Executive Summary

Misdiagnosis is a common malpractice and it is commonly caused by the lack of experience, tiredness or both. These human-based errors may be greatly reduced with the support of automated medical-assisted systems. In developing automated skin lesion diagnosis (ASLD) for common skin lesions, the main problem faced is the extraction of important features from the imaging devices and selection of significant features for classification. There are many features of skin lesions used by practitioners for diagnosis. For example, features identified with ABCD rule (Asymmetry, Border irregularity, Color variegation, Diameter), Menzies’ Scoring Method , the 7-point checklist and CASH algorithm (Colour, Architecture, Symmetry, Homogeneity). However, not all of the features are used. For example in, out of 48 features generated from DBDermo-Mips system, only 13 were selected for classification. From the pre-processing images, segmentation is often the key step in interpreting the image. Image segmentation is a process in which regions or features sharing similar characteristics are identified and grouped together. Segmentation is often the critical step in image analysis and it can be classified as supervised and unsupervised segmentation algorithm. Most presentations of segmentation algorithms contain superficial evaluations which merely display images of the segmentation results and appeal to the reader’s intuition for evaluation. There is a consistent lack of numerical results, thus it is difficult to know which segmentation algorithms present useful results and in which situations they do so. If segmentation is done well, then, all other stages in image analysis are made simpler. If significant features are not extracted from images, it will affect the accuracy of the automated diagnosis. Therefore, selecting the most suitable features is highly important so that as much relevant features can be identified and extract.