Automated System for Acne Vulgaris Grading Using Self-Organizing Map

Javed Khan, Aamir Saeed Malik, Nidal Kamel, Sarat Chandra Dass, Azura Mohd Affandi

Research output: Contribution to journalArticle

Abstract

Acne vulgaris is a chronic skin abnormality that can afflict a person at any stage of life; however, it is more common in adolescent population. Due to the subjectivity and difficulty of the commonly used assessment methods such as photography and lesion counting, discrepancy is usually observed in the severity grading of acne vulgaris patients. In this paper, an automated system is proposed for the assessment of acne vulgaris lesions which consists of three main steps, (1) segmentation of acne vulgaris lesions, (2) identification of lesion types and (3) grading the severity of acne vulgaris patients. Acne vulgaris lesions are segmented from skin in the chromatic components of YIQ color space using a self-organizing map and support vector machine. Moreover, area and texture features are computed for each detected lesion and its type is identified using a rule-based technique. In the last step, the severity of an acne patient is graded according to the modified Global Acne Grading system. The performance of the proposed system is evaluated on a dataset of 500 color images captured under a proper lighting condition. The discriminatory capability of seven color spaces is explored and the effect of luminance components on the segmentation of acne vulgaris lesions is empirically examined. The segmentation and grading results of the proposed system are quantitatively analyzed and compared with several other techniques. The proposed system achieved comparatively better segmentation (sensitivity = 92.20%, specificity= 89.65%) and grading results with kappa value of 0.8193.
Original languageEnglish
Pages (from-to)1705-1713
Number of pages9
JournalJournal of Medical Imaging and Health Informatics
Volume7
Issue number8
DOIs
Publication statusPublished - 1 Dec 2017

Keywords

  • Acne Vulgaris
  • Luminance Component
  • Self-Organizing Map
  • Cluster Analysis
  • Support Vector Machine

Cite this