Computer-aided analysis of morphologic features in dermoscopic images of Port-wine stain birthmarks

Port-wine stain (PWS) birthmarks are characterized by complicated morphologic changes of cutaneous vasculature that cause abnormal discoloration of affected skin. Dermoscopy is a useful tool for visualizing PWS vascular features of skin surface, but dermoscopic images are often analyzed manually.

This study intended to develop a computer-aided image processing method for extracting and analyzing the morphologic features in PWS dermoscopic images.

Dermoscopic images of typical PWS lesions in the face and neck area were selected before and after photodynamic therapy (PDT). Image processing techniques, including image enhancement, noise removal and threshold segmentation, were applied to analyze the vascular characteristics of PWS images. Color, area and morphologic features of PWS vessels were extracted. The extracted features were statistically analyzed and quantified.

Objective evaluation showed that the best contrast effect was obtained by contrast limited adaptive histogram equalization (CLAHE) algorithm and the best image noise reduction results achieved by the non-local means (NLM) algorithm. Morphologic features were identified, extracted, and quantified from processed binarized images. Results showed that the dermoscopic feature of PWS lesion could be dominantly globule-like or network-like or mixed.

Computer-aided image processing method developed in this study might provide a useful tool for the quantitative analysis of dermoscopic images of PWS birthmarks.