A Novel Medical Image Segmentation Using Neutrosophic Sets With Slope Variation

Abstract

Computer-Aided Diagnostic methods demand the precise segmentation of medical
images. The important stage in diagnosis is the extraction of the Region of Interest (ROI).
Illustration of the image in a more meaningful way is the aim of segmentation. Segmentation
finds extensive use in computer vision, object recognition, image recovery based on the
content, etc. In the proposed model, Slope Variation Scatter (SVS) plot of image is obtained
to compute vertices of Neutrosophic Gaussian Function (NGF). The SVS describes global
variation rate of image histogram. In this, the crests represent local mean of pixels/ certainty
mean and valleys represents uncertainty mean. A novel NGF is membership function
designed to convert abdominal Computed Tomography (CT) to Neutrosophic Subsets (NS).
The NS comprises of object, nonobject and edge subsets. The Object Subset (OS) represents
liver or kidney, Nonobject Subset (NOS) represents background of liver/kidney and Edge
Subset (ES) represents edges of liver or kidney. The proposed model is experimented on 106
abdominal CT images to segment the liver and kidney accurately. The experimental
outcomes are compared with Fuzzy C Means algorithm (FCM), show that the anticipated
framework is proficient of segmenting an intended organ automatically and precisely. The
proposed model achieves average accuracy, Relative Volume Difference (RVD) and Dice
Similarity Factor (DSF) for liver are 91.01%, 8.23% and 89.61% respectively. The average
accuracy, RVD and DSF for kidney are 91.11%, 5.96% and 91.45% respectively.

DOI: 10.5281/zenodo.13568338