A medical image complementary fusion method, device, medium and program product based on prior feature guidance

CN122391805APending Publication Date: 2026-07-14ZHONGSHAN HOSPITAL FUDAN UNIV +1
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
ZHONGSHAN HOSPITAL FUDAN UNIV
Filing Date
2026-04-21
Publication Date
2026-07-14

AI Technical Summary

Technical Problem

Existing medical image fusion methods lack deep interaction, making it difficult to effectively preserve the specific information of each feature in a unified feature space. Furthermore, the representational ability after feature fusion is insufficient, resulting in weak robustness and generalization ability.

Method used

A medical image complementary fusion method based on prior features is adopted. The first modality feature vector is used to calculate refined features by guiding the second modality feature vector, and a composite feature sequence containing four complementary information is constructed. The nonlinear dependency relationship between different feature streams is learned by using a Transformer encoder.

Benefits of technology

It significantly improves the purity and robustness of features, enabling the generation of uniform and robust patient-level representations in the face of heterogeneous inputs, thereby improving the accuracy of intelligent diagnosis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122391805A_ABST
    Figure CN122391805A_ABST
Patent Text Reader

Abstract

The application relates to the field of intelligent medicine and discloses a medical image complementary fusion method, equipment, medium and program product based on prior feature guidance. The method comprises the following steps: acquiring a first modality image of a first organ and a second modality image of a second organ of a target object and encoding projection to obtain a first modality feature vector and a second modality feature vector; calculating a first refined feature by guiding the first modality feature vector with the second modality feature vector; simultaneously calculating a second refined feature by guiding the second modality feature vector with the first modality feature vector; splicing the first modality feature vector, the second modality feature vector, the first refined feature and the second refined feature according to a channel dimension to construct a composite feature sequence containing four-way complementary information; and inputting the composite feature sequence into a feature fusion network to output a multi-modality complementary feature. The application can timely find whether a patient has a super progress through joint modeling.
Need to check novelty before this filing date? Find Prior Art