Work

Visual feature extraction from dermoscopic colour images for classification of melanocytic skin lesions

Year: 2020
Type: article
Abstract: The early diagnosis of Melanoma is a challenging task for dermatologists, because of the characteristic similarities of Melanoma with other skin lesions such as typical moles and dysplastic nevi. Aims: This work aims to help both experienced and non-experienced dermatologists in the early detection of cutaneous Melanoma through the development of a computational helping tool based on the “ABCD” rule of dermoscopy. Moreover, it aims to decrease the need for invasive biopsy procedure for each tested abnormal skin lesion. Methods: This is accomplished through the utilization of MATLAB programming language to build a feature extraction tool for the assessment of lesion asymmetry, borders irregularity, and colors variation in the tested lesion. Results: The helping tool obtained a sensitivity of 81.48%, a specificity of 52.83% and accuracy of 62.50% in the assessment of the Asymmetry Index. A new metric for the borders irregularity index was built. Finally, for the Colors Variation Index algorithm a sensitivity of 51.37%, a specificity of 61.51% and accuracy of 57.80% was achieved. Conclusions: This work created a computational tool based on the ABCD-rule, which is helpful for both experienced and non-experienced dermatologists in the early discrimination of Melanoma than other types of skin lesions and to eliminate the need of the biopsy procedure. A new metric for the Borders Irregularity Index was established depending on the number of inflection points in the lesion’s borders.
Language: English
Subfield: Oncology
Field: Medicine
Open Access status: green
OpenAlex ID: W3044065019
FWCI: 0.271
Cites: 14
Cited by: 5
Related to: 10