ASGS — Anatomy-Guided Visceral Fat Segmentation
An anatomy-guided deep-learning study for robust visceral fat segmentation in PET/CT images.
Overview
This study investigates anatomy-guided segmentation for automatic quantification of visceral fat in PET/CT images.
Beyond local image intensity, anatomical structure is incorporated to produce consistent segmentation across patient-specific variations.
Problem
- Ambiguous boundaries between visceral fat and surrounding tissue
- Segmentation variability caused by patient-specific body shape and anatomy
- The need for repeatable automatic quantification in large-scale medical-image analysis
Approach
- PET/CT medical-image dataset construction and preprocessing
- Deep-learning-based visceral fat segmentation with anatomical guidance
- Comparative experiments for model performance and segmentation robustness
Results
- Published in Engineering Applications of Artificial Intelligence · 2026
- SCIE · Impact Factor 9.0 · Top 3.1% in JCR
- Contributed as second author to dataset construction, comparative experiments, and result analysis