Tuberculosis Disease Severity Assessment Using Clinical Variables and Radiology Enabled by Artificial Intelligence.
Authors
Affiliations (9)
Affiliations (9)
- Department of Biomedical Informatics, Harvard Medical School, Boston, Massachusetts, USA.
- National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, Maryland, USA.
- Research Data and Communications Technology Corp, NTT Data Company, New York, New York, USA.
- Department of Bioengineering and Therapeutic Sciences, University of California, SanFrancisco, San Francisco, California, USA.
- UCSF Center for Tuberculosis, University of California, SanFrancisco, San Francisco, California, USA.
- Division of HIV, Infectious Diseases, and Global Medicine, University of California, SanFrancisco, San Francisco, California, USA.
- Center for Biostatistics in AIDS Research, Department of Biostatistics, Harvard T.H. Chan School of Public Health, Boston, Massachusetts, USA.
- Section of Infectious Diseases, Boston University School of Medicine, Boston, Massachusetts, USA.
- Division of Pulmonary and Critical Care Medicine, Massachusetts General Hospital, Boston, Massachusetts, USA.
Abstract
Chest X-ray (CXR) can assess pulmonary tuberculosis (TB) severity and may guide duration of treatment. However, the optimal radiological metric and its integration with clinical variables for predicting treatment outcomes remains unclear. We used logistic regression to associate human-read and commercial artificial intelligence-generated CXR metrics with unfavorable outcome in the TB Portals real-world dataset (n = 2809). We assessed the standalone predictive accuracy for each of 10 radiological features for unfavorable outcomes, and combined the best-performing features with other clinical data. We fine-tuned ensembles of convolutional neural nets (CNNs) to automate human-read percent of lung involved (PLI) measurement directly from CXR images (n = 5261). Human-read PLI is the only CXR finding associated with outcome across drug resistance and human immunodeficiency virus (HIV) subgroups and is optimally combined with age, sex, and smear grade for predicting treatment outcome. PLI predicts outcomes better than cavitation (area under the curve [AUC], 0.654 vs 0.581, respectively), performs better than all tested Qure.ai automated features (qXR v2), and improves outcome prediction when added to sex + age + smear grade (ΔAUC, 0.028 [95% confidence interval, .007-.042]). The CNN ensemble for predicting PLI >25% achieves an AUC of 0.850. PLI performs better than cavitation as a radiological marker for predicting TB treatment outcome, and improves risk stratification when combined with key clinical variables. Automation of PLI can be predicted using CNNs to enable scalability and accurate assessment.