Early detection of melanoma images using gray level co-occurrence matrix features and machine learning techniques for effective clinical diagnosis

摘要

Melanoma is an early stage of skin cancer. The objective of the proposed work is to detect the symptoms of melanoma early through images of the moles obtained from image processing device and classify the types. The procedure involves converting raw melanoma skin image initially into hue, saturation, and intensity for digital processing. The required information for detecting melanoma is available in the intensity part of the color image. The intensity of the image is down sampled to decrease the bit depth. If the illumination of the down sampled image is not uniform, then gamma correction is applied to get the uniform illumination. A K-means clustering is applied on gamma corrected image which segments the melanoma part from the skin. Textural features are extracted from the segmented image using gray level co-occurrence matrix. Machine learning technique is applied to classify the melanoma images into types like lentigo, acral, nodular, and superficial. Melanoma is detected in this process with an accuracy of 90%.