On Performance Analysis of a Neutrosophic Bipolar Fuzzy and Texture-Based Model for COVID-19 Detection from Chest X-Rays

Sanjayprabu S; Senthil Kumar L; Stephan Antony Raj A; Santhakumari R; Mareeswari R; Ajithkumar R

Authors

Sanjayprabu S KPR College of Arts Science and Research, Coimbatore, India;
Senthil Kumar L Dr. Mahalingam College of Engineering and Technology, Pollachi, Coimbatore, India;
Stephan Antony Raj A SNS College of Technology, Coimbatore, India;
Santhakumari R Sri Ramakrishna College of Arts and Science, Coimbatore, India;
Mareeswari R SNS College of Technology, Coimbatore, India;
Ajithkumar R Government Arts and Science College, Thondamuthur, Coimbatore, India;

Keywords:

: COVID-19 detection; Chest X-ray; NeutrosophicBipolar Fuzzy Set; Texture Analysis; Gray Level Co-occurrence Matrix (GLCM); Machine Learning; k-nearestNeighbors (k-NN); Logistic Regression; Decision Tree; Medical Image Processing

Abstract

This study proposes a novel hybrid framework for rapid and accurate COVID-19 detection from
chest X-ray images by integrating Neutrosophic Bipolar Fuzzy Sets (NBFS) with texture-based feature analysis. The methodology comprises four main stages: image acquisition, preprocessing, feature extraction, and
classification. In the preprocessing phase, NBFS enhances image contrast by decomposing each image into
positive, negative, and indeterminate components using intensity thresholds. The neutrosophic domain assigns
truth, falsity, and indeterminacy values, enabling robust handling of noise and uncertainty in medical imaging.
Texture features are extracted using the Gray-Level Co-occurrence Matrix (GLCM), capturing spatial intensity relationships critical for distinguishing infected regions. These features are classified using Decision Tree,
Logistic Regression, and k-Nearest Neighbour (k-NN) models. Experimental evaluations on a benchmark chest
X-ray dataset demonstrate that the Decision Tree classifier outperformed other models, achieving an accuracy
of 95.92%, sensitivity of 96.55%, and specificity of 95.30%. The proposed approach offers a fast, interpretable,
and cost-effective diagnostic aid, potentially supporting large-scale screening in clinical settings. Future research
will explore multi-class classification and application to diverse medical imaging modalities

DOI: 10.5281/zenodo.17041997