Fast Screening of Tuberculosis Patients Based on Analysis of Plasma by Infrared Spectroscopy Coupled with Machine...

Area of Science:

• Biomedical Spectroscopy
• Analytical Chemistry
• Medical Diagnostics

Background:

• Prompt diagnosis of tuberculosis (TB) is crucial for effective treatment and public health.
• Current TB detection methods face challenges in speed, sensitivity, accuracy, and cost-effectiveness.
• Fourier-transform infrared spectrometry (FTIR) offers potential for novel diagnostic approaches.

Purpose of the Study:

• To investigate the utility of FTIR spectroscopy in differentiating plasma samples from TB patients and healthy individuals.
• To compare the performance of transmission and attenuated total reflectance (ATR) FTIR techniques for TB detection.
• To apply machine learning algorithms for spectral analysis and classification.

Main Methods:

• Plasma samples from 20 individuals (10 TB patients, 10 healthy) were analyzed using transmission and ATR-FTIR.
• Spectra were collected with optimized parameters (256 scans, 8 cm⁻¹ resolution) to maximize signal-to-noise ratio.
• Machine learning classifiers (Logistic Regression, Random Forest, XGBoost) were employed to analyze spectral data.

Main Results:

• Distinct spectral differences in protein and lipid regions were observed between TB patients and healthy groups.
• XGBoost classifier achieved an accuracy of 0.749, precision of 0.703, recall of 0.901, and AUC of 0.82 for absorption spectra (3500-2700 cm⁻¹).
• ATR-FTIR data showed performance parameters of approximately 80% accuracy.

Conclusions:

• Absorption FTIR spectroscopy demonstrates superior performance for TB diagnosis compared to ATR-FTIR.
• FTIR spectral analysis of plasma can complement clinical evidence for rapid and accurate TB diagnosis.
• This spectroscopic method is a valuable tool for identifying TB disease in patients.