Facing Big Data in Longterm High Resolution Manometry of the Esophagus by Automated Swallow Detection and Classification

High resolution manometry (HRM) is the gold standard in the diagnosis of esophageal motility disorders. In clinical practice HRM is only performed for about 30 minutes with a very standardized protocol. Esophageal motility disorders can cause severe discomfort and symptoms even if they only occur spontaneously. Such temporary motility disorders are often not detectable in conventional HRM. Longterm HRM is a unique way to study the circadian behavior of the esophagus and thus will contribute to a more effective treatment of esophageal motility disorders. Prolonging the time of measurement to a full of 24 hours leads to an exponential increase in data and work, which makes longterm HRM not yet feasible in clinical routine. 40 longterm HRM measurements of patients and healthy volunteers were meticulously tagged by hand and used as training and validation data for our machine learning model. As each longterm HRM dataset consists of at least 155 million datapoints at total amount of 6 billon datapoints big data in medicine can no longer be handled without support by artificial intelligence. Therefore, we implemented an automated swallow detection and a first classification system to help medi-cal professionals gain faster and easier insight in circadian (dys) motility of the esophagus. This significantly reduces the time and manpower required for longterm HRM evaluation and paves the way for a broad clinical use in the future in patients with temporary symptoms of esophageal motility disorders but though high levels of distress.