Automatic Detection of Granuloma Necrosis in Pulmonary Tuberculosis Using a Two-Phase Algorithm: 2D-TB

Area of Science:

• Pathology
• Computational Biology
• Infectious Diseases

Background:

• Granuloma necrosis is a key visual indicator of pulmonary tuberculosis (TB) in susceptible hosts.
• No automated algorithms currently exist for detecting granuloma necrosis in TB, limiting quantitative analysis.
• Developing such a tool could standardize analysis and enable machine learning applications.

Purpose of the Study:

• To develop and validate an automated algorithm for detecting granuloma necrosis in pulmonary TB.
• To assess the algorithm's performance against expert pathologist evaluation.
• To transform visual TB pathology data into quantitative metrics.

Main Methods:

• Histopathological images from super-susceptible Diversity Outbred (DO) mice infected with Mycobacterium tuberculosis were used.
• A two-phase algorithm, 2D-TB, was trained and validated on lung sections.
• Phase 1 detects granulomas; Phase 2 identifies cell-poor necrotic regions within granulomas.

Main Results:

• The 2D-TB algorithm achieved 100.0% sensitivity and 91.8% positive predictive value.
• Agreement between 2D-TB and an expert pathologist was 95.5%.
• A statistically significant positive correlation was found between the area detected by 2D-TB and the pathologist.

Conclusions:

• The 2D-TB algorithm accurately detects granuloma necrosis in pulmonary TB.
• This validated tool offers a reliable method for quantitative analysis of TB pathology.
• 2D-TB has the potential to reduce variability and advance TB research through data-driven insights.