A novel approach for skin lesion symmetry classification with a deep learning model
Problem
Early skin cancer diagnosis requires assessing lesion symmetry in dermoscopic images, which is difficult to do consistently and automatically.
Why it matters
Abstract states early diagnosis can reduce the malignancy risk of skin cancer, a growing public health problem.
What the AI did
A CNN classifies lesion symmetry into three classes; authors built a labeled dataset of 615 lesions and compared transfer learning against traditional methods.
Results
Pipeline outperforms traditional and pre-trained approaches; weighted-accuracy value truncated in abstract.
Impact
Provides a fast symmetry-classification pipeline and a new labeled dermoscopy dataset for research use.
Record
| Venue | Computers in Biology and Medicine |
|---|---|
| Published | 2022 |
| Work type | article |
| Authors | Lidia Talavera-Martínez, Pedro Bibiloni, Aniza Giacaman, R. Taberner, Luis Javier Del Pozo, Manuel González-Hidalgo |
| Citations | 35 (OpenAlex, at import time) |
| References | 46 |
| Model or method | CNN (no specific architecture named) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1016/j.compbiomed.2022.105450 |
| Open access | hybrid |
| Keywords | skin lesion, symmetry, CNN, dermoscopy, deep learning, transfer learning |
Importance rating
significant — thousands to millions affected
Abstract states skin cancer is a growing public health problem and early diagnosis reduces malignancy risk; a fast symmetry-classification aid plus a new 615-lesion dataset offers significant diagnostic benefit.