Medical image collaborative diagnosis system and method based on image-report bidirectional association

Abstract

The invention relates to the technical field of medical image diagnosis, in particular to a medical image collaborative diagnosis system and method based on image-report two-way association.The system comprises a medical image visualization interaction module, an intelligent structured report editing module, an anchor point object generating module, an index establishing module, a medical image collaborative diagnosis module and an image-report two-way association.The medical image visualization interaction module renders a DICOM image and extracts three-dimensional coordinates and view parameters, and the intelligent structured report editing module generates an anchor point object and establishes an index; the image-text mapping relation management module maintains a dynamic mapping table of anchor points and coordinate parameters, and the view collaborative navigation control module monitors anchor point triggering and calls parameters and drives a view to jump to a corresponding anatomical horizon. According to the method, through construction of bidirectional dynamic mapping of anchor points and three-dimensional coordinates, accurate linkage of description and horizon is realized, response activation is carried out to drive view jump and restore a state, semantic analysis and atlas are utilized to verify spatial consistency so as to early warn conflicts, an original image is not changed by adopting independent image layer marking, and numbers are automatically rearranged and maintained. Cross-sequence linkage is supported based on matrix registration, and diagnosis and treatment data integrity and rechecking accuracy are guaranteed.

Description

Technical Field

[0001] This invention relates to the field of medical imaging diagnostic technology, and in particular to a medical imaging collaborative diagnostic system and method based on bidirectional image-report correlation. Background Technology

[0002] Currently, in the field of medical imaging diagnosis, Picture Archiving and Communication Systems (PACS) and Radiology Information Systems (RIS) are typically two independently operating software platforms. Radiologists, during their daily diagnostic work, need to browse DICOM image sequences containing hundreds of slices from CT and MRI scans on the PACS platform. After identifying lesions, they then switch to the report editing interface of the RIS system to describe the size, shape, and anatomical location of the lesion in plain text (e.g., "A low-density lesion is seen in the lower segment of the right posterior lobe of the liver"). This workflow results in a disconnect between "visual image information" and "report text information" at the data level, lacking a direct, interactive logical connection and relying solely on the doctor's memory and natural language descriptions for a loose link.

[0003] This separation of text and images presents significant efficiency bottlenecks and quality risks to the review and collaboration of imaging diagnoses. On the one hand, when senior reviewing physicians or clinicians read reports, they cannot directly locate the lesion in the original images based on the text descriptions in the reports. They often need to manually search and verify the images in hundreds of layers. Statistics show that the average time to locate a single lesion is as high as 30 seconds, which seriously affects the efficiency of image reading. On the other hand, the text descriptions of lesion locations are highly subjective, especially for junior physicians. If their descriptions of anatomical locations are inaccurate or vague, it is very easy for reviewing physicians to be unable to accurately reproduce the lesion location, leading to missed diagnoses, misdiagnoses, or high communication costs, which makes it difficult to meet the modern medical demand for precise and efficient collaborative diagnosis.

[0004] Existing image archiving and report editing platforms operate independently, and diagnostic work relies on fragmented textual descriptions of lesion attributes. This forces reviewing physicians to manually search for targets in massive image sequences based solely on subjective memory or abstract textual guidance. The disconnect between visual data and textual information leads to time-consuming lesion relocation. The vague or non-standardized anatomical descriptions by junior physicians are prone to causing misunderstandings in subsequent reviews, making it difficult to accurately reproduce the observation perspective at the initial diagnosis. Frequent switching between heterogeneous software interfaces disrupts the coherent cognitive workflow, exacerbates the risk of missed diagnoses and misjudgments, and hinders efficient data verification in multidisciplinary consultations. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a medical image collaborative diagnostic system and method based on bidirectional image-report correlation.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a medical image collaborative diagnosis system based on image-report bidirectional correlation, the system comprising: The medical image visualization interactive module is configured to load and render DICOM format medical image data, provide a multi-dimensional view display interface, and is equipped with a spatial coordinate capture unit. The spatial coordinate capture unit is used to respond to the user's operation commands in the image area and extract the three-dimensional spatial coordinates of the region of interest and the current view status parameters. The intelligent structured report editing module provides a rich text editing environment and is equipped with an anchor object generator, which is used to generate anchor objects with unique identifiers in the report text stream and establish an index relationship between the anchor objects and the report text description content. The image-text mapping relationship management module is communicatively connected to the medical image visualization interaction module and the intelligent structured report editing module, respectively, and is used to build and maintain a dynamic association mapping table. The mapping table stores the bidirectional mapping relationship between the unique identifier of the anchor point object and the three-dimensional spatial coordinates and view state parameters. The view collaborative navigation control module is configured to listen for anchor point trigger events in the intelligent structured report editing module. When a specific anchor point object is detected to be activated, the view collaborative navigation control module retrieves the corresponding three-dimensional spatial coordinates and view status parameters based on the dynamic association mapping table, and sends a view reset command to the medical image visualization interaction module to drive the image view to jump to the corresponding anatomical layer.

[0007] As a further aspect of the present invention, the view state parameters stored in the image-text mapping relationship management module include at least the following: Window width and window level parameters, image scaling factor, and cross-sectional direction identifiers for multi-plane reconstruction; When the view collaborative navigation control module sends a view reset command, it is configured to simultaneously send the view state parameters, so that the medical image visualization interaction module restores the rendering parameters to the display state when the anchor point object was created while jumping to the slice layer.

[0008] As a further aspect of the present invention, the system also includes an anatomical semantic and spatial consistency verification module, which is configured as follows: The text description associated with anchor objects in the intelligent structured report editing module is analyzed in real time using natural language processing algorithms to extract keywords related to anatomical locations; Call the pre-set standard human three-dimensional anatomy atlas library to obtain the standard spatial range corresponding to the keywords of the anatomical parts; Determine whether the three-dimensional spatial coordinates of the anchor point object stored in the image-text mapping relationship management module fall within the standard spatial range; If the judgment result is negative, a logical conflict signal is sent to the intelligent structured report editing module to trigger the generation of a warning label.

[0009] As a further aspect of the present invention, the medical image visualization interaction module is further configured with a dynamic layer overlay unit: The dynamic layer overlay unit is used to construct an independent transparent interactive layer on top of the original image data; When a piece of text in the intelligent structured report editing module is selected or focused, the dynamic layer overlay unit draws a highlighted marker box or guide indicator at the corresponding coordinate position of the transparent interactive layer according to the data of the image-text mapping relationship management module, without modifying the original DICOM pixel data.

[0010] As a further aspect of the present invention, the intelligent structured report editing module is configured with a sequence adaptive rearrangement unit: Used to monitor the sequence of anchor objects in the report text; When anchor objects in the report are deleted, cut, or moved, the sequence adaptive rearrangement unit automatically updates the sequence numbers of all remaining anchor objects and simultaneously notifies the image-text mapping relationship management module to update the index key values ​​in the dynamic association mapping table to maintain consistency between the report text number and the image tag number.

[0011] As a further aspect of the present invention, the image-text mapping relationship management module is also configured with a version control and auditing sub-module: It supports storing multiple independent association mapping tables for the same image sequence, corresponding to different user roles such as the initial diagnosis physician, the reviewing physician, and the clinical physician. When a high-privilege user jumps through the view collaborative navigation control module and corrects the lesion location, the submodule retains the original coordinate data, creates a new corrected coordinate record, and marks the correction status bit and the corrector's timestamp information in the dynamic association mapping table.

[0012] As a further aspect of the present invention, the spatial coordinate capture unit in the medical image visualization interaction module employs a ray projection picking algorithm: Configured to convert the user's click coordinates on the two-dimensional display plane into a projected ray; calculate the intersection of the projected ray with the currently displayed tomographic slice plane to obtain voxel coordinates; In volume rendering or maximum density projection view mode, the configuration is to sample along the projection ray path and identify the first voxel point or the voxel point with the maximum pixel value that meets the preset threshold as the target three-dimensional space coordinates.

[0013] As a further aspect of the present invention, the system also includes a cross-sequence spatiotemporal registration module: The configuration is set to acquire historical image sequences of the same patient and calculate the spatial transformation matrix between the current image sequence and the historical image sequences; When the view collaborative navigation control module responds to the anchor point triggering event reported in the current report, the cross-sequence spatiotemporal registration module uses the spatial transformation matrix to calculate the corresponding anatomical coordinates in the historical image sequence and drives the medical image visualization interaction module to synchronously display the corresponding layer of the historical image in split-screen mode.

[0014] A medical image collaborative diagnosis method based on image-report bidirectional correlation, wherein the method is executed based on the aforementioned medical image collaborative diagnosis system based on image-report bidirectional correlation, includes the following steps: S1: Configured to load and render DICOM format medical image data, providing a multi-dimensional view display interface, and equipped with a spatial coordinate capture unit. The spatial coordinate capture unit is used to respond to the user's operation commands in the image area, extract the three-dimensional spatial coordinates of the region of interest and the current view status parameters. S2: Provides a rich text editing environment, configured with an anchor object generator, used to generate anchor objects with unique identifiers in the report text stream, and establish an index relationship between the anchor objects and the report text description content; S3: It is connected to the medical image visualization interaction module and the intelligent structured report editing module respectively, and is used to build and maintain a dynamic association mapping table. The mapping table stores the bidirectional mapping relationship between the unique identifier of the anchor point object and the three-dimensional spatial coordinates and view state parameters. S4: Configured to listen for anchor point trigger events in the intelligent structured report editing module. When a specific anchor point object is detected to be activated, the view collaborative navigation control module retrieves the corresponding three-dimensional spatial coordinates and view state parameters based on the dynamic association mapping table, and sends a view reset command to the medical image visualization interaction module to drive the image view to jump to the corresponding anatomical layer.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, a bidirectional dynamic mapping mechanism between text anchor points and image 3D coordinates is constructed to achieve precise linkage between diagnostic descriptions and anatomical layers. Upon activation, the view is immediately redirected and the rendering states, such as window width and window level, are synchronously restored, eliminating visual errors associated with manual lesion retrieval. Natural language processing combined with standard anatomical atlases is used to verify the consistency between text semantics and spatial location. Real-time warnings of logical conflicts are provided to avoid descriptive errors. Independent transparent layers are used to draw highlighted markers without modifying the original pixel data. During text editing, the sequence is automatically rearranged to maintain the correspondence between image and text numbers. Cross-sequence registration based on a spatial transformation matrix supports synchronous linkage with historical images, effectively ensuring the integrity of diagnostic data and the accuracy of collaborative review. Attached Figure Description

[0016] Figure 1 This is a system flowchart of the present invention; Figure 2 This is a flowchart of the anatomical semantic verification and dynamic layer overlay processing of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0019] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0020] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0021] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0022] Example 1: System Overall Architecture Deployment, Hardware Environment and Underlying Data Communication Bus This embodiment first describes the basic architecture and data communication mechanism of a medical imaging collaborative diagnostic system based on bidirectional image-report correlation. This is the physical foundation for the collaborative operation of all subsequent functional modules. The system is physically built on top of a PACS (Picture Archiving and Communication System) environment within a hospital's local area network, employing a highly available distributed architecture. The server side deploys an image data center and a structured report database, while the client runs on a radiology diagnostic workstation. In a specific hardware deployment example, the diagnostic workstation is equipped with a high-performance CPU of Intel Core i9 or equivalent, at least 64GB of memory, and a professional-grade graphics card of NVIDIA RTX A4000 or higher to support CUDA parallel computing acceleration. The display devices consist of two 5MP (megapixel) medical grayscale displays for image browsing and one standard color display for report editing. At the software logic level, during system initialization, a global message bus is built in the client's memory, and the medical image visualization interaction module, intelligent structured report editing module, image-text mapping relationship management module, and view collaborative navigation control module are registered as independent nodes on this bus.

[0023] Unlike traditional tightly coupled software designs, the modules in this embodiment communicate asynchronously using a publish / subscribe pattern. For example, when the medical image visualization interaction module loads a chest CT sequence containing 5000 thin-slice scans, it publishes a standardized "Image_Loaded" event via the bus. This event payload includes the Series_Instance_UID (sequence unique identifier) ​​and the spatial resolution parameters of the image. After subscribing to this event, the image-text mapping relationship management module immediately allocates a corresponding hash mapping table space in memory. Meanwhile, when the intelligent structured report... When the report editing module is in focus, it publishes an "Editor_Focus" event. Upon receiving this event, the medical image visualization interaction module automatically activates its dynamic layer overlay unit. This event-driven bus architecture ensures that the system does not block the text input response of the report editor when processing high-load rendering of massive DICOM data. For example, when a doctor is rapidly scrolling through a 512x512 matrix CT image at a frame rate of 60fps, the message distribution latency in the background can still be controlled within 10 milliseconds, thus ensuring the real-time and smooth two-way interaction between text and images, and providing stable underlying support for subsequent accurate coordinate capture and semantic verification.

[0024] Example 2: Coordinate Capture and Dynamic Layer Rendering Mechanism Based on Ray Projection Based on the system architecture constructed in Embodiment 1, this embodiment describes in detail how the medical image visualization interaction module achieves the precise three-dimensional coordinate capture and non-destructive marking involved in the claims. When a doctor discovers a lesion during diagnosis and clicks the mouse to mark it, the spatial coordinate capture unit inside the module immediately starts the ray casting algorithm. The system first obtains the viewport transformation matrix, projection matrix, and model view matrix of the current rendering window, and uses the inverse matrix transformation to map the two-dimensional pixel coordinates of the screen into a ray in the three-dimensional view frustum. If the current view is in the two-dimensional view mode of multi-plane reconstruction (MPR), the system solves the ray equation and the plane equation of the current slice plane simultaneously to find the unique geometric intersection point as the voxel coordinates. In a specific application scenario, suppose a doctor is observing a transverse image of a patient with a lung nodule. The slice thickness is 1mm. The doctor clicks on a tiny nodule at screen coordinates (512, 512). The system instantly calculates the physical coordinates of this point in the patient's anatomical space as (x: -125.5mm, y: 88.0mm, z: 1042.5mm) and caches this precise coordinate. If the current view is in volumetric rendering (VR) or maximum density projection (MIP) 3D view mode, the system performs step sampling of the volume data along the ray vector direction. Combining this with the opacity transfer function or maximum grayscale projection rule, it calculates the first collision point or maximum density point between the ray and the surface of human tissue. For example, in vascular surgery, a doctor switches to MIP mode to observe the direction of blood vessels and clicks on a highlighted coronary artery stenosis. At this time, the algorithm penetrates the voxel data along the line of sight, automatically ignoring low-density lung parenchyma, and accurately captures the location of the contrast-filled blood vessel with the highest CT value (HU) as the target coordinates, thus solving the technical challenge of not being able to directly obtain depth from 3D projected images. Meanwhile, in response to the data security requirements in the claims, this module employs dynamic layer overlay technology. An independent OpenGL or DirectX overlay layer with a transparent background is created above the underlying DICOM image rendering pipeline. After the spatial coordinate capture unit calculates the three-dimensional coordinates, the system does not modify the original pixel data of the underlying layer. Instead, it projects the coordinates back to screen space and draws marker primitives at the corresponding positions on the transparent overlay layer. Specifically, when a doctor marks the aforementioned lung nodules, the system draws a 10mm diameter red circular frame on the overlay layer with accompanying text labels. When the doctor zooms in, zooms out, or pans the underlying image, the transparent overlay layer is synchronously transformed through a shared transformation matrix, ensuring that the red frame remains visually "attached" to the lesion location while achieving strict "layer separation" in terms of data. This ensures both the intuitiveness of the diagnosis and the legal validity of the original medical data.

[0025] Example 3: Intelligent Report Generation, Sequence Rearranging, and Semantic Analysis Verification Process Following the image marking operation completed in Example 2, this example focuses on how the intelligent structured report editing module uses the captured data to generate a document and combines the anatomical semantic and spatial consistency verification module to achieve intelligent quality control. When the doctor completes the marking on the image end and triggers the "insert reference" instruction, the report editing module receives the payload containing UUID and coordinate information through the message bus and generates an anchor object at the current cursor position. This object is rendered as a visual subscript on the front end and is bound to the complete spatial and view parameters captured in Example 2 on the back end. Taking a specific report writing scenario as an example, the doctor finds a ground-glass nodule in the apical posterior segment of the left upper lobe. After marking it, the doctor enters the text "A ground-glass density shadow is visible in the left upper lobe" in the report and clicks insert. The editor automatically generates a blue superscript "[1]".

[0026] In order to maintain the coherence of the report logic, the sequence adaptive reordering unit embedded in this module will monitor the changes of the document object model (DOM) in real time. In specific implementation, assuming that there are already lesions [1] (pulmonary nodules), [2] (lymph node enlargement), and [3] (pleural effusion) arranged in sequence in the report, if the reviewing doctor believes that the lymph node described in [2] has no clinical significance, and deletes the entire text containing the anchor point [2], the reordering algorithm will be triggered immediately. It will traverse the document tree, find that the original sequence is broken, and then automatically update the display text of the original anchor point [3] to [2], and send instructions to the image and text mapping relationship management module at the same time to update the background database index. This makes the numbering in the final printed report always continuous 1 and 2, completely eliminating the confusion that may be caused by manual modification of the numbering.

[0027] More importantly, in order to prevent medical errors such as "misidentification of left and right" or "misattribution of body parts", the system triggers an anatomical semantic verification process while generating anchor points. The natural language processing (NLP) engine running in the background of the system will analyze the text description around the anchor point in real time and extract the anatomical entity keywords. For example, a specific error correction example: Suppose that the doctor marked a low-density lesion in the right lobe of the liver on the image due to fatigue, but mistakenly wrote "low-density lesion visible in the spleen" in the report text description.[1] The NLP engine extracts the keyword "spleen" and calls the preset standard human three-dimensional anatomical atlas library to map the patient's image coordinates to the standard space through rigid registration. It is found that the coordinates of the marked point are within the "Liver Mask" range of the standard atlas, while the text description is "spleen". It is judged as a serious logical conflict. The system immediately sends an abnormal signal to the report editing module. The report editor will render a conspicuous yellow warning triangle icon next to the anchor point[1]. When the mouse hovers, a floating window pops up with the message: "Anatomy inconsistency warning: It was detected that you described it as 'spleen', but the marked point is located in the 'liver' anatomical area. Please verify." Thus, in the editing stage before the report is submitted, potential diagnostic quality accidents are intercepted by technical means.

[0028] Example 4: Cooperative navigation and cross-sequence registration applications based on full-state recovery This embodiment describes an advanced application in review and follow-up scenarios, mainly involving the linkage between the view collaborative navigation control module and the cross-sequence spatiotemporal registration module. When the reviewing physician or clinician opens the report generated by the aforementioned embodiment, the view collaborative navigation control module enters a listening state. Once the user clicks on an anchor point object in the report, the module immediately retrieves the complete data package associated with that anchor point from the image-text mapping relationship management module. Unlike the traditional method of only jumping to slices, this system performs "full-state scene restoration". Specifically, assuming that the initial physician is observing a fine fracture line, adjusts the image view to "bone window" (window width 2000HU, window level 500HU), magnifies the image by 3.5 times, and rotates it by 45 degrees to obtain the best viewing angle, when the reviewing physician (who may currently be looking at the soft tissue window) clicks on the anchor point link in the report, the system will send a compound command to the image interaction terminal, forcibly resetting the current rendering state instantly to the aforementioned bone window, 3.5x magnification, and 45-degree rotation state. This means that the image seen by the reviewing doctor is completely consistent with the image seen by the initial diagnosis doctor, achieving seamless transmission of diagnostic thinking and avoiding the risk of the reviewing doctor missing minor lesions due to improper window width settings.

[0029] Furthermore, in scenarios involving multiple examinations, such as tumor follow-up, the cross-sequence spatiotemporal registration module plays a core role; The specific implementation scenario is as follows: Patient Zhang underwent a follow-up examination after lung cancer surgery (Current Study). The baseline examination (Prior Study) three months ago was automatically retrieved. When the workstation loads these two sets of sequences, the background uses the Mutual Information algorithm to calculate the spatial transformation matrix between the two. When the doctor clicks the anchor point describing "new nodule [1]" in the current report, the view collaborative navigation control module not only drives the main screen to display the current new lesion, but also calls the transformation matrix to calculate the corresponding spatial position of the lesion coordinates in the historical images three months ago, and drives the old images displayed on the split screen to automatically jump. The system draws a dashed cursor at the corresponding position in the historical images, allowing doctors to see intuitively that three months ago, there was no nodule at that position, or only a tiny punctate shadow. This mechanism eliminates the need for doctors to manually search through hundreds of historical images to find the same anatomical level. They can directly compare changes in lesion volume (e.g., diameter from 0mm to 5mm) or density evolution (e.g., from ground-glass opacity to solid) in the same field of view, significantly improving the efficiency and accuracy of efficacy assessment (e.g., RECIST standard assessment). If the reviewing doctor corrects the lesion position during this process, the version control submodule in the image-text mapping relationship management module will retain the original record and create a correction record with the reviewer's signature and timestamp, thus forming a complete and traceable chain of collaborative diagnostic evidence.

[0030] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A medical image collaborative diagnostic system based on image-report bidirectional correlation, characterized in that, The system includes: The medical image visualization interactive module is configured to load and render DICOM format medical image data, provide a multi-dimensional view display interface, and is equipped with a spatial coordinate capture unit. The spatial coordinate capture unit is used to respond to the user's operation commands in the image area and extract the three-dimensional spatial coordinates of the region of interest and the current view status parameters. The intelligent structured report editing module provides a rich text editing environment and is equipped with an anchor object generator, which is used to generate anchor objects with unique identifiers in the report text stream and establish an index relationship between the anchor objects and the report text description content. The image-text mapping relationship management module is communicatively connected to the medical image visualization interaction module and the intelligent structured report editing module, respectively, and is used to build and maintain a dynamic association mapping table. The mapping table stores the bidirectional mapping relationship between the unique identifier of the anchor point object and the three-dimensional spatial coordinates and view state parameters. The view collaborative navigation control module is configured to listen for anchor point trigger events in the intelligent structured report editing module. When a specific anchor point object is detected to be activated, the view collaborative navigation control module retrieves the corresponding three-dimensional spatial coordinates and view state parameters based on the dynamic association mapping table, and sends a view reset command to the medical image visualization interaction module to drive the image view to jump to the corresponding anatomical layer.

2. The medical image collaborative diagnostic system based on image-report bidirectional correlation according to claim 1, characterized in that: The view state parameters stored in the image-text mapping relationship management module include at least the following: Window width and window level parameters, image scaling factor, and cross-sectional direction identifiers for multi-plane reconstruction; When the view collaborative navigation control module sends a view reset command, it is configured to simultaneously send the view state parameters, so that the medical image visualization interaction module restores the rendering parameters to the display state when the anchor point object was created while jumping to the slice layer.

3. The medical image collaborative diagnostic system based on image-report bidirectional correlation according to claim 1, characterized in that: The system also includes an anatomical semantic and spatial consistency verification module, which is configured as follows: The text description associated with anchor objects in the intelligent structured report editing module is analyzed in real time using natural language processing algorithms to extract keywords related to anatomical locations; Call the pre-set standard human three-dimensional anatomy atlas library to obtain the standard spatial range corresponding to the keywords of the anatomical parts; Determine whether the three-dimensional spatial coordinates of the anchor point object stored in the image-text mapping relationship management module fall within the standard spatial range; If the judgment result is negative, a logical conflict signal is sent to the intelligent structured report editing module to trigger the generation of a warning label.

4. The medical image collaborative diagnostic system based on image-report bidirectional correlation according to claim 1, characterized in that: The medical image visualization interactive module is further configured with a dynamic layer overlay unit: The dynamic layer overlay unit is used to construct an independent transparent interactive layer on top of the original image data; When a piece of text in the intelligent structured report editing module is selected or focused, the dynamic layer overlay unit draws a highlighted marker box or guide indicator at the corresponding coordinate position of the transparent interactive layer according to the data of the image-text mapping relationship management module, without modifying the original DICOM pixel data.

5. The medical image collaborative diagnostic system based on image-report bidirectional correlation according to claim 1, characterized in that: The intelligent structured report editing module is equipped with a sequence adaptive rearrangement unit: Used to monitor the sequence of anchor objects in the report text; When anchor objects in the report are deleted, cut, or moved, the sequence adaptive rearrangement unit automatically updates the sequence numbers of all remaining anchor objects and simultaneously notifies the image-text mapping relationship management module to update the index key values ​​in the dynamic association mapping table to maintain consistency between the report text number and the image tag number.

6. The medical image collaborative diagnostic system based on image-report bidirectional correlation according to claim 1, characterized in that: The image and text mapping relationship management module is also configured with a version control and auditing sub-module: It supports storing multiple independent association mapping tables for the same image sequence, corresponding to different user roles such as the initial diagnosis physician, the reviewing physician, and the clinical physician. When a high-privilege user jumps through the view collaborative navigation control module and corrects the lesion location, the submodule retains the original coordinate data, creates a new corrected coordinate record, and marks the correction status bit and the corrector's timestamp information in the dynamic association mapping table.

7. The medical image collaborative diagnostic system based on image-report bidirectional correlation according to claim 1, characterized in that: The spatial coordinate capture unit in the medical image visualization interactive module uses a ray projection picking algorithm. Configured to convert the user's click coordinates on the two-dimensional display plane into a projected ray; calculate the intersection of the projected ray with the currently displayed tomographic slice plane to obtain voxel coordinates; In volume rendering or maximum density projection view mode, the configuration is to sample along the projection ray path and identify the first voxel point or the voxel point with the maximum pixel value that meets the preset threshold as the target three-dimensional space coordinates.

8. The medical image collaborative diagnostic system based on image-report bidirectional correlation according to claim 1, characterized in that: The system also includes a cross-sequence spatiotemporal registration module: The configuration is set to acquire historical image sequences of the same patient and calculate the spatial transformation matrix between the current image sequence and the historical image sequences; When the view collaborative navigation control module responds to the anchor point triggering event of the current report, the cross-sequence spatiotemporal registration module uses the spatial transformation matrix to calculate the corresponding anatomical coordinates in the historical image sequence and drives the medical image visualization interaction module to synchronously display the corresponding layer of the historical image in split-screen mode.

9. A medical image collaborative diagnosis method based on image-report bidirectional correlation, characterized in that, The method is used to execute the medical image collaborative diagnostic system based on image-report bidirectional correlation as described in any one of claims 1-8, and includes the following steps: S1: Configured to load and render DICOM format medical image data, providing a multi-dimensional view display interface, and equipped with a spatial coordinate capture unit. The spatial coordinate capture unit is used to respond to the user's operation commands in the image area, extract the three-dimensional spatial coordinates of the region of interest and the current view status parameters. S2: Provides a rich text editing environment, configured with an anchor object generator, used to generate anchor objects with unique identifiers in the report text stream, and establish an index relationship between the anchor objects and the report text description content; S3: It is connected to the medical image visualization interaction module and the intelligent structured report editing module respectively, and is used to build and maintain a dynamic association mapping table. The mapping table stores the bidirectional mapping relationship between the unique identifier of the anchor point object and the three-dimensional spatial coordinates and view state parameters. S4: Configured to listen for anchor point trigger events in the intelligent structured report editing module. When a specific anchor point object is detected to be activated, the view collaborative navigation control module retrieves the corresponding three-dimensional spatial coordinates and view state parameters based on the dynamic association mapping table, and sends a view reset command to the medical image visualization interaction module to drive the image view to jump to the corresponding anatomical layer.

Images