Intracranial aneurysm multi-center clinical image data standardized labeling and sharing platform

Abstract

The invention belongs to the technical field of clinical image processing, and particularly relates to an intracranial aneurysm multi-center clinical image data standardized labeling and sharing platform which comprises a multi-center operation end and a sharing platform management end, the multi-center operation end receives and preprocesses data, labeling is carried out in an artificial dominant and AI auxiliary mode, and the sharing platform management end is connected with the multi-center operation end. The model is optimized based on local manual correction data increment; and the sharing platform management end synchronously and uniformly marks specifications, generates a standardized data set through automatic checking and expert rechecking, and performs grading sharing according to roles. According to the method, data cross-center unified multiplexing is realized, the labeling efficiency is improved, sharing and privacy security are balanced, and multi-center research and clinical diagnosis and treatment are assisted.

Description

Technical Field

[0001] This invention belongs to the field of clinical image processing technology, specifically relating to a standardized annotation and sharing platform for multi-center clinical image data of intracranial aneurysms. Background Technology

[0002] Intracranial aneurysms are sac-like or fusiform protrusions formed by localized weakness or structural abnormalities in the walls of intracranial arteries under the impact of blood flow. Essentially, they are "pathological bulging" of the arterial wall. The core risk lies in rupture and bleeding. Once ruptured, they can cause subarachnoid hemorrhage, which has a high mortality rate. Even if the patient survives, there is a high probability of neurological dysfunction. It is a cerebrovascular disease with an extremely high mortality and disability rate in the field of neurosurgery.

[0003] Clinical imaging is the core basis for the "diagnosis-assessment-follow-up" of intracranial aneurysms, and promoting the sharing of clinical images is the key to breaking through the current bottlenecks in diagnosis, treatment, and research. Single-center imaging data has inherent limitations; a tertiary hospital can only diagnose a limited number of intracranial aneurysm cases annually, and these cases are relatively homogeneous due to regional patient characteristics, treatment equipment, and technological preferences (e.g., a regional hospital mainly deals with hypertension-related saccular aneurysms, with rare fusiform aneurysms or pediatric aneurysms). This makes it difficult to cover a full range of cases with different morphologies, risk factors, and treatment options. However, through image sharing, multiple centers can integrate thousands or even tens of thousands of diverse cases: on the one hand, this provides more comprehensive data support for rupture risk prediction; on the other hand, it allows for cross-institutional comparison of treatment outcomes, promoting unified treatment standards across different levels of hospitals and reducing prognostic disparities caused by regional technological differences.

[0004] Clinical image annotation, as a core component of "structured image information," suffers from inconsistent annotation standards across centers, severely hindering the realization of the value of image sharing. Furthermore, annotation remains primarily manual, requiring radiologists to browse DICOM format images frame by frame, manually outlining tumor contours and adjusting measurement tool parameters, resulting in time-consuming processes and inconsistent quality.

[0005] Therefore, there is an urgent need for a unified standard and sharing platform for the annotation of clinical images of intracranial aneurysms, which can enable rapid annotation and information sharing of clinical images of intracranial aneurysms. Summary of the Invention

[0006] To address the aforementioned shortcomings in the existing technology, this invention provides a standardized annotation and sharing platform for multi-center clinical imaging data of intracranial aneurysms, thereby resolving the problems mentioned in the background technology.

[0007] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: A standardized annotation and sharing platform for multicenter clinical imaging data of intracranial aneurysms, including a multicenter operation terminal and a sharing platform management terminal; The multi-center operating terminal is deployed in various medical institutions and includes a clinical imaging data access module, an intelligent image annotation module, and a local interaction module. The clinical imaging data access module is used to receive clinical imaging data and related clinical information of intracranial aneurysms uploaded by medical institutions and preprocess them. The intelligent image annotation module, based on the unified annotation specifications issued by the shared platform management terminal, uses an AI model to assist operators in identifying and annotating parameters of clinical imaging data. The local interaction module is used by operators to view annotation progress, retrieve historical annotation data, and upload structured annotation data to the shared platform management terminal. The shared platform management terminal includes a labeling standard management module, an image data labeling quality verification module, and a multi-center sharing module. The labeling standard management module has a built-in unified labeling specification for intracranial aneurysm images and needs to synchronously update the unified labeling specification to the multi-center operation terminals. The image data labeling quality verification module is used to receive structured labeled data uploaded by the multi-center operation terminals, automatically verify it through a built-in medical rule base, and generate a standardized dataset. The multi-center sharing module sets hierarchical access permissions according to user roles and supports anonymized query and download of the standardized dataset.

[0008] Furthermore, the clinical imaging data includes at least one of CTA, MRA, and DSA imaging data; the associated clinical information includes patient history, treatment plan, and postoperative follow-up results; the preprocessing includes format unification conversion, quality inspection, and privacy processing of the clinical imaging data and associated clinical information, specifically: unifying the conversion of clinical imaging data of different formats to the DICOM 3.0 standard format; performing quality inspection on the clinical imaging data, including evaluating signal-to-noise ratio, resolution, and artifact level; performing adaptive enhancement processing on clinical imaging data that does not meet the quality standards, including noise reduction, artifact removal, and contrast enhancement; and using hash encryption to desensitize patient identity information in the associated clinical information.

[0009] Furthermore, the AI ​​model is an auxiliary decision-making model that integrates the U-Net segmentation algorithm and the multimodal feature extraction algorithm. Based on clinical imaging data, it automatically identifies potential areas of intracranial aneurysms and generates initial annotation suggestions. The initial annotation suggestions include candidate lines of the aneurysm contour, predicted values ​​of core parameters, and measurement path prompts. At the same time, when the operator is annotating, the AI ​​model compares the operator's annotation content with the unified annotation standard in real time. When the annotated parameters deviate from the limits of the unified annotation standard or the measurement path does not meet the standard, it automatically triggers prompts and records them.

[0010] Furthermore, the training data for the AI ​​model comes from historical labeled data of intracranial aneurysms, and the training process uses cross-validation to optimize model parameters; and the AI ​​model supports dynamic optimization using an incremental learning mechanism.

[0011] Furthermore, the core parameters include the maximum diameter of the tumor, the width of the tumor neck, the diameter of the tumor-bearing artery, and the spatial distance between the tumor and surrounding nerves and blood vessels; the prompts triggered by the AI ​​model include the parameter reference range defined by the unified annotation standard, the recommended measurement section, the path calibration guide, and the annotation reference values ​​of similar historical cases; the recorded content includes the initial annotation suggestion, the operator's modified trajectory, the parameter deviation value, the prompt trigger time, and the operator's response result.

[0012] Furthermore, the unified annotation specifications of the annotation standard management module clarify the measurement standards for core parameters: the maximum diameter of the aneurysm is measured by constructing a three-dimensional model of the aneurysm using 3D image reconstruction technology, measuring the maximum straight-line distance between any two points on the model surface; the aneurysm neck width is measured by the narrowest diameter at the connection between the aneurysm and the parent artery, with the measurement direction perpendicular to the centerline of the long axis of the parent artery, and the measurement path avoiding calcified areas; the diameter of the parent artery is measured by the maximum transverse diameter of the normal vascular segment proximal to the aneurysm.

[0013] Furthermore, the structured labeled data is image-text related interactive data, including: Graphical annotation: In clinical images, the aneurysm area is marked with contour lines, the aneurysm neck is marked with arrows, and the course of the parent artery is marked with lines; Textual parameters: Records include the specific values ​​of core parameters, measurement methods, associated clinical information, and information of the annotators; The graphic annotations and text parameter data are bound to a unique case identifier, and clicking on the text parameter will automatically locate the graphic annotation area and highlight it.

[0014] Furthermore, the medical rule base of the image data annotation quality verification module includes logical consistency verification and parameter rationality verification; the logical consistency verification is used to verify the matching relationship between structured annotation data and core parameters; the parameter rationality verification is used to verify the proportional relationship between core parameters; data that fails automatic verification is submitted to experts for cross-verification, and after the verification is consistent, it is included in the standardized dataset.

[0015] Furthermore, the medical rule base supports dynamic updates: when the guidelines for the diagnosis and treatment of intracranial aneurysms or multicenter clinical consensus are updated, the platform administrator initiates a rule update application. After being reviewed and approved by experts from three or more core medical institutions, the new rules are automatically synchronized to the medical rule base and used for the verification of subsequent labeled data.

[0016] Furthermore, the hierarchical access permissions for the multi-center sharing module are as follows: researchers can only query and download the anonymized standardized labeled data and associated clinical statistical information, after which patient identifiable information is removed; clinicians can access the complete structured labeled data of their own patients and anonymized reference data from other hospitals; platform administrators are only responsible for user role configuration and access log management, and have no data access permissions.

[0017] Compared with the prior art, the present invention has the following beneficial effects: Through the annotation standard management module of the shared platform management terminal, the annotation specifications (including core parameter measurement standards and verification rules) are synchronized and unified to all multi-center operation terminals. Combined with the "automatic verification + expert cross-review" mechanism of the image data annotation quality verification module, it is ensured that the generated standardized datasets meet the unified cross-center standards. This transforms the annotation data of different medical institutions from "fragmented and unusable" to "universally reusable". It can not only support multi-center joint research, but also provide clinicians with de-identified reference data from other hospitals, and help make diagnosis and treatment decisions for rare cases. This invention adopts a "human-led, AI-assisted" annotation mode. The AI ​​model first outputs initial annotation suggestions (including tumor outline and predicted parameter values) and compares the human-annotated content with the specifications in real time, promptly indicating deviations. At the same time, incremental learning is initiated based on single-center human-corrected data (after verification), and the optimization process does not require the transmission of original data. This reduces the workload of manual correction (AI pre-annotation reduces repetitive operations) and continuously improves annotation accuracy through dynamic optimization of the local model, forming an efficiency closed loop of "annotation-optimization-re-annotation", which is suitable for the daily high-frequency annotation needs of medical institutions. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the structure of a standardized annotation and sharing platform for multi-center clinical imaging data of intracranial aneurysms according to the present invention. Detailed Implementation

[0019] To enable those skilled in the art to better understand the present invention, the technical solution of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0020] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual images. They should not be construed as limiting the scope of this application. To better illustrate the embodiments of the present invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual dimensions of the product. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0021] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "inner," and "outer" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present application. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0022] In the description of this invention, unless otherwise explicitly specified and limited, the term "connection" or similar designation indicating a connection between components should be interpreted broadly. For example, it can refer to a fixed connection, a detachable connection, or an integral part; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can refer to the internal communication between two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. Example

[0023] like Figure 1 As shown, this invention discloses a standardized annotation and sharing platform for multi-center clinical imaging data of intracranial aneurysms, aiming to solve problems such as inconsistent annotation standards, insufficient privacy protection during the sharing process, and low annotation efficiency in existing multi-center clinical imaging data. The overall architecture of this standardized annotation and sharing platform includes a multi-center operation terminal and a sharing platform management terminal. The two achieve bidirectional interaction through an encrypted data channel. The multi-center operation terminal is deployed in each medical institution to be responsible for local data processing and annotation, while the sharing platform management terminal realizes cross-center standard control, quality verification, and secure sharing.

[0024] Specifically, the multi-center operating terminal serves as the core carrier for local data processing and annotation in medical institutions. It includes a clinical imaging data access module, an intelligent image annotation module, and a local interaction module. These modules work together to complete the entire process of clinical imaging data processing from access to upload.

[0025] The core function of the clinical imaging data access module is to receive intracranial aneurysm clinical imaging data and related clinical information uploaded by medical institutions, and to ensure that the data meets the requirements for subsequent annotation and sharing through preprocessing. Clinical imaging data specifically includes at least one of CTA, MRA, and DSA imaging data, and related clinical information includes patient history, treatment plans, and postoperative follow-up results. The preprocessing process sequentially performs format unification conversion, quality inspection and enhancement, and privacy processing. The format unification conversion stage converts all non-standard format imaging data uploaded by different medical institutions into the DICOM 3.0 standard format to ensure subsequent cross-module data compatibility. The quality inspection stage uses image analysis algorithms to evaluate the signal-to-noise ratio, resolution, and artifact level of the data. The evaluation indicators refer to the minimum requirements for clinical imaging annotation, and adaptive enhancement processing is initiated for imaging data that does not meet the quality standards. All enhancement algorithms used are mature technologies in the field of medical image processing: noise is eliminated by searching for neighboring pixels similar to the target pixel in the image and calculating a weighted average, which is particularly suitable for preserving the small edges of aneurysms while reducing salt-and-pepper noise in CTA / MRA images; super-resolution reconstruction based on generative adversarial networks reconstructs low-resolution images into high-resolution images through adversarial training between the generator and discriminator, clearly restoring the subtle branching structure of the aneurysm-bearing artery; the adaptive artifact removal algorithm first locates metal artifacts (such as surgical clips) or motion artifact regions through edge detection, and then performs pixel filling repair based on the grayscale features of surrounding normal tissue, avoiding artifacts obscuring the aneurysm neck and affecting annotation. These algorithms are integrated into the clinical image data access module, automatically matching the image problem type and initiating corresponding processing without manual intervention. The accuracy of aneurysm boundary recognition in the enhanced images is improved, providing a clear data foundation for subsequent accurate annotation; in the privacy processing stage, patient identity information in associated clinical information is desensitized using hash encryption technology, removing identifiable information such as name, ID number, and hospital number, avoiding the risk of privacy leakage during data upload and sharing.

[0026] The intelligent image annotation module, based on the unified annotation specifications issued by the shared platform management terminal, achieves standardized annotation of clinical image data in a "human-led, AI-assisted" model. Its core function is to reduce manual annotation errors, improve annotation efficiency, and ensure that annotation results conform to unified standards across centers through AI models. This AI model is an auxiliary decision-making model that integrates the U-Net segmentation algorithm and multimodal feature extraction algorithm. During the AI ​​model training phase, historical annotation data of intracranial aneurysms from multiple tertiary hospitals are used as training samples. The training process optimizes model parameters through 5-fold cross-validation, ultimately achieving a Dice coefficient of no less than 0.92 for aneurysm region segmentation, with the deviation between the predicted values ​​of core parameters and the accurate manual annotation results being less than 5%. In the actual annotation process, the AI ​​model first automatically identifies potential intracranial aneurysm areas and generates initial annotation suggestions based on the grayscale, texture, and spatial structure features of clinical imaging data. These initial suggestions include candidate aneurysm contour lines, predicted values ​​of core parameters, and measurement path prompts. Core parameters cover the maximum diameter of the aneurysm, the neck width, the diameter of the parent artery, and the spatial distance between the aneurysm and surrounding neurovascular structures. When the operator performs manual annotation, the AI ​​model collects the operator's annotation trajectory and input parameters in real time, comparing them dimension-by-dimensionally with the unified annotation specifications. If the annotated parameters deviate from the specified range or the measurement path does not meet the standard, the AI ​​model immediately triggers a prompt. The prompt includes the parameter reference range specified by the unified annotation specifications, recommended measurement sections, path calibration guidelines, and annotation reference values ​​from similar historical cases, helping the operator correct deviations promptly. Simultaneously, the AI ​​model fully records key information during the annotation process, including initial annotation suggestions, operator-modified trajectories, parameter deviations, prompt trigger times, and operator responses. The AI ​​model backs up this content using timestamps and hash encryption, creating an immutable annotation operation log, providing a basis for subsequent quality verification and accountability. Furthermore, this module supports dynamic optimization of AI models, employing an incremental learning mechanism. The samples required for optimization originate from manually corrected data generated during the annotation process. This data is a natural byproduct of operators' standardized corrections to the initial pre-annotation results of the AI. After generation, it undergoes preliminary verification via a local interactive module at the multi-center operation terminal to ensure data integrity and traceability of corrections. When the accumulated amount of such manually corrected data at a single center reaches 100 cases, the platform automatically initiates a sample screening process, selecting only samples that have been uploaded to the shared platform management terminal and verified by the image data annotation quality verification module as new training data. This ensures that the data used for optimization conforms to a unified cross-center standard. The AI ​​model optimization phase employs 5-fold cross-validation to update parameters, and the entire optimization process is completed solely based on the manually corrected data from this center. There is no need to transmit the original image data or annotation data externally. This approach focuses on local AI annotation deviations for targeted optimization while further ensuring data security for medical institutions through data transmission.

[0027] The local interaction module provides operators with a visual interface, with core functions including viewing annotation progress, retrieving historical annotation data, and uploading structured annotation data. Operators can view the annotation completion status of current cases and the number of unannotated cases in real time through the interface, facilitating the rational allocation of annotation tasks. When retrieving historical annotation data, the system quickly retrieves corresponding image data and annotation records using the unique identifier of each case, allowing operators to compare annotation differences between different cases and assisting in current annotation decisions. In the structured annotation data upload phase, the module encapsulates the annotated data in a specific format and transmits it to the shared platform management terminal via an encrypted channel. The module provides real-time feedback on transmission progress and status during the upload process to ensure successful data submission.

[0028] The shared platform management terminal serves as the core of cross-center standard control and secure sharing. It includes a labeling standard management module, an image data labeling quality verification module, and a multi-center sharing module. These modules work together to achieve unified standard distribution, labeling quality control, and cross-center secure sharing.

[0029] The core function of the labeling standards management module is to formulate and maintain unified labeling standards for intracranial aneurysm images and to synchronously update these standards across all multi-center operating systems, ensuring consistency in labeling practices across medical institutions. The unified labeling standards clearly define the specific measurement standards for core parameters. For example, the maximum diameter of the aneurysm must first be measured by constructing a three-dimensional model of the aneurysm using 3D image reconstruction technology, and then measuring the maximum straight-line distance between any two points on the model surface. The aneurysm neck width is measured as the narrowest diameter at the connection between the aneurysm and the parent artery; the measurement direction must be perpendicular to the centerline of the parent artery's long axis, and the measurement path must avoid calcified areas. When calcification occupies more than 30% of the measurement section, a new measurement section without calcification must be selected. The diameter of the parent artery is measured as the maximum transverse diameter of the normal proximal segment of the aneurysm, avoiding the influence of vascular lesions on measurement accuracy. When intracranial aneurysm diagnosis and treatment guidelines or multi-center clinical consensus are updated, the module supports dynamic updates of the standards. An update request is initiated by the shared platform administrator and approved by associate chief physicians or higher from the neurosurgery or radiology departments of three or more core medical institutions. The new standards are then synchronously transmitted to all multi-center operating systems via an encrypted channel, ensuring that the labeling standards remain consistent with the latest clinical understanding.

[0030] The core function of the image data annotation quality verification module is to receive structured annotation data uploaded from multi-center operating terminals, filter datasets that meet the standards through a multi-level verification mechanism, and generate a unified standardized dataset across centers. This module has a built-in medical rule base, which includes two core rule categories: logical consistency verification and parameter rationality verification. Logical consistency verification checks the matching relationship between annotation labels and core parameters; for example, when the maximum diameter of the aneurysm is less than 3mm, the annotation label should be "microaneurysm." Parameter rationality verification checks the proportional relationship between core parameters; for example, the ratio of the aneurysm neck width to the diameter of the parent artery should be controlled within the range of 0.2 to 0.8. During the verification process, the system first automatically verifies the uploaded data through the medical rule base. Data that passes verification is directly included in the standardized dataset. For data that fails automatic verification, the system automatically assigns at least two experts (associate chief physicians or above in neurosurgery or radiology) for cross-review. Experts view the image data, annotation records, and reasons for verification failure through a dedicated review interface, and independently provide review opinions. Only when the two experts' review opinions are consistent and confirm that the annotation results meet the standards is the data included in the standardized dataset, ensuring that the final standardized dataset has clinical credibility. In addition, the medical rule base supports dynamic updates. When the guidelines for the diagnosis and treatment of intracranial aneurysms or multicenter clinical consensus are updated, the platform administrator initiates a rule update application. After being reviewed and approved by experts at the level of associate chief physician or above in neurosurgery or radiology from three or more core medical institutions, the new rules are automatically synchronized to the medical rule base and immediately applied to the verification of subsequent labeled data to ensure the timeliness of the verification standards.

[0031] The multi-center sharing module is based on the principles of "tiered authorization and secure sharing," setting differentiated access permissions according to user roles. Its core function is to meet the usage needs of different roles for standardized datasets while ensuring data privacy. Specific permission settings are as follows: Researchers can only query and download anonymized standardized labeled data and associated clinical statistical information through the platform. The anonymization process removes all patient-identifiable information, retaining only annotation parameters, associated clinical information summaries, and unique case identifiers, supporting researchers in AI model training or clinical statistical analysis. Clinicians can access complete structured labeled data of their own patients, including original labeled images, detailed parameter records, and associated clinical information, for case comparison and treatment plan development in daily diagnosis and treatment. They can also access anonymized reference data from other hospitals, providing multi-center experience support for the diagnosis and treatment of rare cases. Platform administrators are only responsible for user role configuration, permission changes, and access log management, and have no data access permissions. This separation of permissions prevents administrators from abusing their privileges to obtain data. During the sharing process, all data transmission is completed through an encrypted channel to ensure data security during transmission. The system automatically records access logs for all users, including access time, data type, and operation content, providing a basis for data security audits.

[0032] The overall workflow of this invention is as follows: Each medical institution uploads and preprocesses data through the clinical imaging data access module of the multi-center operating terminal. After standardized annotation is completed by the image intelligent annotation module, the data is uploaded to the shared platform management terminal through the local interaction module. The image data annotation quality verification module of the shared platform management terminal automatically verifies and reviews the uploaded data with experts, generating a standardized dataset. Finally, the multi-center sharing module opens up the query and download services of the corresponding data to researchers and clinicians according to user role permissions. At the same time, the annotation standard management module continuously updates the annotation specifications to the multi-center operating terminal. The AI ​​model continuously optimizes its performance through incremental learning, forming a closed loop of "data processing - standardized annotation - quality verification - secure sharing - model optimization", realizing the efficient utilization and secure sharing of multi-center clinical imaging data of intracranial aneurysms.

[0033] The above are merely embodiments of the present invention. The circuits, electronic components, and modules involved are all prior art, fully achievable by those skilled in the art, and require no further explanation. The scope of protection in this application does not involve improvements to the software and methods. Commonly known structures and characteristics in the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all prior art in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent.

Claims

1. A standardized annotation and sharing platform for multi-center clinical imaging data of intracranial aneurysms, characterized in that... This includes multi-center operation terminals and shared platform management terminals; The multi-center operating terminal is deployed in various medical institutions and includes a clinical imaging data access module, an intelligent image annotation module, and a local interaction module. The clinical imaging data access module is used to receive clinical imaging data and related clinical information of intracranial aneurysms uploaded by medical institutions and preprocess them. The intelligent image annotation module, based on the unified annotation specifications issued by the shared platform management terminal, uses an AI model to assist operators in identifying and annotating parameters of clinical imaging data. The local interaction module is used by operators to view annotation progress, retrieve historical annotation data, and upload structured annotation data to the shared platform management terminal. The shared platform management terminal includes a labeling standard management module, an image data labeling quality verification module, and a multi-center sharing module. The labeling standard management module has a built-in unified labeling specification for intracranial aneurysm images and needs to synchronously update the unified labeling specification to the multi-center operation terminals. The image data labeling quality verification module is used to receive structured labeled data uploaded by the multi-center operation terminals, automatically verify it through a built-in medical rule base, and generate a standardized dataset. The multi-center sharing module sets hierarchical access permissions according to user roles and supports anonymized query and download of the standardized dataset.

2. The standardized annotation and sharing platform for multi-center clinical imaging data of intracranial aneurysms as described in claim 1, characterized in that: The clinical imaging data includes at least one of CTA, MRA, and DSA imaging data; the associated clinical information includes patient history, treatment plan, and postoperative follow-up results; the preprocessing includes format unification conversion, quality inspection, and privacy processing of the clinical imaging data and associated clinical information, specifically: unifying the conversion of clinical imaging data of different formats to the DICOM 3.0 standard format; performing quality inspection on the clinical imaging data, including evaluating signal-to-noise ratio, resolution, and artifact level; and performing adaptive enhancement processing on clinical imaging data that does not meet the quality standards, including noise reduction, artifact removal, and contrast enhancement. Patient identity information in the associated clinical information is desensitized using hash encryption.

3. The standardized annotation and sharing platform for multi-center clinical imaging data of intracranial aneurysms as described in claim 1, characterized in that: The AI ​​model is an auxiliary decision-making model that integrates the U-Net segmentation algorithm and the multimodal feature extraction algorithm. It automatically identifies potential areas of intracranial aneurysms based on clinical imaging data and generates initial annotation suggestions. The initial annotation suggestions include candidate lines of the aneurysm contour, predicted values ​​of core parameters, and measurement path prompts. At the same time, when the operator is annotating, the AI ​​model compares the operator's annotation content with the unified annotation standard in real time. When the annotated parameters deviate from the limits of the unified annotation standard or the measurement path does not meet the standard, it automatically triggers prompts and records them.

4. The standardized annotation and sharing platform for multi-center clinical imaging data of intracranial aneurysms as described in claim 3, characterized in that: The training data for the AI ​​model comes from historical labeled data of intracranial aneurysms, and the training process uses cross-validation to optimize model parameters; moreover, the AI ​​model supports dynamic optimization using an incremental learning mechanism.

5. The standardized annotation and sharing platform for multi-center clinical imaging data of intracranial aneurysms as described in claim 3, characterized in that: The core parameters include the maximum diameter of the tumor, the width of the tumor neck, the diameter of the tumor-bearing artery, and the spatial distance between the tumor and surrounding nerves and blood vessels; the prompts triggered by the AI ​​model include the parameter reference range defined by the unified annotation standard, the recommended measurement section, the path calibration guide, and the annotation reference values ​​of similar historical cases; the recorded content includes the initial annotation suggestion, the operator's modified trajectory, the parameter deviation value, the prompt trigger time, and the operator's response result.

6. The standardized annotation and sharing platform for multi-center clinical imaging data of intracranial aneurysms as described in claim 1, characterized in that: The unified annotation specifications of the annotation standard management module clearly define the measurement standards for core parameters: the maximum diameter of the aneurysm is measured by constructing a three-dimensional model of the aneurysm using 3D image reconstruction technology, and measuring the maximum straight-line distance between any two points on the model surface; the aneurysm neck width is measured by the narrowest diameter at the connection between the aneurysm and the parent artery, with the measurement direction perpendicular to the center line of the long axis of the parent artery, and the measurement path avoiding calcified areas; the diameter of the parent artery is measured by the maximum transverse diameter of the normal vascular segment proximal to the aneurysm.

7. The standardized annotation and sharing platform for multi-center clinical imaging data of intracranial aneurysms as described in claim 1, characterized in that: The structured labeled data is image-text related interactive data, including: Graphical annotation: In clinical images, the aneurysm area is marked with contour lines, the aneurysm neck is marked with arrows, and the course of the parent artery is marked with lines; Textual parameters: Records include the specific values ​​of core parameters, measurement methods, associated clinical information, and information of the annotators; The graphic annotations and text parameter data are bound to a unique case identifier, and clicking on the text parameter will automatically locate the graphic annotation area and highlight it.

8. The standardized annotation and sharing platform for multi-center clinical imaging data of intracranial aneurysms as described in claim 1, characterized in that: The medical rule base of the image data annotation quality verification module includes logical consistency verification and parameter rationality verification. The logical consistency verification is used to verify the matching relationship between structured annotation data and core parameters. The parameter rationality verification is used to verify the proportional relationship between core parameters. Data that fails the automatic verification is submitted to experts for cross-verification. After the cross-verification is consistent, it is included in the standardized dataset.

9. The standardized annotation and sharing platform for multi-center clinical imaging data of intracranial aneurysms as described in claim 8, characterized in that: The medical rule base supports dynamic updates: when the guidelines for the diagnosis and treatment of intracranial aneurysms or multicenter clinical consensus are updated, the platform administrator initiates a rule update application. After being reviewed and approved by experts from three or more core medical institutions, the new rules are automatically synchronized to the medical rule base and used for the verification of subsequent labeled data.

10. The standardized annotation and sharing platform for multi-center clinical imaging data of intracranial aneurysms as described in claim 1, characterized in that: The hierarchical access permissions of the multi-center sharing module are as follows: researchers can only query and download the desensitized standardized labeled data and related clinical statistical information, after which patient identifiable information is removed; clinicians can access the complete structured labeled data of patients in their own hospital and desensitized reference data from other hospitals; Platform administrators are only responsible for user role configuration and access log management, and do not have data access permissions.

Images