Magnetic resonance image segmentation method based on artificial intelligence and linear attention
By combining convolutional neural networks and linear attention modules, a magnetic resonance image segmentation method is developed, which solves the problems of high computational resources and overfitting in high-resolution image segmentation of Transformer-CNN, and achieves efficient and accurate magnetic resonance image segmentation.
WO2026124412A1PCT designated stage Publication Date: 2026-06-18INNOVATION ACAD FOR PRECISION MEASUREMENT SCI & TECH CAS
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- INNOVATION ACAD FOR PRECISION MEASUREMENT SCI & TECH CAS
- Filing Date
- 2025-12-08
- Publication Date
- 2026-06-18
Smart Images
Figure CN2025140761_18062026_PF_FP_ABST
Abstract
Disclosed in the present invention is a magnetic resonance image segmentation method based on artificial intelligence and linear attention. The method comprises: acquiring a magnetic resonance image and a corresponding ground truth label, and generating a training set and a test set; constructing a magnetic resonance image segmentation network model; using a data set to train the magnetic resonance image segmentation network model, and acquiring an optimal weight of the magnetic resonance image segmentation network model with the objective of minimizing a total loss function; and using the magnetic resonance image segmentation network model with the optimal weight to segment a magnetic resonance image to be segmented, so as to output a segmentation map. The present invention proposes a linear multi-head self-attention module based on Taylor expansion, which module retains the capability of performing modelling on a long-distance dependency relationship between data, thereby enabling efficient and accurate segmentation with less memory and fewer parameters.
Need to check novelty before this filing date? Find Prior Art