Skin Lesion Segmentation Using Wavelet Transform on MATLAB

Melanoma, a form of skin cancer, is a growing concern in the medical world. This cancer, originating in the epidermal layer in pigment producing cells called melanocytes, has metastatic tendencies with high possibilities of reaching nerves and bones and causing fatally adverse effects. Melanoma’s visible symptoms are lesions on cutaneous surfaces with characteristic properties which are key determinants for doctors to differentiate between a benign or malignant lesion. Thus, a very significant step to reduce the mortality rate of melanoma is early diagnosis with high accuracy during the primary development period of lesion. Clinical images of such skin anomalies are examined using the non-invasive practice of dermoscopy. Dermoscopic images are obtained through Medical Imaging Procedures; however, their assessment was physical and depended heavily on the dermatologist’s understanding. Currently, the chief method used for examination of a lesion is ABCD criteria which sets standards for four parameters of a lesion viz Asymmetry, Border Irregularity, Colour Pigmentation and Diameter (> 6 mm). Lesions fulfilling ABCD criteria need immediate expert attention. Efforts for simulating ABCD criteria on computerised systems using methods of image processing for diagnostic accuracy and speed have been made in the past. However, major concerns with these modalities include ambiguity within human understanding, resolution limitations, distortion and poor contrast, algorithmic error of assigning same numerical values to dissimilar lesion parameters and limitations of spectral methods by inability of obtaining precise frequency content of the lesion's border. Our work simulates two features of ABCD criteria, independently calculating Asymmetry and Border Irregularity of a lesion, on a MATLAB model utilizing Wavelet Transform for processing of lesion images.