Epilepsy prediction method based on adaptive spatiotemporal attention and dynamic fusion network

By using the method of adaptive spatiotemporal attention and dynamic fusion network in epilepsy prediction, the problems of local or short-term dependence modeling, low efficiency and high computational complexity in the prior art are solved, and high accuracy and real-time epilepsy prediction are achieved.

CN119344751BActive Publication Date: 2025-04-29ANHUI UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202411897459.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-12-23
Publication Date
2025-04-29
Estimated Expiration
2044-12-23

AI Technical Summary

Technical Problem

The prior art has the limitations of local or short-term dependency modeling, low efficiency of long-term dependency modeling and high computational complexity in epilepsy prediction.

Method used

Using an adaptive spatiotemporal attention and dynamic fusion network, multi-level dependence relationships and space-time dynamic characteristics in EEG signals are captured through brain signal preprocessing module, context attention coding layer, long-distance content dependence network, adaptive timing state space modeling module and epilepsy prediction module.

Benefits of technology

The multi-level dependency modeling is realized, which improves the accuracy and real-time prediction, reduces the computational complexity, and makes the model suitable for low-latency and real-time prediction scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119344751B_ABST
    Figure CN119344751B_ABST
Patent Text Reader

Abstract

The present invention discloses an epilepsy prediction method based on an adaptive spatiotemporal attention and dynamic fusion network, which involves a brain signal preprocessing module, a contextual attention coding layer, a long-range content-dependent network, an adaptive temporal state space modeling module and an epilepsy prediction module; the brain signal preprocessing module obtains corresponding time domain information and frequency domain information and an adjacency matrix; the adjacency matrix is ​​regarded as a graph structure, the weights of each edge of the graph structure are input into the contextual attention coding layer, the spatial features between each channel are extracted, the weights are adjusted dynamically through the aggregation of neighbor node information and attention, and the node feature matrix is ​​output; the node feature matrix is ​​input into the long-range content-dependent network to output global spatial features and corresponding time series; the spatiotemporal dynamic fusion features are obtained through the adaptive temporal state space modeling module; the spatiotemporal dynamic fusion features are input into the epilepsy prediction module and the epilepsy prediction result is output. The present invention can accurately predict.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Epileptic seizure prediction model training method and epileptic seizure prediction system

    CN117562551A

  • Epilepsy prediction method and system in combination with spatial-temporal characteristics

    CN118072947A