July 2024
·
74 Reads
British Journal of Dermatology
Whether dermatoscopy deep features could serve as biomarker for the prediction of melanoma metastasis, remains an underexplored area in medical research. In this cohort of 712 patients from ten centers from 3 different continents, a support vector machine (SVM) classifier, analyzing deep features from dermatoscopic images, demonstrated similar prognostic performance for metastasis, in terms of AUC and true positive rate, to current benchmarks of melanoma staging, namely Breslow thickness and ulceration. Deep features derived from dermatoscopy could predict, at the time of diagnosis, early-stage melanomas with high metastatic potential, tailoring further treatment strategies.