Data augmentation importance for classification of skin lesions via deep learning
Problem
Detecting malignant skin lesions (melanoma) from images is hard when training data is insufficient; the study asked whether data augmentation improves deep-learning classifiers.
Why it matters
Melanoma is fatal but mostly curable if detected early, so accurate automated classification has life-or-death stakes.
What the AI did
The same deep-learning network was trained on augmented versus non-augmented skin lesion images; the augmented training achieved better malignant-lesion detection.
Results
Augmented-data training achieved better results than non-augmented training; no numeric metrics stated in abstract.
Impact
Shows data augmentation can improve deep-learning skin-lesion classification when datasets are limited.
Record
| Venue | Electric Electronics, Computer Science, Biomedical Engineerings' Meeting (EBBT) |
|---|---|
| Published | 2018 |
| Work type | conference-paper |
| Authors | Enes Ayan, Halil Murat Ünver |
| Citations | 77 (OpenAlex, at import time) |
| References | 22 |
| Model or method | not stated in source |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1109/ebbt.2018.8391469 |
| Open access | closed |
| Keywords | deep learning, data augmentation, melanoma, skin lesion, classification, medical imaging |
Importance rating
significant — thousands to millions affected
Abstract states melanoma is fatal but mostly curable if detected early, giving life-or-death stakes for accurate automated classification; study is a narrower augmentation-method advance with no scale stated.