Identification of Skin Tumours using Statistical and Histogram Based Features LC11-LC15
Assistant Professor, Department of Biomedical Engineering, Vels Institute of Science, Technology and Advanced studies (VISTAS),
Pallavaram, Chennai -600117, Tamil Nadu, India.
Introduction: Skin tumour is uncontrolled growth of cells in skin. Skin tumour is becoming predominant in different parts of the world. Basal carcinoma, squamous carcinoma and melanoma are the skin cancer types common in India. The rate of survival depends on the cancer stages, if diagnosed early it can be treated completely. Statistical and histogram features can be defined as part of image processing algorithm used to identify the type of skin tumours based on the probabilistic occurrence and intensity of pixel values respectively.
Aim: The aim was to illustrate easy identification process of skin tumours from dermal images using statistical and histogram features.
Materials and Methods:Dermal images were obtained from the PH2 database for identification of two different types of skin tumours such as melanocytic nevi and malignant melanoma. Colour Histogram was used to differentiate the two categories. Pre-processing and segmentation was performed for extraction of statistical and histogram based features from the lesion. From the extracted features, mean and standard deviation values were calculated for proper identification of skin tumours. Further to improve the accuracy of the identification, neural network classifiers were used which defines more enhanced efficiency in detection of skin tumours.
Results: Colour histogram was used to differentiate the two categories of skin tumours. Malignant melanoma possesses high peaks of channel pixels at both extremities of the histogram. Histogram and statistical based features derived from the lesion describes that malignant melanoma has higher values of mean and standard deviation of features derived from segmented lesions. Neural network classifiers were used for further accuracy of identification which distinguishes the two different categories of skin tumours.
Conclusion: Colour histogram, statistical and histogram based features were derived for differentiation and identification of two categories of skin tumours. Thus, a simple and effective technique for description of skin tumours was determined.