Integration of Localized, Contextual, and Hierarchical Features in Deep Learning for Improved Skin Lesion Classification
Problem
Classifying skin lesions from dermoscopic images is difficult because existing methods struggle with complex information and long-range dependencies.
Why it matters
Accurate skin lesion classification enables early detection and timely treatment of skin diseases.
What the AI did
A novel dual-track deep learning model combines local features from a modified DenseNet-169 with a Coordinate Attention Module and global features from a custom CNN with a Feature Pyramid Network and Global Context Network.
Results
not stated in source (abstract truncated before benchmark results)
Impact
Evidence supports a proposed improved architecture; performance claims are not shown in the abstract.
Record
| Venue | Diagnostics |
|---|---|
| Published | 2024 |
| Work type | article |
| Authors | R. Karthik, Illakiya Thayumanaswamy, Menaka Radhakrishnan, Daehan Won, Sindhia Lingaswamy |
| Citations | 24 (OpenAlex, at import time) |
| References | 38 |
| Model or method | dual-track model: modified DenseNet-169 + CoAM; custom CNN with FPN and GCN |
| Organization | not stated in source |
| Primary source | https://doi.org/10.3390/diagnostics14131338 |
| Open access | gold |
| Keywords | deep learning, skin lesion, dermoscopy, classification, attention, CNN |
Importance rating
real but narrower advance
Skin lesion classification aids early detection of skin diseases, a real benefit, but abstract presents an architecture improvement without stating disease burden or scale.