Related Experiment Video
Updated: Jul 5, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
MSDEnet: Multi-scale detail enhanced network based on human visual system for medical image segmentation.
Yuangang Ma1, Hong Xu2, Yue Feng1
1Department of Intelligent Manufacturing, Wuyi University, China.
Computers in Biology and Medicine
|January 23, 2024
Summary
This study introduces a novel multi-scale detail enhanced network for medical image segmentation. The new method improves accuracy on images with unclear boundaries, outperforming existing algorithms.
Area of Science:
- Medical Imaging
- Computer Vision
- Artificial Intelligence
Background:
- Medical image segmentation accuracy decreases for images with unclear boundaries, like certain skin or CT scans.
- Existing segmentation methods struggle with subtle details and similar feature values between adjacent structures.
Purpose of the Study:
- To develop an advanced deep learning network for enhanced medical image segmentation.
- To improve segmentation accuracy, particularly for images with low contrast or ambiguous boundaries.
Main Methods:
- Proposed a multi-scale detail enhanced network inspired by the human visual system.
- Introduced a detail enhanced module using asymmetric and standard convolutions to improve contrast.
- Incorporated a channel multi-scale module (adapted from Res2net) to increase effective receptive field and utilize redundant information.
Main Results:
- The proposed network achieved superior performance compared to common medical image segmentation algorithms across four datasets.
- Ablation studies confirmed the significant contribution of each proposed module to the overall performance.
Conclusions:
- The multi-scale detail enhanced network effectively addresses limitations in segmenting medical images with subtle or ambiguous features.
- This approach offers a promising advancement for accurate and reliable medical image segmentation.
