Please use this identifier to cite or link to this item:
Appears in Collections:Computing Science and Mathematics eTheses
Title: Towards the early detection of melanoma by automating the measurement of asymmetry, border irregularity, color variegation, and diameter in dermoscopy images
Author(s): Ali, Abder-Rahman
Supervisor(s): Li, Jingpeng
Shankland, Carron
Keywords: Machine learning
Deep learning
Skin lesion
Image processing
Issue Date: 11-Sep-2020
Publisher: University of Stirling
Abstract: The incidence of melanoma, the most aggressive form of skin cancer, has increased more than many other cancers in recent years. The aim of this thesis is to develop objective measures and automated methods to evaluate the ABCD (Asymmetry, Border irregularity, Color variegation, and Diameter) rule features in dermoscopy images, a popular method that provides a simple means for appraisal of pigmented lesions that might require further investigation by a specialist. However, research gaps in evaluating those features have been encountered in literature. To extract skin lesions, two segmentation approaches that are robust to inherent dermoscopic image problems have been proposed, and showed to outperform other approaches used in literature. Measures for finding asymmetry and border irregularity have been developed. The asymmetry measure describes invariant features, provides a compactness representation of the image, and captures discriminative properties of skin lesions. The border irregularity measure, which is preceded by a border detection step carried out by a novel edge detection algorithm that represents the image in terms of fuzzy concepts, is rotation invariant, characterizes the complexity of the shape associated with the border, and robust to noise. To automate the measures, classification methods that are based on ensemble learning and which take the ambiguity of data into consideration have been proposed. Color variegation was evaluated by determining the suspicious colors of melanoma from a generated color palette for the image, and the diameter of the skin lesion was measured using a shape descriptor that was eventually represented in millimeters. The work developed in the thesis reflects the automatic dermoscopic image analysis standard pipeline, and a computer-aided diagnosis system (CAD) for the automatic detection and objective evaluation of the ABCD rule features. It can be used as an objective bedside tool serving as a diagnostic adjunct in the clinical assessment of skin lesions.
Type: Thesis or Dissertation

Files in This Item:
File Description SizeFormat 
PhD-Thesis.pdfPhD Thesis83.03 MBAdobe PDFView/Open

This item is protected by original copyright

Items in the Repository are protected by copyright, with all rights reserved, unless otherwise indicated.

The metadata of the records in the Repository are available under the CC0 public domain dedication: No Rights Reserved

If you believe that any material held in STORRE infringes copyright, please contact providing details and we will remove the Work from public display in STORRE and investigate your claim.