Machine learning in the loop for tuberculosis diagnosis support

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Public Health

Background:

• Machine learning (ML) is increasingly utilized for diagnostic support in healthcare.
• Tuberculosis (TB) diagnosis remains a challenge, particularly in resource-limited settings.

Purpose of the Study:

• To evaluate the effectiveness of various ML techniques for TB diagnosis within a limited-resource healthcare setting.
• To compare ML model performance against established diagnostic methods.

Main Methods:

• Five ML models (logistic regression, classification trees, random forest, support vector machines, artificial neural networks) were trained and supervised by physicians.
• Models utilized seven key variables collected during patient intake.
• Analysis focused on variable importance and model limitations.

Main Results:

• Artificial neural networks demonstrated superior performance in accuracy, sensitivity, and area under the receiver operating curve.
• ML model performance showed improvement compared to smear microscopy for specific TB detection cases.
• Key variables influencing diagnostic accuracy were identified.

Conclusions:

• ML techniques, particularly artificial neural networks, can serve as a valuable alternative diagnostic tool for TB in resource-limited areas.
• Leveraging available data with ML can enhance diagnostic capabilities where health infrastructure is constrained.
• Further integration of ML into diagnostic workflows is recommended for improved patient outcomes.