Segmentation of Acne Vulgaris Images Techniques: A Comparative and Technical Study

Abstract

Background: Acne vulgaris is the most common dermatological pathology worldwide. The currently used methodologies for the evaluation and monitoring of acne have been analyzed in several studies, highlighting important limitations that can be concretely addressed using image processing methods by performing segmentation on different acne vulgaris image modalities. These techniques reduce the costs of treatment and acne severity grading, since they improve objectivity and are less time-consuming. That is why, in the last decade, several studies that propose segmentation methodologies on acne patients’ images have been published. The aim of this work is to analyze the segmentation methods developed for acne vulgaris images until now, including an analysis of the processing techniques and image modalities used, as well as the results. Results: Following the PRISMA statement and PICO model, 27 studies were included in the systematic review, and subsequently, they were divided into two groups: those that discuss methods based on classical image processing techniques, such as contrast adjustment and conversion of RGB images to other color spaces, and those discussing methods based on machine learning algorithms. Conclusions: Currently, there is no preference between one group of segmentation methods or the other. Moreover, the lack of uniformity in the evaluation of results for each study makes the comparison of methods difficult. The preferred image modality for segmentation is conventional photography, which shows a research gap in the application of segmentation algorithms to other acne vulgaris image modalities that could be useful, such as fluorescence imaging.