Title: Melanoma skin cancer identification from dermatoscopy images by machine learning using Thepade SBTC and triangle thresholding

Authors: Sudeep D. Thepade; Deepa Abin; Aasim Sayyad; Zahid Akthar; Rik Das

Addresses: Computer Engineering Department, Pimpri Chinchwad College of Engineering, Savitribai Phule Pune University, Pune, India; Pimpri Chinchwad University, Pune, India ' Vishwakarma Institute of Technology, Pune, India ' Computer Engineering Department, Pimpri Chinchwad College of Engineering, Savitribai Phule Pune University, Pune, India ' Department of Network and Computer Security, State University of New York Polytechnic Institute, USA ' A.K. Chaudhari School of Information Technology, Kolkata, India

Abstract: Melanoma skin cancer is prevalent in all types of cancer and can be completely treated if detected in the initial stages. Even for medical practitioners, discrimination of melanoma skin lesions from other skin scars resulting from sunburn or rashes is difficult. Machine learning (ML) can help in skin cancer detection (SCD) from dermatoscopy images. The study proposes the fusion of global and local features computed with Thepade_Sorted_Block_ Truncation_Coding (TSBTC) and the triangle thresholding method. Nine variations of the TSBTC are explored for feature formation (TSBTC 2-ary to TSBTC 10-ary). Each of the variations of the features is provided to eight classifiers like BayesNet, RandomForest, J48, NBTree, RandomTree, REPTree, IBK, and Kstar, with three ensembles. The classification is performed with dataset HAM10000. Compared to individual feature considerations, better SCD performance is observed with feature fusion of TSBTC and triangle thresholding. Ensemble 'RandomForest+IBK+NBTree' gives good accuracy for melanoma SCD.

Keywords: melanoma identification; triangle thresholding; ML classifiers; Thepade SBTC; global and local features.

DOI: 10.1504/IJCVR.2026.155202

International Journal of Computational Vision and Robotics, 2026 Vol.17 No.1, pp.60 - 77

Received: 30 Mar 2023
Accepted: 15 Nov 2023
Published online: 29 Jul 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article