Efficient and Parallel Medical Image Segmentation Model (EPSM) Based on Brink-MCET Using Heterogeneous Distributions

Abstract

Medical image segmentation is becoming increasingly popular in the field of image analysis. In computer vision, medical image segmentation is a challenging task but is crucial for identifying and analyzing diseased areas in the body. It is especially crucial for detecting conditions such as brain tumors, skin cancer, and other serious illnesses. This paper presents a novel thresholding technique based on minimum cross-entropy thresholding (MCET), specifically designed for precise segmentation of dermoscopy images. The suggested bimodal technique was evaluated using three benchmark datasets from PH2, HAM10000, and ISIC 2017. To determine the MCET of each input image, three different combinations of statistical distributions— Gaussian, Gamma, and Lognormal—were employed. To further raise the effectiveness of the effective and parallel segmentation model (EPSM) model, a novel parallel boosting segmentation technique was created and applied. By comparing the proposed image segmentation method’s output with that of the Entropy-Li approach, its effectiveness was assessed. Both supervised and unsupervised evaluation methods were used. Based on the obtained outcomes, it can be inferred that the EPSM segmentation model is a reliable, accurate, and consistent method with outstanding performance characteristics.