A disease risk early warning and efficacy prediction model fusing phenotype-image-molecular network

By integrating phenotypic, imaging, and molecular network modeling methods, and utilizing large language and visual models for feature extraction and self-supervised learning, the problem of insufficient information mining in multimodal learning methods is solved, enabling efficient and accurate prediction of disease risk warning and drug efficacy.

CN122266748APending Publication Date: 2026-06-23TSINGHUA UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
TSINGHUA UNIVERSITY
Filing Date
2026-02-12
Publication Date
2026-06-23

AI Technical Summary

Technical Problem

Existing multimodal learning methods have failed to fully exploit image modal information, making it difficult to reveal the multi-scale dynamic mechanisms of disease development. Furthermore, traditional models lack interpretability, resulting in limited accuracy in early disease detection and drug efficacy prediction.

Method used

We employ a modeling approach that integrates phenotype, imaging, and molecular networks. We extract features using large language and visual models, and combine attention mechanisms and self-supervised learning to perform cross-modal feature alignment and fusion, thereby constructing a disease risk warning and efficacy prediction model.

Benefits of technology

It enables efficient and accurate disease risk warning and drug efficacy prediction, improves the accuracy of early disease detection, and provides a biologically interpretable intelligent diagnosis and treatment system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122266748A_ABST
    Figure CN122266748A_ABST
Patent Text Reader

Abstract

The application provides a disease risk early warning and efficacy prediction model of a fusion phenotype-image-molecular network. The modeling method corresponding to the model comprises the following steps: firstly, an attention mechanism network is constructed to extract multi-modal features; secondly, a unified public embedding space is constructed, and cross-modal feature alignment and fusion are realized by using self-supervised learning; further, combined with disease progression and drug information, disease risk early warning and drug efficacy prediction are carried out based on the fusion features, and the whole process intelligent evaluation of the disease from occurrence, development to intervention is realized. The application is significantly better than the baseline method in tumor occurrence risk grading and drug efficacy prediction. The multi-modal disease risk early warning and efficacy prediction model provided by the application provides a new way for revealing the disease mechanism from the macro (Chinese and Western phenotypes), the medium (image features) and the micro (molecular level), and promotes the development of precision medicine and intelligent health management of Chinese and Western medicine integration.
Need to check novelty before this filing date? Find Prior Art