Automated Inflammatory Bowel Disease Detection Using Wearable Bowel Sound Event Spotting

Annalisa Baronetto ^1,2* Sarah Fischer ^3,4 Markus Neurath ^3,4 Oliver Amft ^1,2

• ¹ University of Freiburg, Freiburg, Germany
• ² Hahn-Schickard-Gesellschaft für angewandte Forschung, Villingen-Schwenningen, Germany
• ³ Department of Medicine 1, University Hospital Erlangen, Erlangen, Bavaria, Germany
• ⁴ Deutsches Zentrum Immuntherapie, Medizinische Fakultät, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Bavaria, Germany

We employ pattern spotting to detect rare Bowel Sound (BS) events in continuous abdominal recordings using a smart T-Shirt with embedded miniaturised microphones. Subsequently, we investigate the clinical relevance of BS spotting in a classification task to distinguish patients diagnosed with Inflammatory Bowel Disease (IBD) and healthy controls. Abdominal recordings were obtained from 24 IBD patients with varying disease activity and 21 healthy controls across different digestive phases. In total, approx. 281 h of audio data were inspected by expert raters and thereof 136 h were manually annotated for BS events. A deep-learning-based audio pattern spotting algorithm was trained to retrieve BS events. Subsequently, features were extracted around detected BS events and a Gradient Boosting Classifier was trained to classify IBD patients vs. healthy controls. Stratified group K-fold cross-validation experiments yielded a mean Area Under the Receiver Operating Characteristic curve ≥ 0.83 regardless of whether BS were manually annotated or detected by the BS spotting algorithm. We further explore classification window size, feature relevance, and the link between BS-based IBD classification performance and IBD activity. Our results confirm that automated BS retrieval and our BS event classification approach have the potential to support diagnosis and treatment of IBD patients.