Abstract
The advent of high-resolution (HR) electrical mapping of slow wave activity has significantly improved the understanding of gastric slow wave activity in normal and dysrhythmic states. One of the current limitations of this technique is it generates a vast amount of data, making manual analysis a tedious task for research and clinical development. In this study we present new automated methods to classify, identify, and locate patterns of interest in gastric slow wave propagation. The classification method uses a similarity metric to classify slow wave propagations, while the identification algorithm uses the divergence and mean curvature of the slow wave propagation to identify and regionalize patterns of interest. The methods were applied to synthetic and experimental datasets and were also compared to manual analysis. The methods classified and identified patterns of slow wave propagation in less than 1 s, compared to manual analysis which took up to 40 min. The automated methods achieved 96% accuracy in classifying AT maps, and 95% accuracy in identifying the propagation pattern with a mean spatial error of 1.5 mm in comparison to manual methods. These new methods will facilitate the efficient translation of gastrointestinal HR mapping techniques to clinical practice.