Tuberculosis diagnosis support analysis for precarious health information systems

Area of Science:

• Computational intelligence
• Medical diagnostics
• Public health

Background:

• Pulmonary tuberculosis is a global health emergency.
• Developing countries like Colombia need cost-effective diagnostic tools.
• Artificial neural networks (ANNs) can aid tuberculosis diagnosis.

Purpose of the Study:

• To explore the use of ANNs for pulmonary tuberculosis diagnosis.
• To develop a tool supporting medical decision-making in resource-limited settings.
• To assess the feasibility of unsupervised learning for risk stratification.

Main Methods:

• Utilized a database of 105 subjects suspected of pulmonary tuberculosis.
• Extracted data included demographics, comorbidities, and clinical information.
• Applied supervised and unsupervised artificial neural network models.

Main Results:

• ANN models achieved 97% sensitivity and 71% specificity.
• Results are comparable to traditional tuberculosis detection methods.
• Demonstrated advantages in speed and low implementation costs.

Conclusions:

• ANNs provide valuable decision-making support for physicians.
• The approach is particularly beneficial in areas with limited infrastructure and data.
• This AI-driven method enhances tuberculosis diagnosis capabilities.