Preprandial and Postprandial Body Surface GastroIntestinal Mapping Source Separation based Denoising allows Intestinal Spike Bursts Analysis.

Body Surface GastroIntestinal Mapping (BSGIM) could be useful in the evaluation of gastrointestinal motility disorders affecting, for example, patients with Crohn's disease or functional gastrointestinal disorders. BSGIM is the mixture of stomach, intestinal and colonic electrical activity, it is composed of multiple frequential components generally divided in two categories: Slow waves (SWs) and Spike Bursts (SBs). The interpretation of BSGIM is made difficult by the presence of electrocardiographic and respiratory noises, which are in similar frequency range as SBs and SWs respectively, rendering their analysis challenging. In this paper, we propose and compare five BSGIM denoising techniques: a bandpass filter and four semi-blind source separation methods based on the pseudo-periodic component analysis (PICA) differentiated by strategies for component sorting and reconstruction. Firstly, the methods are compared using a simulation framework able to synthetize realistic noisy BSGIM. Denoised signals are compared to the ground truth using two metrics: bandpower and main frequency. The best method is then applied to extract band power metrics from preprandial and postprandial BSGIM data (28 electrodes, 3 hours of recording at 250 Hz) recorded on 3 healthy volunteers. Simulation results show that PICA with peaks correlation sorting and component filtering is the more efficient to reconstruct both SWs and SBs, in particular intestinal SBs. Denoised BSGIM data from volunteers show a marked increase in SWs and intestinal SBs power in the postprandial state compared to the preprandial state, which is consistent with an increase of motility caused by digestion reported in the literature.