Qure.ai

Qure.ai applies deep learning to interpret radiological images, primarily chest X-rays (CXR), to detect and localize abnormalities associated with tuberculosis (TB).

Key Features:

• Deep Learning Algorithms: Employs deep learning techniques to interpret medical images including CXR, MRI, and CT, and to identify and localize abnormalities such as cavities, nodules, pleural effusions, hilar lymphadenopathy, opacity, atelectasis, consolidation, indicators of TB disease, and fibrosis.
• Automated CXR Interpretation (qXR): qXR provides automated interpretation of chest X-rays and was tested on the TB Portals Program (TBPP) database containing drug-resistant tuberculosis (DR-TB) cases, with predictions closely matching human experts.
• Correlation with Treatment Outcomes: Evaluated CXR descriptors for correlation with treatment success and identified ten descriptors as statistically significant predictors of treatment outcomes.

Scientific Applications:

• Tuberculosis Management: Supports large-scale CXR annotation and uniform descriptor extraction to aid patient monitoring and treatment planning in tuberculosis.
• Drug-Resistant Tuberculosis (DR-TB): Validated on TBPP DR-TB cases to enable analysis of complex radiological patterns in drug-resistant tuberculosis.

Methodology:

Trains deep learning models on extensive datasets and blind-tests qXR against a subset of the TB Portals Program (TBPP) database annotated by radiologists and pulmonologists.