Early Detection of Acral Lentiginous Melanoma with a Focus on Nail and Sole Areas Using YOLO

Melanoma is a highly aggressive and potentially deadly form of skin cancer, with Acral Lentiginous Melanoma (ALM) being more commonly found among individuals with darker skin tones, including those in Indonesia. Unlike other types of melanoma that are usually caused by sun exposure, ALM typically appears on parts of the body that are rarely exposed to ultraviolet (UV) light, such as the soles of the feet, palms of the hands, and under the nails. This unique characteristic often results in late diagnosis, as its symptoms are difficult to recognize and awareness among both the public and healthcare providers remains low. Early and accurate detection is crucial to improving survival rates and reducing the burden on the healthcare system. This study investigates the effectiveness of a Convolutional Neural Network (CNN) model based on YOLOv8 for detecting ALM lesions specifically on the soles of the feet and nail areas. The model was trained and evaluated using an annotated dermatological image dataset under three different training configurations to assess its robustness. The results indicate that the YOLOv8 model delivered promising performance, achieving mAP@0.5 scores ranging from 81.2% to 84.7%, along with significant improvements in F1 score, recall, and precision. The highest accuracy was obtained when the model was trained on a more varied and diverse dataset. The novelty of this research lies in its specific focus on detecting ALM in subungual and plantar regions using the YOLOv8 architecture and a custom augmented dataset, setting it apart from previous studies that typically rely on general melanoma datasets. These findings highlight the potential of YOLOv8 as an intelligent diagnostic tool for early ALM detection.