Machine Learning Based Skin Disease Detection Using Semi-Automatic Image Segmentation

The human skin is a remarkable structure, susceptible to numerous known and unknown diseases. Thus, diagnosing skin conditions is one of the most uncertain and complex areas in medical science, making clinical image analysis for skin diseases a formidable task. Moreover, diagnosing skin conditions manually by medical professionals is not only time-consuming but also subjective. This paper presents an approach for diagnosing and categorizing skin diseases utilizing machine learning classification techniques. This approach is based on a popular Semi-automatic image segmentation algorithm known as Interactive Superpixel Segmentation. It combines manual user input with automated algorithms to partition images into meaningful regions. Morphological operations are used in this system to remove hair in image processing. Median filtering technique is employed for noise removal. Computer-based methods will be employed to analyze, process, and classify image data, taking into account different attributes of the photos under examination. To start with, skin photographs go through filtering to eliminate unwanted noise, followed by processing to enhance the overall quality of the image. Gray-Level Co-occurrence Matrix (GLCM) is implemented to extract the features from the skin disease images. Following feature extraction, the image can be classified using the K-Nearest Neighbors (KNN) algorithm, ultimately generating an evaluative report as the output. This paper summarizes the basic workflow for interactive superpixel segmentation in machine learning. Medical image dataset is taken from different websites to predict the skin diseases. This application is more efficient and producing more accuracy results comparing to others.