ATS 2024: Using Cough Sounds to Predict Chest Radiography Results
Summary
In this pilot study, we were able to collect coughs in the course of routine patient workflows and demonstrate that AI enabled algorithms are predictive of which individuals are likely to have an abnormal CXR.
A triage tool that could identify patients who are most likely to have chest radiographic (CXR) abnormalities wouldimprove the efficiency of CXR utilization. Acoustic artificial intelligence techniques are being developed to extract diagnosticallyrelevant features from cough. Here we report the feasibility and preliminary performance of an AI-enabled cough classifier’s abilityto identify which patients are most likely to have abnormal CXRs based on cough sound.
