Feature Extraction from Dermoscopy Images for an Effective Diagnosis of Melanoma Skin Cancer

The aim of this paper is to extract some distinct geometric features from dermoscopy images to classify benign and malignant melanomas. To avoid skin biopsy which is an invasive technique, diagnosis of melanoma skin cancer from dermoscopy images was developed. It is a very challenging task due to some reasons. Firstly, high degree of intraclass variation exists among melanoma images while low interclass variation is found between melanoma and non-melanoma images. Secondly, benign and malignant melanoma images are visually similar to a great extent. And finally noises like hair are always present in skin images which make difficult to analyse the images. In this work, we used fundamental ABCD rule to detect malignant melanoma and benign lesion based on quantitative measures. In our proposed technique, we extracted a new feature which is the difference between maximum and minimum Feret diameters of the best fit ellipse to skin lesion. This discriminative feature alone classified the melanomas with 86.5% accuracy. In our approach, we applied the feature extraction block containing all parameters to 200 images and the overall accuracy of 98% was achieved to detect malignant and benign melanoma from the images. A Back-propagation Neural Network (BNN) model was developed and eventually used as a classifier in this proposed method.

I. Introduction

Mainly three types of skin cancers, Basal-Cell Carcinoma (BCC), Squamous Cell Carcinoma (SCC) and melanoma are found frequently. Among them melanoma is the deadliest form of skin cancer although it is not the most common. It is responsible only for 4% of all cancers occurred in human skin while it is accountable for 75% of deaths caused by skin cancers [1]. This cancer is primarily caused due to severe exposure of skin to the sunlight. It is prime need to detect melanoma in the early stage because if it is treated in the initial state, it is almost curable. But if the cancer is not detected and treated from its early stage, it spreads adjacent parts of the body rigorously when it becomes tenacious to treat and ultimately causes death. The American Cancer Society estimates that in 2018, about 91,270 new melanomas are suspected to be diagnosed (about 55,150 in male and 36,120 in female) in the United States. Almost 9,320 people are presumed to die of melanoma (about 5,990 male and 3,330 female) [2]. Australia and New Zealand have the highest rates of melanoma in the world [3]. The American Cancer Society [2] provides the ABCD rule which is fundamental method for the dermatologists and steerage of self-examination for the patient for Malignant Melanoma. “A” stands for Asymmetry: one half of the lesion does not match another. “B” is for Border: the verges are irregular, uneven, blur, ragged or notched. “C” represents Color: the color is not uniform all over the lesion. “D” is for Diameter: the lesion is larger than 6 millimeters, although melanoma lesions may sometimes be smaller than this. According to Patwardhan et al. [4], this ABCD rule diagnoses melanoma only for thin melanocytic lesions and it has 59-88% accuracy in detection of malignant melanoma, but for precise diagnosis, biopsy is essential. About 75-84% diagnostic accuracy was determined even when the expert dermatologists use the dermoscopy images [5]. There are mainly four proposed rules and methods such as seven-point checklist method [6], ABCD rule [2], the Menzies method [7] and pattern analysis [8] to determine the melanoma skin cancer. In 2015, Marin et al. [9] noticed that the ABCD rule became inefficient while detecting micro-melanomas. This evaluation determines a specificity of 91% and sensitivity of 43%. Amarathunga et al. [10] proposed a system which enabled patient to detect skin diseases via online and the system provided therapeutic advice also. They employed some data mining classification algorithms such as AdaBoost, BayesNet, J48, MLP and Naive Bayes to classify three types of skin diseases and achieved 80-85% accuracy for melanoma detection. Jain et al. in 2015 used ABCD features [11]. Krishna et al. [12] proposed several clustering techniques for segmentation process and extracted features by using ABCD. In this paper, we developed five effective features to classify malignant and benign melanoma. These features are Asymmetry score (AS), Border Irregularity (B), Color variegation (C), Diameter (D1) and Difference between maximum and minimum Feret diameters of the best fit ellipse to the lesion (D2). In our proposed technique, we achieved accuracy of 98% with 95% sensitivity and 98.8% specificity.