Analysis of tuberculosis disease through Raman spectroscopy and machine learning

Area of Science:

• Biomedical Spectroscopy
• Computational Biology
• Medical Diagnostics

Background:

• Tuberculosis (TB) diagnosis relies on invasive methods.
• Early and non-invasive screening methods are crucial for effective TB control.
• Biochemical profiles in blood sera can indicate disease presence.

Purpose of the Study:

• To evaluate the effectiveness of Raman spectroscopy (RS) and machine learning for TB screening.
• To analyze blood sera from TB patients and healthy controls.
• To identify specific biomolecules indicative of active pulmonary tuberculosis.

Main Methods:

• Acquisition of Raman spectra from blood sera using a 785 nm laser system.
• Application of Support Vector Machine (SVM) and Principal Component Analysis (PCA) for spectral analysis.
• Discrimination between healthy and TB patient sera based on spectral intensity variations.

Main Results:

• SVM model with Gaussian radial basis effectively discriminated between healthy and TB patients.
• Differences in concentrations of lactate, β-carotene, and amide-I were identified.
• Achieved diagnostic accuracy of 92% with high precision (95%) and specificity (98%).

Conclusions:

• Raman spectroscopy combined with machine learning offers a promising non-invasive method for early TB screening.
• Analysis of specific biomolecules in blood sera can aid in TB diagnosis.
• This approach facilitates rapid and minimally invasive detection of tuberculosis.