current case and the integration of the intermediate outputs of each reasoning link, extract the global semantic features of the current case through the feature integrator based on the "self-attention + mutual attention" mechanism, and retrieve the multi-phase comparison report and multi-phase comparison clinical decision recommendation results with high cosine similarity from the multi-phase comparison typical medical record knowledge base and the image clinical decision knowledge base, splice them behind the integration of the intermediate outputs of the above-mentioned reasoning links to form a unified "data link", and then input them into the multi-phase comparison analysis module to generate an interpretable multi-phase comparison report or a preliminary report update based on multi-phase comparison, and at the same time generate a multi-phase comparison-based clinical decision support containing corresponding clinical decision recommendations; Among them, the intermediate output integration of each reasoning link specifically includes [current film visual features, current film lesion area visual features, previous film 1 visual features, previous film 1 lesion area visual features, …, previous film n visual features, previous film n lesion area visual features, medical record clinical data text features, preliminary report text features, multi-phase comparison reminder text features, multi-phase image comparison text features]; The update of the preliminary report based on the multi-period comparison is the revised interpretability multi-period comparison report.

5. The medical image multi-phase analysis system according to claim 4, characterized in that: The interpretable multi-period comparison report and the preliminary report update based on the multi-period comparison include, on the one hand, accurate multi-period comparison longitudinal diagnosis conclusions, and on the other hand, detailed multi-period comparison longitudinal diagnosis basis; The explainable multi-phase comparison report and the preliminary report update based on the multi-phase comparison include the specific coordinates of each lesion area, support the display of the corresponding multi-phase comparison longitudinal diagnosis conclusion and the multi-phase comparison longitudinal diagnosis basis, thereby forming a complete explainable link.

6. The medical image multi-phase analysis system according to any one of claims 1 to 5, characterized in that: The image diagnosis and analysis module, the multi-phase contrast warning module, the multi-phase image association diagnosis module and the multi-phase contrast and analysis module are all constructed based on a multimodal large model.

Images