TBscreen: A passive cough classifier for tuberculosis screening with a controlled dataset.

Recent respiratory disease screening studies suggest promising performance of cough classifiers, but potential biases in model training and dataset quality preclude robust conclusions. To examine tuberculosis (TB) cough diagnostic features, we enrolled subjects with pulmonary TB (N = 149) and controls with other respiratory illnesses (N = 46) in Nairobi. We collected a dataset with 33,000 passive coughs and 1600 forced coughs in a controlled setting with similar demographics. We trained a ResNet18-based cough classifier using images of passive cough scalogram as input and obtained a fivefold cross-validation sensitivity of 0.70 (±0.11 SD). The smartphone-based model had better performance in subjects with higher bacterial load {receiver operating characteristic-area under the curve (ROC-AUC): 0.87 [95% confidence interval (CI): 0.87 to 0.88], P < 0.001} or lung cavities [ROC-AUC: 0.89 (95% CI: 0.88 to 0.89), P < 0.001]. Overall, our data suggest that passive cough features distinguish TB from non-TB subjects and are associated with bacterial burden and disease severity.

Tuberculosis Infectiousness is Associated with Distinct Clinical and Inflammatory Profiles.

Interrupting transmission events to prevent new acquisition of infection and disease is a critical part of tuberculosis (TB) control efforts. However, knowledge gaps in understanding the biology and determinants of TB transmission, including poor estimates of individual infectiousness and the lack of accurate and convenient biomarkers, undermine efforts to develop interventions. Cough-generated aerosol cultures have been found to predict TB transmission better than any microbiological or clinical markers in cohorts from Uganda and Brazil. We hypothesized that highly infectious individuals with pulmonary TB (defined as positive for cough aerosol cultures) have elevated inflammatory markers and unique transcriptional profiles compared to less infectious individuals (negative for cough aerosol cultures). We performed a prospective, longitudinal study using a cough aerosol sampling system as in other studies. We enrolled 142 participants with treatment-naïve pulmonary TB in Nairobi, Kenya, and assessed the association of clinical, microbiologic, and immunologic characteristics with Mtb aerosolization and transmission in 143 household members. Contacts of the forty-three aerosol culture-positive participants (30%) were more likely to have a positive IGRA (85% vs 53%, P = 0.005) and a higher median IGRA IFNγ level (P < 0.001, median 4.25 IU/ml (0.90-5.91) vs. 0.71 (0.01-3.56)) compared to aerosol culture-negative individuals. We found that higher bacillary burden, younger age, and larger mean upper arm circumference were associated with positive aerosol cultures. In addition, novel host inflammatory profiles, including elevated serum C-reactive protein and sputum cytokines, were associated with aerosol culture status. Notably, we found pre-treatment whole blood transcriptional profiles associated with aerosol culture status, independent of bacillary load. Together, these findings suggest that TB infectiousness is associated with epidemiologic characteristics and inflammatory signatures and that these features may be used to identify highly infectious persons. These results provide new public health tools and insights into TB pathogenesis.