Differentiation of digital tb images using texture analysis and rbf classifier

Area of Science:

• Medical Imaging
• Computational Pathology
• Digital Health

Background:

• Tuberculosis (TB) diagnosis relies on sputum smear microscopy.
• Accurate differentiation of positive and negative TB sputum smears is crucial for effective treatment.
• Automated image analysis can potentially improve diagnostic efficiency and accuracy.

Purpose of the Study:

• To develop and evaluate a statistical method for differentiating positive and negative Tuberculosis (TB) sputum smear images.
• To assess the efficacy of Gray Level Co-occurrence Matrix (GLCM) texture analysis for TB image classification.
• To explore the use of Principal Component Analysis (PCA) and Radial Basis Function (RBF) classifier for enhanced differentiation.

Main Methods:

• Acquisition of 100 sputum smear images under a standard protocol.
• Second-order statistical texture analysis using GLCM to derive 19 features.
• Feature set reduction using Principal Component Analysis (PCA) and classification with Radial Basis Function (RBF).

Main Results:

• GLCM texture analysis successfully differentiated between positive and negative TB sputum smear images.
• PCA effectively reduced the feature set to four key components, enhancing efficiency.
• The RBF classifier achieved high classification accuracy for TB image differentiation.

Conclusions:

• The GLCM-based statistical texture analysis is a promising approach for automated characterization of digital TB sputum smear images.
• This method has the potential to support the development of automated systems for TB diagnosis.
• The combination of GLCM, PCA, and RBF classifiers offers a robust solution for TB image classification.