A medical record image de-identification method based on a multi-modal large model

By using a multimodal large model for medical record image preprocessing and sensitive information identification, the problem of insufficient generalization ability in traditional medical record image processing is solved, and efficient and automated medical record image de-identification is achieved, adapting to diverse medical institution medical record formats.

CN122413461APending Publication Date: 2026-07-17广州中康数字科技有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
广州中康数字科技有限公司
Filing Date
2026-03-05
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

Traditional methods for de-identifying medical record images lack generalization ability when faced with diverse medical record formats from different medical institutions, resulting in low processing efficiency and requiring manual intervention to adjust the rules.

Method used

A multimodal large model is used for medical record image preprocessing, multimodal information fusion and sensitive information identification. NLP and reasoning capabilities are used for semantic analysis to identify and locate sensitive information, and de-identified medical record images are generated through pixel filling.

Benefits of technology

It achieves efficient de-identification of medical records from different medical institutions, reduces manpower and time costs, has strong generalization capabilities, and does not rely on fixed templates or formats.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

本发明公开了一种基于多模态大模型的病历图像去标识化方法,本发明的核心在于免人工标注以及强泛化能力。不再依赖于固化的规则或固定的版式,而是利用多模态大模型同时理解图像视觉布局与文本深层语义的综合能力,在不经过针对性训练的情况下,高效处理各种未知格式的病历图像,从根本上解决了传统方法的泛化难题。本发明对不同医院、不同科室、不同类型的病历文书(如入院记录、出院记录、检验报告等)均具有良好的适应性,具有优秀的泛化能力。由于不依赖大量针对性的标注数据和复杂的匹配规则,本发明还可以显著降低因病历格式变更而带来的人力与时间成本。
Need to check novelty before this filing date? Find Prior Art