Related Experiment Video
Updated: Jan 10, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.3K
HHBSNet: a global channel-spatial attention and multi-scale dilated convolution network for automatic melasma
Shange Wang1, Lin Xu1, Linshuai Zhang1
1School of Intelligent Medicine, Chengdu University of Traditional Chinese Medicine, Chengdu, China.
Frontiers in Physiology
|November 21, 2025
Summary
A new deep learning model, HHBSNet, accurately segments melasma, a common facial hyperpigmentation. This lightweight network effectively handles irregular boundaries and lighting variations for improved computer-aided diagnosis.
Area of Science:
- Dermatology
- Computer Vision
- Medical Imaging
Background:
- Melasma is a prevalent facial hyperpigmentation disorder with irregular, blurred boundaries, posing segmentation challenges.
- Automated segmentation of melasma is difficult due to variations in lighting, skin reflections, and lesion morphology.
Purpose of the Study:
- To develop a lightweight and effective deep learning model for accurate melasma segmentation.
- To address the challenges of low contrast and irregular boundaries in melasma images.
Main Methods:
- Proposed HHBSNet, a novel lightweight segmentation network incorporating Global Channel-Spatial Attention (GCSA) and Multiscale Cavity Fusion (MCF) modules.
- GCSA module suppresses lighting interference and enhances feature discrimination.
- MCF module captures lesions at various scales without resolution loss; combined loss strategy addresses class imbalance.
Main Results:
- HHBSNet achieved superior performance on a dataset of 501 facial melasma images.
- Achieved a mean Intersection over Union (Miou) of 79.69%, accuracy (ACC) of 96.68%, F-score of 88.10%, recall of 88.18%, and precision of 87.80%.
Conclusions:
- HHBSNet demonstrates robust and superior segmentation for melasma.
- The model's lightweight design and generalization ability show potential for computer-aided diagnosis and clinical screening of pigmentary disorders.
