Date of Award

31-8-2024

Document Type

Thesis

School

School of Computing

Programme

Ph.D.-Doctoral of Philosophy

First Advisor

Dr.J.Premaladha

Keywords

Medical Image Processing, Artificial Intelligence, Deep Neural Networks, Melanoma Skin cancer, Explainable AI

Abstract

Several cancer types are commonly prevalent, and skin cancer is one among them, becoming even more widespread worldwide in the last few decades. To diagnose skin cancer at an early stage and obtain appropriate therapy to treat it, there is a demand to know more about the disease’s characteristics or severity. Skin cancer is caused mainly by various reasons, including damage of the sun or tanning beds by ultraviolet light exposure.

Failing to treat skin cancer might substantially impair an individual’s quality of life as the victim. They likely to experience physical issues linked with the deformities caused by psychological distress and the spreading of cancer cells to other healthy organs. Image processing and Artificial Intelligence (AI) play a crucial role in the early detection of melanoma skin cancer by analysing skin images for irregularities.

This technology helps in identifying potential cancerous lesions, leading to early diagnosis and potentially lifesaving treatment interventions. The ultimate goal of this research is to aid the melanoma skin cancer diagnosis in its early stage. Using the revolutionary AI algorithms and image processing techniques introduced the novel methods that will assist the clinical diagnosis with fair decisions.

Several artefacts are present, and the low image quality makes it challenging to evaluate the clinical features of malignant melanoma. Additionally, it is challenging to discern between benign lesions and malignant melanomas. This research initially proposed the MELIIGAN (Melanoma Information Improved Generative Adversarial Network) architecture to extract fine features from intermediate skin lesion images for early noninvasive diagnosis to overcome a few clinical diagnosis obstacles.

Undiagnosed skin lesions, also known as intermediate skin lesions with image enhancement. Secondly, proposed a new DDCNN-F (Double Decker Convolutional Neural Network – Feature fusion) framework for melanoma classification that includes a novel ‘F’ Flag feature for early detection. This novel ‘F’ indicator efficiently distinguishes benign skin lesions from malignant ones known as melanoma. Additionally, it deals with artefacts such as occluded hair. The scope of the study is extended to deal with multimodal images, and the wavelet method of extracting eight statistical features and seven distinct entropy features from a segmented image is utilised as an early diagnosis tool for a computer-aided diagnosis system.

Finally, the interpretable findings are used for multiclass skin malignancies to make oppressive choices about the clinical diagnosis of cancer in an early period by metadata ingestion in the proposed YOLOv7-XAI paradigm. The suggested research obtained a 9.1% improvement relative to the current state of the art in accurately classifying worrisome lesions. The proposed research shows a 93.75% accurate detection of melanoma and a 7.34% reduction in mistakes. Accuracy with multimodal image classification has been increased to 93.62%, and the multiclass classification model with explainability gave an accurate output of 96.89% and a precision of 94.60%.

Download

Share

COinS