A placenta image segmentation method based on a learnable visual center mechanism and a multilayer perception

By using a placental image segmentation algorithm based on LVC and MLP, the problem of insufficient global information extraction in existing technologies is solved, and high-precision placental image segmentation is achieved. By combining global and local features, the model complexity is reduced.

CN119540548BActive Publication Date: 2026-07-21ZHENGZHOU UNIV
View PDF 3 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
ZHENGZHOU UNIV
Filing Date
2024-10-10
Publication Date
2026-07-21

AI Technical Summary

Technical Problem

Existing placental image segmentation algorithms cannot achieve the segmentation level of professional physicians and cannot be used as a diagnostic basis. Furthermore, convolutional neural networks neglect the effective extraction of global information in placental image segmentation, resulting in insufficient segmentation accuracy.

Method used

A placental image segmentation algorithm based on Learnable Visual Center (LVC) and Multilayer Perceptron (MLP) is adopted. By marking MLP blocks to reduce parameters and computational complexity, the algorithm captures global long-distance dependencies by combining shift operations, and uses the LVC module to realize bottom-up interlayer feature interaction to enhance the representation of local detail features.

Benefits of technology

It achieves high-precision placental image segmentation, improves segmentation results, effectively combines global and local features, and reduces model complexity and computational burden.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119540548B_ABST
    Figure CN119540548B_ABST
Patent Text Reader

Abstract

The present application relates to a kind of placenta image segmentation algorithm based on learnable visual center mechanism and multilayer perception, can utilize fetal magnetic resonance imaging data to realize the automatic segmentation of placenta.The algorithm uses encoder-decoder architecture, introduces the labeling multilayer perception module to capture long distance long-distance dependence relationship of global feature and local corner area, realizes the complement of global and local feature information, makes up the deficiency of traditional convolution operation in this respect;In addition, the model also uses learnable visual center mechanism, by aggregating the local key area of input image, the context information of the region of interest obtained using deep features is used to regulate all positive shallow features, to realize the feature fusion between levels from top to bottom.Experimental results show that the proposed algorithm has achieved good results in placenta image segmentation task.
Need to check novelty before this filing date? Find Prior Art