|
|
 |
| |
|
|
|
Skin Cancer and Benign Lesion Classification Using Machine Learning Algorithms |
|
|
|
PP: 1093-1107 |
|
|
|
doi:10.18576/amis/190510
|
|
|
|
Author(s) |
|
|
|
Saleh Ali Alomari,
|
|
|
|
Abstract |
|
|
| In recent years, skin cancer proved to be one of the fatal types of cancer, which include basal carcinoma, squamous cell carcinoma, and melanoma. Early detection of skin cancer is highly important for efficient cure of this disease. Interest in automatic techniques for image analysis has been growing among the medical and computer research communities in an effort to provide early, reliable information on nature and types of lesions. In this respect, computer vision can play vital role in diagnosis of diseases using non- invasive methods such as medical imaging and images. This paper proposes a new system for detection of skin cancer and benign lesions in medical images using machine learning algorithms. In this system, detection progresses through four main steps: establishment of database of dermoscopy images, image pre-processing, image segmentation using thresholding, and feature filtering and statistical feature extraction via the Gray Level Co-occurrence Matrix (GLCM) and Asymmetry, Border, Color, and Diameter (ABCD) analysis. The proposed system was applied on dataset of 5,500 dermoscopy images and the features extracted by this system were classified into cancer lesions and normal lesions by using three classification methods: the k-nearest neighbor (KNN) algorithm, support vector machine (SVM), and nave bias (NB) classifier. The experimental results uncovered that the highest overall accuracy of skin lesion detection and diagnosis (94.35%) was produced by the SVM classifier. |
|
|
|
|
|
|
|
