Diagnostics · 12 authors, 8 centres
This summary was generated by AI from a single paper. It has not been reviewed by a clinician and is not clinical advice. Verify against the source before acting on it.
This narrative review examines how artificial intelligence can improve the measurement of pH-impedance metrics for diagnosing gastroesophageal reflux disease (GERD). The main finding is that AI demonstrates high performance in identifying reflux episodes and other metrics, but limitations include limited training data and the need for further validation.
Gastroesophageal reflux disease (GERD) is common, with a global prevalence of 8–33%, and pH-impedance monitoring is used for diagnosis, but interpretation can be time-consuming. Artificial intelligence (AI) is being explored to enhance measurement of metrics like reflux episodes, baseline impedance, and post-reflux swallow-induced peristaltic wave (PSPW) index. Methods: This is a narrative review updating literature on AI applications in pH-impedance metrics, discussing machine learning and deep learning techniques, including decision trees and convolutional neural networks (CNNs), and outlining steps for AI model development such as data collection, model development, interpretability, clinical validation, and collaboration. Key Results: AI models have shown high accuracy in identifying reflux episodes (e.g., 87% accuracy with inter-rater correlation coefficient [ICC] of 0.965) and PSPW events (82% accuracy with ICC of 0.921). AI can also extract baseline impedance metrics, such as the upright:recumbent AIBI ratio, which may predict treatment response with an area under the curve (AUC) of 0.661. Studies used datasets from patients with GERD symptoms, with sample sizes like 106 patients and thousands of impedance events. Limitations: The review notes that training data is limited, AI models need validation on diverse populations, compatibility with different software is unverified, interpretability is not studied, and accuracy is currently less than 90%. Implications: AI has the potential to automate and enhance pH-impedance analysis, particularly when acid exposure time (AET) is inconclusive (4–6%), aiding in precise GERD diagnosis. However, further research and collaboration are needed for clinical implementation.
