An adaptive depression processing method based on spatial perturbation

By employing adaptive spatiotemporal manifold curvature saliency filtering and topological isovariance feature space perturbation enhancement strategies, combined with a multimodal decision network based on evidence theory, the long-range dependency modeling and noise interference problems in rs-fMRI depression diagnosis were solved, achieving high-precision and robust auxiliary screening for depression.

CN122696306APending Publication Date: 2026-09-04HARBIN UNIV OF SCI & TECH
0 Cites 0 Cited by

Patent Information

Application Number
CN202610812419.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-06-07
Publication Date
2026-09-04

AI Technical Summary

Technical Problem

Existing rs-fMRI-based diagnostic methods for depression suffer from problems such as insufficient long-term dependency modeling, severe background noise interference, easy overfitting in small samples, and lack of adaptability in multimodal fusion, resulting in insufficient diagnostic accuracy and robustness.

Method used

We employ an adaptive spatiotemporal manifold curvature saliency filtering mechanism (AS-MCSF) to screen key voxel features. Combined with the topological isovariant feature space perturbation enhancement strategy (TE-FSPE) and the evidence theory-based multimodal cognitive uncertainty dynamic decision network (ET-MEUDD), we construct a super-order voxel state space evolution network (HV-SSEN) to achieve adaptive feature extraction and decision fusion.

Benefits of technology

It improves the accuracy and robustness of depression identification, effectively models long-range dependencies in the whole brain, removes background noise, alleviates overfitting in small samples, and enables personalized adaptive diagnosis.

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

Abstract

The application discloses a spatial disturbance-based adaptive depression processing method, and belongs to the technical field of medical image processing and artificial intelligence auxiliary diagnosis. The method comprises the following steps: firstly, performing multi-order space-time preprocessing on resting-state functional magnetic resonance imaging (rs-fMRI) data; secondly, constructing an adaptive space-time manifold curvature significant filtering mechanism (AS-MCSF) to screen out an intrinsic feature subset with high discriminability; extracting macroscopic brain region functional connection features and microscopic voxel discriminative features. Further, a hyper-order voxel state space evolution network (HV-SSEN) is constructed by introducing the Mamba state space modeling idea, and the network is optimized by combining a topological equivariant feature space disturbance enhancement strategy (TE-FSPE) and a density-sensitive focusing loss function. Finally, a multi-modal cognitive uncertainty dynamic decision network (ET-MEUDD) based on evidence theory is proposed, different modal cognitive uncertainty dynamic allocation fusion weights are allocated according to different modal cognitive uncertainty, adaptive fusion of multi-source heterogeneous features is realized, and finally, a depression diagnosis result is output.
Need to check novelty before this filing date? Find Prior Art