Article Preview
Top1. Introduction
Skin forms the first layer of defense mechanism against diseases and infections, primarily at prima-face protection from UV radiation is ensured by skin. If proper care is not taken after some year the skin itself can turn into a source of a diseases. Most dangerous kind of skin disease is skin cancer and melanoma is one of them which accounts for almost 4% of all skin cancer. If melanoma is not identified at early stages, it can grow deeper into the skin and may spread across other parts of the body which is dangerous as it is very difficult to cure (Bafounta, Beauchet, Aegerter, & Saiag, 2001). Figure 1 shows images of skin lesions.
The diagnosis accuracy of melanoma stays between 75% to 84% even after being examined by the expert dermatologists. For increasing speed and diagnosis accuracy the computer aided diagnostics is required (Argenziano et al., 2003; Garnavi, Aldeen, & Bailey, 2012). Some information can be extracted using computers, like asymmetry, texture features, color variation, which can not be readily perceived using human eyes. Currently most used methods for classification of skin lesion rely on hand-crafted features like ABCDE rule (Nachbar, Stolz, Merkle, & Cognetta, 1994), Menzies method (Menzies, 2001), 3-point checklist (Soyer et al., 2004), CASH (Henning et al., 2007) and 7-point checklist (Argenziano et al., 2011). Personal experience is what the physicians depends on evaluating each patient’s lesions over case-by-case basis by captivating into patterns of patient’s local lesion in contrast to that of the whole body (Gachon et al., 2005). Without any type of computer-based assistance, the accuracy of clinical diagnosis for melanoma lesion detection is reported to be arround 65% to 80% (Argenziano & Soyer, 2001). Use of dermoscopic images, pictures taken skin by a skin surface microscopy (Kittler, Pehamberger, Wolff, & Binder, 2002), can increase diagnostic accuracy of the skin lesions by arround 49% (Kawahara, BenTaieb, & Hamarneh, 2016). In Figure 1, it can be visually seen the differences between melanoma and benign skin lesions, making it very difficult to distinguish between the two cases, even for trained medical experts. For the reasons described above, the physician should be provided with an intelligent medical imaging-based skin lesion diagnosis system that can assist a physician in classifying skin lesions. The Menzies, ABCD Rule and Seven Point Checklist are one of the few most referred algorithms which may improve the Melanoma diagnostics in Skin Cancer (Kawahara, BenTaieb, & Hamarneh, 2016). A computer image based melanoma diagnosis follows a sequence which contains the following steps: 1. Scanning the image of skin lesion, 2. Skin Lesion segmentation, 3. Extraction of Lesion’s features, 4. Segmentation of Skin Lesion and 5. Feature classification (Celebi, Iyatomi, Schaefer, & Stoeck, 2009).
This research article is divided into seven sections; Section II gives an explaination of different types of skin lesion. Section III presents the different melanoma diagnosis techniques. Section IV describes the work that has been done related to skin lesion image recognition and detection. In section V, a generalized model for improved detection and prediction of skin lesion is presented. In Section VI, The mobile implementation of the dermoscopy image analysis is presented which helps to detect the skin lesion. In Section VII, we have shown the results and in Section VIII conclusion of paper is discussed.