Unique Color Circle Design for A Novel Screening Tool to Identify Cancerous Skin Lesions

Diagnosis of skin cancer is mostly based on clinical examination and biopsy. Despite advances in diagnostic tools, both invasive and non-invasive, more than 2 people die of the disease every hour and more than 9,500 people are diagnosed with skin cancer every day in the US. This emphasizes the need to have a screening tool which can differentiate between non-cancerous (NC) and cancerous (C) skin lesions, quickly and effectively. As most of the skin lesions are apparent to naked eyes and have different color textures, it was hypothesized that segmental analyses of photographs of suspicious lesions should be able to differentiate NC from C lesions on skin. For this study, a new and unique color circle design (CCD) for establishing a relation between wavelength (λ) and hue of the HSV (hue Saturation and Value) model is proposed. Using MATLAB, enlarged dermatoscopic images of 23 authentic skin lesions (test samples) with known diagnoses were segmented in such a manner that the region of interest (ROI) included all the suspicious areas. Thereafter, the pixels data from ROI locations were standardized and transformed from RGB to HSV space. Again using MATLAB, hue (h) and Value (V) data were extracted from HSV data. Since each h represents a unique wavelength in the visible range of the spectrum, the CCD was used to identify the cancerous lesions aided by the V parameter. Using trial and error on several other skin cancer lesions (not included in the test samples), two thresholds and a set of criteria were selected to discriminate between C and NC. Results show that use of CCD has great potential as a screening tool for skin cancer detection, achieving over 90 % accuracy on the test samples.