A Unified Neural Network Framework for Consistent and Efficient Real-time Object Detection of Early Longitudinal Melanonychia

This study presents a unified neural network framework for the early detection of longitudinal melanonychia, leveraging deep learning and computer vision technologies to achieve consistent and efficient object detection in real-time. Longitudinal melanonychia, a nail disease often overlooked until it necessitates surgical intervention or long-term treatment, has various causes and potential consequences that are not widely understood. The proposed framework aims to address this gap by enabling early identification, thereby reducing the condition's progression and raising awareness among individuals unfamiliar with nail diseases. The study employs the YOLOv10, a state-of-the-art model for fast and efficient computer vision system development. Manually curated datasets, focused exclusively on the early stages of longitudinal melanonychia, were used for training and validation. The model achieved a mean average precision (mAP) of 98.97% after 100 training epochs, demonstrating high accuracy and reliability. Detection was tested using a laptop webcam, confirming the framework's practical application for live monitoring and early intervention.