Image Perceptual Similarity Metrics for the Assessment of Basal Cell Carcinoma

Efficient management of basal cell carcinomas (BCC) requires reliable assessments both tumors and post-treatment scars. We aimed to estimate image similarity metrics that account for BCC’s perceptual color texture deviation from perilesional skin. In total, 176 clinical photographs BCC were assessed by six physicians using a visual scale. Internal consistency inter-rater agreement estimated Cronbach’s α, weighted Gwet’s AC2, quadratic Cohen’s kappa. The mean scores used validate range employing different spaces, distances, embeddings pre-trained VGG16 neural network. calculated similarities transformed into discrete values ordinal logistic regression models. Bray–Curtis distance in the YIQ model rectified ‘fc6’ layer minimized squared error demonstrated strong performance representing similarities. Box plot analysis Wilcoxon rank-sum test visualize compare levels agreement, conducted on random validation round between two groups: ‘Human–System’ ‘Human–Human.’ proposed comparable terms internal with human raters. findings suggest offer robust cost-effective approach monitoring treatment outcomes settings.