SKIN CANCER DETECTION USING ARTIFICIAL INTELLIGENCE

Abstract

Malignant melanoma is deadliest form of skin cancer but can be easily treated if detected in early stages. Due to increasing incidence of melanoma, researches in field of autonomous melanoma detection are accelerated. Malignant melanoma is the most severe kind of skin cancer. It can grow anywhere on the body. Its exact cause is still unclear but typically it’s caused by ultraviolet exposure from sun or tanning beds. Its detection plays a very significant role because if detected early then it’s curable, before the spread has begun. It can be 95% recovered if it is early diagnosed. Melanoma cases are rapidly increasing in Australia, New Zealand and Europe. Australia took highest place in the world with this deadly disease. Early diagnose of melanoma totally depends upon the accuracy and talent of practitioners. So automatic detection of melanoma is highly in demand as computer aided diagnosis methods give great accuracy and they are non-invasive methods for the detection of melanoma. This thesis investigates different methods for melanoma classification. In long run it will offer a source to test new and existing methodologies for skin cancer detection. The main objective of this thesis is to present detailed investigation for CAD in melanoma detection. Further thesis objective is to improve and build up relevant segmentation, feature extraction, feature selection and classification techniques that can cope up with the complexity of dermoscopic, clinical or histopathological images. Several algorithms were developed during the path of thesis. These algorithms have been used in skin cancer detection but they can be also used in other machine learning applications. The most significant assistance of this thesis can be summarized as below:  Developing novel feature extraction technique and optimization of parameters. The proposed work has two stages. In first stage, a new method for color and texture features in one features are extracted with the help of CLCM. This method is known as 3D CTF extraction. This method is applied on 200 images with improved results for skin cancer detection. Second stage is applied with 3D CTF with PCA. This technique is used for dimensionality reduction to improve accuracy of the classifiers