Background
Bowel sounds reflect gastrointestinal motility but are rarely used in clinical decision-making, largely because conventional auscultation is brief, subjective, and poorly standardized. Advances in digital stethoscopes and acoustic signal processing create opportunities to transform bowel sounds into quantitative biomarkers.
Aims
To explore whether acoustic features extracted from pre-endoscopic bowel sound recordings are associated with clinical characteristics and endoscopic findings in patients undergoing colonoscopy.
Methods
We conducted a monocentric, observational, cross-sectional study in the gastroenterology department of Mongi Slim Hospital (Tunisia) between March and June 2023. Consecutive adult patients scheduled for colonoscopy after a standardized low-residue diet and bowel preparation underwent a structured interview, physical examination and a five-minute abdominal sound recording within one hour before the procedure. Recordings were obtained over the right iliac fossa using a modified Littmann CORE digital stethoscope equipped with a piezoelectric sensor (1) connected directly to a smartphone. Sounds were manually labelled as bowel sounds, environmental noise or artefacts and subclassified into single bursts [sb], multiple bursts [mb] and harmonic sounds [h]. Acoustic features were extracted per labelled event and aggregated per patient (method A) and per sound subtype (method B), then entered multivariable logistic regression models to assess associations with selected clinical and endoscopic outcomes (significance threshold: p < 0.05).
Results
Ninety-five patients were included (mean age 56 ± 15 years; 48 men, 47 women). Colonoscopy was performed in 83 patients (87%), and at least one endoscopic abnormality was identified in 47 (57%). A total of 33,837 bowel sounds were labelled across all recordings. Sex, BMI, and difficulty passing gas were significantly associated with amplitude-related acoustic parameters (RMS and envelope). The presence of endoscopic abnormalities, particularly colonic polyps, was also significantly associated with these parameters and additionally with zero-crossing rate. Bowel preparation quality and time since last liquid intake showed consistent associations with spectral contrast, number of labels and zero-crossing.
Conclusions
In this pilot study, standardized pre-colonoscopy bowel sound recordings analysed using multivariable logistic regression yielded acoustic features nominally associated with clinical characteristics and endoscopic findings. These results are consistent with the feasibility of intestinal sound analysis as a non-invasive biomarker, but require confirmation in larger, multicentric and externally validated cohorts before clinical integration.
