Screening for active pulmonary tuberculosis: Development and applicability of artificial neural network models

Area of Science:

• Medical Informatics
• Public Health
• Artificial Intelligence in Medicine

Background:

• Tuberculosis (TB) presents a significant global health challenge due to diverse epidemiological settings.
• Effective TB control necessitates cost-effective screening and advanced diagnostic tools.
• Secondary clinics require efficient methods for pulmonary TB (PTB) screening.

Purpose of the Study:

• To introduce a novel decision support tool (DST) for screening pulmonary TB (PTB) patients.
• To enhance the accuracy and efficiency of PTB diagnosis in secondary healthcare settings.
• To provide a low-cost, rapid pre-testing solution for presumptive PTB patients.

Main Methods:

• Development of a DST integrating an adaptive resonance model (iART) for risk stratification (low, medium, high).
• Implementation of a multilayer perceptron (MLP) neural network for active vs. inactive PTB classification.
• Evaluation of the DST's performance using sensitivity, specificity, and predictive values.

Main Results:

• The DST achieved an overall sensitivity of 92% and specificity of 58%.
• High sensitivity was observed for smear-positive (96%) and smear-negative (82%) PTB cases.
• Negative predictive values exceeded 95% even at 20% prevalence, with performance above 83% in low and high-risk groups.

Conclusions:

• The proposed DST offers a quick, low-cost pre-screening method for PTB.
• It effectively guides confirmatory testing and patient management, particularly in low and middle-income countries.
• The tool demonstrates significant potential for improving TB control strategies in resource-limited environments.