FIELD NOTES / 12 OCTOBER 2026
Listen closely.
Question everything.
A closer look at the sounds, the interference and the work between a working prototype and useful evidence.
01 / SOUND EXPLORER
Hear what we’re studying.
Two short excerpts from a published bowel-sound dataset, alongside a synthetic interference example. The waveforms match the playable audio. These are not ANTRAM sensor results; no microphone is used.
A cluster of bursts
Illustrative interference
Recordings: Zahra Mansour (2025), Bowel sounds signal, under CC BY 4.0. Source annotations, not independently reviewed by ANTRAM. Excerpts trimmed, mean removed, resampled to 16 kHz and volume normalised; no denoising. Exact excerpt times and processing.
Sound starts only when you press play. Use a comfortable volume.
Sound or noise?
Compare the two research excerpts with the synthetic interference. Real speech, clothing friction and handling can resemble acoustic events. The illustration is not a recorded negative example or a benchmark: confirming an event requires context and independent review.
Loudness has been adjusted for listening. Do not compare these clips as measures of digestion strength.
02 / RESEARCH, WITHOUT THE SHORTCUTS
Two things worth understanding.
Sound is a signal. Not a verdict.
An abdominal microphone can capture acoustic activity alongside speech, clothing friction and movement. Before interpreting a pattern, we need to determine which events are genuine and how much recording time is usable.
The source labels describe acoustic patterns, not diagnoses. These excerpts illustrate the researchers’ single-burst and multiple-burst categories. Our own hardware still needs independently reviewed recordings and testing on people absent from training.
Read the original study ↗Editorial explainer · not clinician-reviewed · no diagnostic guidance
A working PCB is the beginning.
A connected sensor tells us the electronics can communicate. It does not establish sound-detection accuracy, physiological meaning or user benefit. ANTRAM’s next questions concern repeatable capture, contact quality, interference rejection and whether a sensor adds useful information beyond meal and symptom logs.
- Capture repeatable signals.
- Compare events with independent acoustic labels.
- Test on people absent from model training.
- Evaluate comfort and usefulness under an appropriate study protocol.
Proposed roadmap · not completed validation · not clinician-reviewed
Patterns, not diagnoses.
These examples do not measure digestion quality, identify food intolerance or diagnose a condition. Published research on other systems does not validate ANTRAM.
Explore our evidence and limitations ↗