X-ray image based on Gray Level Cooccurrence Matrices (GLCM) k-nearest neighbor (KNN) to detect tuberculosis

• Published in: AIP conference proceedings Vol. 2536; no. 1
• Main Authors: Suhariningsih, Bastomi, Mohammad Yazid, Purwanti, Endah, Hariyani, Dita Aprilia, Permatasari, Perwira Annissa Dyah, Astuti, Suryani Dyah
• Format: Journal Article Conference Proceeding
• Language: English
• Published: Melville American Institute of Physics 19.05.2023
• Subjects: Accuracy; Correlation; Entropy; Homogeneity; Infectious diseases; K-nearest neighbors algorithm; Medical imaging; Sneezing; Thorax; Tuberculosis
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0118948

Tuberculosis is an infectious disease caused by a bacterium called bacillus mycobacterium tuberculosis. Tuberculosis is spread through coughing and sneezing which affects the lungs of people infected with pulmonary tuberculosis. One of the methods is using the thorax image. However, accuracy without a standard is the problem in this topic. It's caused by the analysis result depend on the ability of the medical experts only. In this study, a Tuberculosis detection program was designed using the k-nearest neighbor classification method and Gray Level Cooccurrence Matrices (GLCM) features as classification input. So that the detection program was expected to be a tool for medical experts who had standardized accuracy. The GLCM features were to input the k-nearest neighbor (kNN) classification which are contrast, correlation, energy, entropy, and homogeneity. The program output was divided into 2 classes namely abnormal (tuberculosis) and normal. The combination of entropy-correlation and entropy-energy-correlation features by an optimal level of accuracy, sensitivity, and specificity showed a value of k=1 that is 92%, 92%, 92%.