Skin Cancer Classification Using Image Processing and Machine Learning

Abstract

One of the most rapidly spreading cancers among various other types of cancers known to humans is skin cancer. Melanoma is the worst and the most dangerous type of skin cancer that appears usually on the skin surface and then extends deeper into the layers of skin. However, if diagnosed at an early stage; the survival rate of Melanoma patients is 96% with simple and economical treatments. The conventional method of diagnosing Melanoma involves expert dermatologists, equipment, and Biopsies. To avoid the expensive diagnosis, and to assist dermatologists, the field of machine learning has proven to provide state of the art solutions for skin cancer detection at an earlier stage with high accuracy. In this paper, a method for skin lesion classification and segmentation as benign or malignant is proposed using image processing and machine learning. A novel method of contrast stretching of dermoscopic images based on the methods of mean values and standard deviation of pixels is proposed. Then the OTSU thresholding algorithm is applied for image segmentation. After the segmentation, features including Gray level Co-occurrence Matrix (GLCM) features for texture identification, the histogram of oriented gradients (HOG) object, and color identification features are extracted from the segmented images. Principal component analysis (PCA) reduction of HOG features is performed for dimensionality reduction. Synthetic minority oversampling technique (SMOTE) sampling is performed to deal with the class imbalance problem. The feature vector is then standardized and scaled. A novel approach of feature selection based on the wrapper method is proposed before classification. Classifiers including Quadratic Discriminant, SVM (Medium Gaussian), and Random Forest are used for classification. The proposed approach is verified on the publicly accessible dataset of ISIC-ISBI 2016. Maximum accuracy is achieved using the Random Forest classifier. The classification accuracy of the proposed system with the Random Forest classifier on ISIC-ISBI 2016 is 93.89%. The proposed approach of contrast stretching before the segmentation gives satisfactory results of segmentation. Further, the proposed wrapper-based approach of feature selection in combination with the Random Forest classifier gives promising results as compared to other commonly used classifiers.