No Detail Found!

Optimal solution for segmentation of malignant melanoma dermoscopic images

Author(s)
Tahir Abbas Department of Computer Science, TIMES Institute, Multan, 60000, Pakistan.
Muhammad Kashan Basit Department of Computer Science, MNS-UET, Multan, 60000, Pakistan.
Jamshaid Iqbal Janjua Al-Khawarizmi Institute of Computer Science, University of Engineering & Technology (UET), Lahore, Pakistan, School of Computer Sciences, National College of Business Administration & Economics, Lahore, Pakistan.
Bushra Tanveer Naqvi University of Azad Jammu and Kashmir, Muzaffarabad, Azad Jammu and Kashmir.
Muhammad Irfan School of Computer Sciences, National College of Business Administration & Economics, Multan, Pakistan.
Abstract
Melanoma Malignant (MM) is the most common and dangerous form of skin cancer, which is analyzed by using Dermoscopic images in computer sciences. Segmentation technique is used to separate lesion part from healthy part in Dermoscopic images. In this research, comparison of different most popular segmented Dermoscopic image technique like Type-2 Fuzzy, Hybrid Threshold, Wavelet, Gradient Vector Flow (GVF), and Watershed etc. is approached and then better segmentation technique is proposed. In these segmentation techniques different issues like problem of hair, different color lesion, specular reflection and smoothing transaction between lesion and skin were not taken under consideration. Our methodology involves three levels of hierarchy. In the preprocessing step, it deals with problem of hair, bubble noise, smoothing and reflection noise in Dermoscopic images. These noise removals are achieved by using different filters like “Derivative of Gaussian filter and Bootomhat filter”. After region of interest is extracted then combination of threshold, image enhancement and morphological filter are used to produce the efficient algorithm for segmentation. At the end step, segmented crop image is compared with dice coefficient and experimental results of gross error rate are evaluated. For this purpose, PH² Dataset is used that contains 200 Dermoscopic images with the lesion images. The lesion images are extracted by the expert dermatologists.
Publication Details
Page(s) 39-49
DOI DOI not available
Published Journal: Journal of Statistics, Computing and Interdisciplinary Research, Volume: 5, Issue: 2, Year: 2023

Request the full text from PASTIC

Keywords
Image Processing, segmentation, Malignant melanoma., Dermoscopic images
0

Citations

0

Downloads

31

Views