Smart Image Follow-Up of Black Pigmentation on the Nail With Convolutional Neural Networks

Abstract

Background: It is a common reason for many patients to consult the dermatologists about black pigmentations on the nail because it maybe be related to the life-threatening malignant disease-melanoma. Although the incidence of nail melanoma is not too high, unfortunately, it is usually diagnosed at a later stage than at the other body sites and directly leads to a worse prognosis.\xa0 \n \nMethods: We have built the dataset which contains 550 standard, high-quality nail dermoscopy images from the outpatients. Relying on the dataset, we created an intelligent analysis system based on deep learning image segmentation. Meanwhile, we realized the quantitative calculation of black pigmentation on the features of shape, color, area and other indicators according to the clinical ABCDEF criteria to achieve medically interpretable index analysis.\xa0 \n \nFindings: Nail image segmentation model has a significant segmentation effect on the target area. The result of index analysis on 5 representative black line images may indicate that the black line is a benign lesion. Comparing the sample images with the quantitative indicators, our intelligent analysis system showed reliablity and interpretability. \n \nInterpretation: We developed a deep learning model based on the segmentation model to quantified multiple indexes of the ABCDEF rule. The intelligent analysis system can be helpful to analyze the dynamic index changes of black pigmentations on the nail, which is of significant aid in clinical diagnosis of subungual melanoma. \n \nFunding Information: None. \n \nDeclaration of Interests: Authors declare no Conflict of Interests for this article. \n \nEthics Approval Statement: This research has got the ethics committee approval from the Fifth Affiliated Hospital, Sun Yatsen University and its protocol number is [2021](K225-1).