NPJ Digit Med. 2026 May 27. doi: 10.1038/s41746-026-02796-y. Online ahead of print.
ABSTRACT
Gastroesophageal reflux disease (GERD) diagnosis traditionally relies on acid exposure time (AET) obtained from 24-h multichannel intraluminal impedance-pH (MII-pH) monitoring, the gold standard for GERD diagnosis. However, a negative result (AET < 4%) does not always exclude GERD, as the limited 24-h monitoring window may fail to capture reflux events in patients with intermittent or low-frequency reflux. To address this limitation, we proposed a complementary machine learning-based framework targeting exclusively patients with negative MII-pH results (AET < 4%) to identify potential false-negative cases within this cohort, by integrating statistical and waveform-derived features from pH signals to enhance anomaly detection. Using one-class support vector machine and support vector data description models trained on real-world MII-pH datasets, the framework achieved an score of approximately 0.9 and identified potential anomalies undetected by the conventional AET criteria. Explainable AI techniques using Shapley additive explanations showed that features such as kurtosis and peak-to-peak amplitude contributed significantly to the identification of subtle reflux patterns within this cohort. These anomalies may indicate additional candidates for clinical reassessment within the AET-negative cohort. This complementary approach, operating downstream of the conventional MII-pH diagnostic system, could help identify potential false-negative cases among patients with negative MII-pH results, potentially assisting in their proper clinical management.
PMID:42204253 | DOI:10.1038/s41746-026-02796-y
