DEVELOPING A MELANOMA SKIN CANCER STAGE IDENTIFICATION MODEL USING IMAGE PROCESSING AND MACHINE LEARNING APPROACHES

YEMER, KELEBIE

dc.contributor.author	YEMER, KELEBIE
dc.date.accessioned	2021-08-13T07:31:28Z
dc.date.available	2021-08-13T07:31:28Z
dc.date.issued	2021-03
dc.identifier.uri	http://ir.bdu.edu.et/handle/123456789/12409
dc.description.abstract	Skin cancer is one of the most common cancers in the world. From skin cells melanocyte cells develop melanoma cancer and this cancer is dangerous and more spread out quickly to other parts of the skin as well as organ systems of the body unless treated early. In melanoma skin cancer there are three main stages: early melanoma; high risk melanoma and advanced melanoma and they have their own characteristics. As per the related works there is not conducted a research on stage identification of melanoma skin cancer. Besides, it is a challenging task but it is crucial to diagnose and treat the patient in its identical treatment mechanism. For this skin cancer and its stage have been mostly diagnosed by the human eye and it has its own drawbacks like no fast determination of stage, less accuracy of correct result and different noises in the skin. For those cases we have developed a computer vision technique that can identify the stage of melanoma skin cancer. In this we have followed design science methods followed by data collection, image preprocessing, feature extraction and classification. Firstly, the data is collected from ALERT Hospital, BLACK LION Hospital and machine repository sites that are PH2, HAM1000. We have collected a total of 1176 with 392 each class. image preprocessing tasks like histogram equalization, adaptive median filtering is done. Secondly, we have used the combination of the automatic feature extraction technique called convolutional neural network and local feature descriptor called SURF are used. Finally, we have applied flipping and rotation augmentation techniques used to produce a total of 3000 images with 1000 for each using convolutional neural network and SVM classifier. We have got a test accuracy of 84.6% in SoftMax classifier using CNN features,78.8% test accuracy in SVM using SURF feature,91.1% test accuracy in SVM using CNN feature and 94.3% accuracy in SVM using hybrid features of SURF & CNN.	en_US
dc.language.iso	en_US	en_US
dc.subject	computer science	en_US
dc.title	DEVELOPING A MELANOMA SKIN CANCER STAGE IDENTIFICATION MODEL USING IMAGE PROCESSING AND MACHINE LEARNING APPROACHES	en_US
dc.type	Thesis	en_US