An automatic evaluation system for Treatment Effects on Port-wine Stains based mapping algorithm and deep learning

Abstract

Objectives A port-wine stain (PWS) is a common type of capillary malformation that often occurs on the head and neck, seriously affecting patients' appearance. Currently, laser phototherapy devices are mainly used to treat PWS. The accuracy of lesion efficacy evaluation results affects the rational selection of treatment plans. In clinical practice, visual assessment methods are commonly used to judge the efficacy of this disease by estimating the degree of improvement in lesions, which is highly subjective and difficult to quantify. To achieve convenient and accurate efficacy evaluation, many image-based evaluation schemes have been proposed. However, these schemes usually require doctors to manually select random areas for color comparison, making it difficult to ensure comprehensiveness and objectivity when evaluating results. To address existing problems in previous studies, we propose an automatic method for evaluating PWS treatment effects. Methods By implementing steps such as image correction, lesion area segmentation, and image mapping, color difference comparisons based on all lesion areas and all normal skin color areas can be achieved to realize more objective and accurate treatment evaluations. Results We verify the proposed method through consistency experiments. In the experiments, the highest consistency between our proposed method and the findings of three professional doctors reached 88.89%, which was higher than their highest consistency rate of 85.19%. Conclusions The experimental results show that this method can significantly improve both efficiency and accuracy in evaluating the effects of PWS treatment.