Table 3 Characteristics and results of included studies

Craine et al. 1999 [13]

IBS

Case-control using healthy controls (2 gate): healthy (15) and IBS (18) matched for gender bias. Heterogeneity test and ROC analysis on one dataset.

GI clinic in a county medical centre. USA

Low

Electronic stethoscope lower right quadrant. Analysis of sounds (150–450 Hz)

Six metrics and difference between fasting and fed states. Useful: fasting s-s interval (2 min recording)

Rome criteria for IBS. Symptoms >  6 months.

Craine et al. 2001 [14]

IBS and Crohn’s disease

Case-control using healthy controls and alternative diagnosis groups (3 gate): healthy (37), IBS (45), and Crohn’s (25). Crohn’s younger. ‘Healthy’ comprised community volunteers and patients attending GI clinic for diagnoses unrelated to abdominal discomfort. Heterogeneity test and ROC analysis on one dataset.

GI Clinic in a county medical centre. USA

Low

One electronic stethoscope to the right of umbilicus and ‘Enterotach’ analysis (start and stop times of sounds 150 to 450 Hz)

Fasting s-s interval (2 min recording)

Rome criteria. Clinical, radiological, and biopsy findings for Crohn’s disease.

Craine et al. 2002 [15]

IBS and non-ulcer dyspepsia (NUD)

Case-control using healthy controls and alternative diagnosis groups (3 gate): controls (10), IBS (11), and NUD (19) (split into 2 groups on bowel sound characteristic). Controls attending clinic for diagnoses unrelated to abdominal discomfort. Heterogeneity testing for multiple features.

GI Clinic in a county medical centre. USA

Moderate

Three electronic stethoscopes. Analysis enterotachogram (start and stop times of sounds and magnitude of sound envelope)

Sound mapping and measurement of freq and s-s interval.

Rome II criteria.

Hadjileontiadis et al. 1999 [16]

IBS, diverticular disease (DD), bowel polyp (2 cm) and ulcerative colitis (UC)

Case-control (multi-gate): healthy (9), pre-confirmed diseases of the large bowel (7) (IBS, diverticular disease, 2 cm bowel polyp, ulcerative colitis.

No details on incl. criteria given. Scatterplot analysis.

Unclear, probably hospital. Greece.

Moderate

Audioscope.

Denoised with WTST-NST filter

Three procedures for analysis based on higher order crossings (HOC): scatter plots, distance from white Gaussian noise, calculation of the weighted φ2 statistic. 16 min BS recordings

(8 min each) above right and left anterior superior iliac spine.

Unclear—“pre-confirmed”

Yoshino et al. 1990 [17]

Intestinal obstruction (large and small bowel)

Case-control with healthy controls (2 gate): healthy (4) and intestinal obstruction (21) (later split into 17 with simple obstructions and 4 with strangulating). Tests of association.

Hospital. Japan

Low

Foam covered microphone. Right lower abdomen.

Categorised sounds into three types based on frequency characteristics: Peak freq the most common freq). The upper and lower limits of freq range. Freq over 900 kHz present or absent (15 mins recording)

Obstructions were diagnosed based on clinical examinations including plain film based X-rays or laparotomy.

Ching et al. 2012 [18]

Small and large bowel obstruction

Cross-sectional (single -gate): 71 patients with suspected bowel obstruction (split into acute bowel obstruction, subacute bowel obstruction and no bowel obstruction). Little info on incl. criteria. Tests of heterogeneity across groups.

General hospital. Singapore

Low

Electronic stethoscope.

Sound duration, sound to sound interval, dominant frequency and peak frequency from 6 tracks of 8 s: 2 at each of 3 locations on lower abdomen.

Radiological imaging: plain film radiology in all, and CT in 85.9% of patients, and symptoms and physical signs.

Sugrue and Redfern 1994 [19]

Acute abdomen, varying severity (appendicitis, cholecystitis, and intestinal obstruction)

Case-control (multi- gate): healthy (63) and patients with an acute abdomen (61) (multiple conditions: appendicitis (25, 18 acute, 7 perforated), obstructions (21, 12 large, 9 small), cholecystitis (15))

Teaching hospital. Ireland

Low

Microphone (range 30–15,000 Hz) taped to right iliac fossa. 4 computer programs.

Ten-minute recordings (after fasting for 2 h in the case of controls). Five features: sound length, number of sounds, sound amplitude, silence length, sound/silence ratio.

Surgery and histology (for all but one)

Kim et al. 2011A [20]

Delayed gastric emptying

Case-control with healthy controls (2 gate): healthy (12), patients with spinal cord injury and delayed gastric emptying (4).

Tests of heterogeneity across groups. Test of correlation between index and reference tests.

Unclear, assume University Hospital, Korea

High

Piezo-polymer sensor (range 8–2200 Hz) multiple filters. Denoising, segmentation, feature extraction (jitter and shimmer).

R3 channel recordings (right upper quadrant, left upper, and left lower quadrant.

Nine features (jitter, shimmer, and trace) used to model eCTT. Fasting conditions, 200 g test meal at 9:00 am, then 10-min recordings at 9:30 am, 1:00 pm, and 5:00 pm.

Metcalf’s method. Ingestion of radiopaque marker and X-rays, to calculate total CTT.

Kim et al. 2011B [21]

Delayed gastric emptying

Case-control with healthy controls (2 gate): healthy (12), patients with spinal cord injury, and delayed gastric emptying (6). All male, controls younger.

Tests of heterogeneity across groups. Used K-fold cross-validation to calculate correlation between index and reference tests.

Unclear, probably university hospital, Korea

Super high

Piezo-polymer sensor (range 8–2200 Hz) multiple filters. Denoising, segmentation, feature extraction (jitter and shimmer). Training and estimation of the back propogation neural network.

3 channel recordings (right upper quadrant, left upper, and left lower quadrant.

Six jitter and shimmer features used to model eCTT. Model refined through an artificial (back propogation) neural network. Fasting conditions, 200 g test meal at 9:00 am, then 10-min recordings at 9:30 am, 1:00 pm, and 5:00 pm.

Metcalf’s method. Ingestion of radiopaque marker and X-rays, to calculate total CTT.

Tomomasa et al. 1999 [22]

Pyloric stenosis and impaired gastric emptying in infants

Case-control with healthy controls (2 gate): healthy (6), infants with infantile hypertrophic pyloric stenosis (15). Similar ages. Heterogeneity study (SI across 2 groups HPS and healthy) and correlations between SI and standard measure of gastric emptying.

Unclear, probably children’s hospital, Japan

Low

Condenser microphone sound sensor attached with electrocardiograph tape 3 cm below umbilicus, for 60 min (when fasted) before pyloromyotomy, and at 9 to 12 h, 20 to 24 h, 40 to 48 h, and 112 to 120 h after the operation. Recordings made when sleeping (for at least 20 min.)

Sound index (SI) as the sum of absolute signal amplitudes expressed as volts per minute.

Gastric emptying measured using marker dilution-double sampling method. Diagnosis of pyloric stenosis (based on ?).

Spiegal et al. 2014 [23]

Post-operative ileus

Case-control (3 gate): healthy controls (8), patients with post-operative ileus (25), post-operative patients tolerating feeding (7).

Controls 62.5% male, patients 100% male. Heterogeneity test across groups and an ROC analysis on differentiation between healthy and POI on the same dataset (note, those tolerating feeding not in ROC analysis)

1 teaching hospital. USA

Unclear (few details provided)

AGIS’ sensor with microelectronic microphone and computer to calculate motility scores.

Intestinal rate (average rate of pulses resulting from motility events), in a 60 min recording (post meal in controls).

Pragmatic definition of POI: presence of one or more of (1) nausea that precluded advancement of diet beyond sips on POD #1 or later, (2) post-op vomiting that precluded any oral intake, or (3) nasogastric tube decompression

Kaneshiro et al. 2016 [24]

Post-operative ileus

Cross-sectional (single gate) longitudinal prospective: subjects recovering from colorectal surgery (28). Consecutive sample. Identified an algorithm that maximised predictive discrimination. ROC analysis to assess sensitivity and specificity using the same data.

3 hospitals. USA

Unclear (few details provided)

AGIS’ sensor with microelectonuc microphone. 2 sensors either side of the umbilicus.

Intestinal rate (number of acoustic motilty events per minute. Metrics used were a drop in IR between POD1 and POD 2, plus % time that the subject had an IR below the 5th percentile.

Definition of POI was pragmatic. 3 criteria used.

Campbell et al. 1989 [25]

Diarrhoea—severe (post-gastrectomy) and mild idiopathic

Case-control (3 gate): healthy (22), severe diarrhoea (5), and mild diarrhoea (7). Ages differed between groups. Heterogeneity test and correlation between ref standard and index test values (additional test on effect of cisapride)

Unclear, probably teaching hospital. UK

Low

One transducer. Filter, integration, fast fourier transform, and ‘SVA’ analysis.

Post-prandial 3.5-h recording. SVA values expressed in linear energy units.

Oral caecal transfer time (hydrogen breath technique)

Liatsos et al. 2003 [26]

Small volume ascites

Case-control with healthy controls (2 gate): healthy (20), cirrhotic patients (with proven small-volume ascites) (20). Healthy slightly younger. Analysis of bowel sounds using a higher order crossings based technique to distinguish between the two groups.

Hepatobiliary and Liver Transplantation Unit, teaching hospital. UK

Moderate

Electronic stethoscope in a semi-soundproofed room. Right upper and lower abdomen. Subject lying supine.

16 min per patients. Fasted. WTST-NST Filter, denoised bowel sounds, HOC analysis, linear discrimination classification.

Small volume ascites picked up on ultrasound, but not clinical examination.