February 2026 in “International journal of intelligent engineering and systems” This study proposes a new method for hair segmentation that improved performance in skin lesion images, as indicated by an increase in the Dice score from 76.97% to 79.08%.
3 citations
,
February 2024 in “arXiv (Cornell University)” In this study, researchers used Google Search ads to gather an open access dataset of 10,408 dermatological images from over 5,000 U.S. internet users, enhancing the diversity and representativeness of skin condition images available for research and artificial intelligence development.
In this study, researchers developed a method to create a synthetic dataset of facial acne images using generative techniques, achieving 97.6% classification accuracy with InceptionResNetv2, which helps overcome privacy concerns in biomedical applications by using anonymized data.
61 citations
,
June 2022 in “IEEE Journal of Biomedical and Health Informatics” This study introduced a novel deep clustering approach for melanoma detection from dermoscopic images, demonstrating improved performance over existing methods by mitigating class imbalance issues using a center-oriented margin-free triplet loss.
1 citations
,
January 2023 in “IEEE access” This review examines advancements in deep learning methods for detecting dermatological conditions from dermoscopic images, summarizing available datasets and suggesting future research directions, but reports no new results.