Computer Aided Diagnosis (CAD) for Segmentation and Classification of Burnt Human skin

Abstract

Human Skin Burn injuries are viewed as the most genuine general medical issue because of which numerous patients died every year all around the globe. Pakistan is a Low-Income Country (LIC) and death rate due to burn injuries is much greater in such countries. To classify burn depths is an under-researched area in Pakistan and has got the great attention of researchers and practitioners. One of the significant issues coming in the Health centers is that Non-Expert doctors are not ready to recognize the burnt area of skin which isn't obvious by bare eyes and hence can't make on the spot decision for correct first treatment according to burn depths, and this may cause a noteworthy issue later on. The objective of this paper is to identify the depth of burnt human skin and to analyze the burns by classifying among first, second and third-degree burns. In this regard, we used the Otsu method of thresholding for segmentation and then applied the statistical method to obtain the feature vector. The maximum average accuracy obtained by using multiple classifiers is reported round about 74.86%. The obtained results will help nonexpert doctors to make on the spot decision by evaluating between first, second and third-degree burns and correct first treatment. The dataset (Images of Burnt Patients) for segmentation and analysis of burnt human skin have been collected from the burn center of Allied Hospital Faisalabad, Pakistan.