Automatic Characterization of User Errors in Spirometry

Andrew Luo, Eric Whitmire, James W. Stout, Drew Martenson, Shwetak Patel
An example of a spirometry device


Spirometry plays a critical role in characterizing and improving outcomes related to chronic lung disease. However, patient error in performing the spirometry maneuver, such as from coughing or taking multiple breaths, can lead to clinically misleading results. As a result, spirometry must take place under the supervision of a trained specialist who can identify and correct patient errors. To reduce the need for specialists to coach patients during spirometry, we demonstrate the ability to automatically detect four common patient errors. Creating separate machine learning classifiers for each error based on features derived from spirometry data, we were able to successfully label errors on spirometry maneuvers with an F1-score between 0.85 and 0.92. Our work is a step toward reducing the need for trained individuals to administer spirometry tests by demonstrating the ability to automatically detect specific errors and provide appropriate patient feedback. This will increase the availability of spirometry, especially in low resource and telemedicine contexts.