Skin Lesion Classification using Deep Learning Architectures
Problem
Classifying skin lesions from images to enable early detection of lesion types that can become malignant.
Why it matters
Some skin lesions (e.g., actinic keratosis, basal cell carcinoma) can become malignant, so early detection reduces cancer risk.
What the AI did
Deep learning architectures (MobileNet, VGG-16, and a custom model) classified six types of skin lesions from images to aid pre-diagnosis.
Results
Accuracy above 80% across six lesion classes (basal cell carcinoma, benign keratosis, dermatofibroma, vascular lesions, melanoma, melanocytic nevi)
Impact
Intended to aid doctors in pre-diagnosis; no clinical adoption or external validation stated in source.
Record
| Venue | International Conference on Communication System, Computing and IT Applications (CSCITA) |
|---|---|
| Published | 2020 |
| Work type | conference-paper |
| Authors | Abhishek C. Salian, Shalaka Vaze, Pragya Singh, Gulam Nasir Shaikh, Santosh Chapaneri, Deepak Jayaswal |
| Citations | 49 (OpenAlex, at import time) |
| References | 7 |
| Model or method | MobileNet, VGG-16, custom deep learning model |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1109/cscita47329.2020.9137810 |
| Open access | closed |
| Keywords | skin lesion, deep learning, classification, melanoma, early detection, MobileNet |
Importance rating
real but narrower advance
Abstract states skin cancer is one of the major types of cancers and that some lesion types can become malignant, so early detection cuts malignancy risk; solution is a pre-diagnosis aid without external validation.