Diagnostic Performance Evaluation

Uses

How it works

Highlighted claims

• The primary performance metric for CAD evaluation in CAPTURE is AUC, computed using the R package pROC with 95% confidence intervals from 2,000 stratified bootstrap replicates. — Catalysing Artificial Intelligence for Paediatric Tuberculosis Research (CAPTURE): protocol for a global multicentre study establishing a paediatric chest X-ray repository to evaluate computer-aided detection algorithms
• Sensitivity and specificity are calculated at two thresholds: the threshold achieving 90% sensitivity, and the manufacturer-recommended threshold. — Catalysing Artificial Intelligence for Paediatric Tuberculosis Research (CAPTURE): protocol for a global multicentre study establishing a paediatric chest X-ray repository to evaluate computer-aided detection algorithms
• The primary reference standard is a composite clinical standard distinguishing confirmed and unconfirmed TB from unlikely TB using NIH 2015 case definitions. — Catalysing Artificial Intelligence for Paediatric Tuberculosis Research (CAPTURE): protocol for a global multicentre study establishing a paediatric chest X-ray repository to evaluate computer-aided detection algorithms
• Results will be stratified by age group, sex, HIV status, nutritional status, WHO region, CXR image format, and presence of specific radiological features. — Catalysing Artificial Intelligence for Paediatric Tuberculosis Research (CAPTURE): protocol for a global multicentre study establishing a paediatric chest X-ray repository to evaluate computer-aided detection algorithms