Share Email Print
cover

Proceedings Paper

Asymmetry and irregularity border as discrimination factor between melanocytic lesions
Author(s): David Sbrissa; Sebastião Pratavieira; Ana Gabriela Salvio; Cristina Kurachi; Vanderlei Salvadori Bagnato; Luciano Da Fontoura Costa; Gonzalo Travieso
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

Image processing tools have been widely used in systems supporting medical diagnosis. The use of mobile devices for the diagnosis of melanoma can assist doctors and improve their diagnosis of a melanocytic lesion. This study proposes a method of image analysis for melanoma discrimination from other types of melanocytic lesions, such as regular and atypical nevi. The process is based on extracting features related with asymmetry and border irregularity. It were collected 104 images, from medical database of two years. The images were obtained with standard digital cameras without lighting and scale control. Metrics relating to the characteristics of shape, asymmetry and curvature of the contour were extracted from segmented images. Linear Discriminant Analysis was performed for dimensionality reduction and data visualization. Segmentation results showed good efficiency in the process, with approximately 88:5% accuracy. Validation results presents sensibility and specificity 85% and 70% for melanoma detection, respectively.

Paper Details

Date Published: 19 June 2015
PDF: 8 pages
Proc. SPIE 9531, Biophotonics South America, 953122 (19 June 2015); doi: 10.1117/12.2186180
Show Author Affiliations
David Sbrissa, Univ. de São Paulo (Brazil)
Sebastião Pratavieira, Univ. de São Paulo (Brazil)
Ana Gabriela Salvio, Amaral Carvalho Hospital (Brazil)
Cristina Kurachi, Univ. de São Paulo (Brazil)
Vanderlei Salvadori Bagnato, Univ. de São Paulo (Brazil)
Luciano Da Fontoura Costa, Univ. de São Paulo (Brazil)
Gonzalo Travieso, Univ. de São Paulo (Brazil)


Published in SPIE Proceedings Vol. 9531:
Biophotonics South America
Cristina Kurachi; Katarina Svanberg; Bruce J. Tromberg; Vanderlei Salvador Bagnato, Editor(s)

© SPIE. Terms of Use
Back to Top