Quantifications of asymmetries on the spectral bands of malignant melanoma using six sigma threshold as preprocessor

• Published in: Proceedings of third International Conference on Computational Intelligence and Information Technology pp. 80 - 86
• Main Authors: Sankaran, S, Sethumadhavan, G
• Format: Conference Proceeding
• Language: English
• Published: Stevenage, UK IET 2013; The Institution of Engineering & Technology
• Subjects: Biology and medical computing; Biomedical measurement and imaging; Computer vision and image processing techniques; Inspection and quality control; Medical and biomedical uses of fields, radiations, and radioactivity; health physics; Optical, image and video signal processing; Patient care and treatment; Patient diagnostic methods and instrumentation; spectral band asymmetry quantification; malignancy degree; noninvasive techniques; preprocessor; skin; diseased region border detection; RGB spectrum; skin cancer identification; six sigma (quality); image segmentation; Shewhart's control charts; excess melanin secretion; image colour analysis; medical image processing; skin lesion images; six sigma threshold; asymmetric index; color variations; AI; ROI symmetries; heterogeneous color variation; control charts; malignant melanoma; skin cancer diagnosis; cancer
• ISBN: 9781849198592; 1849198594
• DOI: 10.1049/cp.2013.2575

Identifying and diagnosing skin cancer using non-invasive techniques have gained momentum in recent years. Such analysis requires trained medical professionals and hence it has become more of subjective. This paper presents a method to quantify the asymmetries of skin lesion. It is achieved by identifying the variations of the RGB spectrum of skin lesion images using Six Sigma threshold as the preprocessor. These identifications are achieved based on the underlying principles of Shewhart's Control Charts, which focus on the fact that the variability does exist in all repetitive processes. The heterogeneous color variation within the skin is considered as an assignable cause and is due to the secretion of excess melanin. These variations possess greater magnitude as compared to the chance causes due to the color variations found in normal skins. Based on these color variations, image is segmented into different homogeneous regions and the borders of the diseased regions are identified. A novel method to quantify the symmetries of the ROI after extracting the border of the lesion is proposed. It is achieved by counting the number of border pixels falls either sides of all the possible axes passing through the center of mass of ROI and by calculating an Asymmetric Index (AI). Results show that the method produces robust border detection and also ascertains the existence of asymmetries in the images of different degree of malignancy.