DIFF-4 · VALIDATED ACCURACY

How accurate is SomniSense — really?

We don't believe in vague accuracy claims. Here are the numbers, the methodology, and the limits.

Quick Answer

SomniSense's snoring and breathing irregularity detection algorithm has been validated against in-lab polysomnography (PSG) in our internal study.

Key results (against PSG ground truth, internal study n=70+):

  • 91% Snoring detection sensitivity
  • 98.5% Snoring detection precision
  • 80% Breathing irregularity sensitivity
  • 91% Breathing irregularity precision
  • 87% BRI agreement vs clinical AHI (±5) (at AHI ≤ 30)

Our algorithm builds on years of sleep apnea research by our founder. SomniAI LLC is currently preparing the first peer-reviewed publication of the upgraded algorithm and the US patent application. We will update this page once published. These will become SomniSense's primary publicly cited scientific basis going forward.

SomniSense reports a Breathing Irregularity Index (BRI) — the same per-hour event-rate metric clinicians call AHI — but SomniSense is a wellness tool and does not diagnose sleep apnea or any other condition.

The four numbers that matter

Question Number What it means
Of the snores you make, what % does the app catch? 91% Snoring detection sensitivity. 9% of snores get missed.
When the app calls something a snore, how often is it actually a snore? 98.5% Snoring detection precision. False positives are rare.
Of breathing pauses that occur, what % does the app flag? 80% Breathing irregularity sensitivity. Conservative — borderline events are missed rather than over-reported.
When the app flags a pause, how often is it real? 91% Breathing irregularity precision. Combined breathing pauses + reduced breathing.
How close is your BRI to a clinical AHI? 87% BRI agreement vs clinical AHI (±5). Bland–Altman agreement on n=70+ paired nights (at AHI ≤ 30).

How we tested

Detection performance was measured against gold-standard in-lab polysomnography (PSG) recordings, audio-annotated by certified sleep technicians. Our algorithm builds on years of sleep apnea research by our founder. SomniAI LLC's first peer-reviewed publication of the upgraded algorithm and US patent application are in active preparation. Full methodology and limitations are in the white paper linked below.

  • Sample: n=70+ participants, paired nights (smartphone + PSG simultaneously) — internal validation
  • Population: adults with and without diagnosed sleep breathing issues
  • Recording: bedside smartphone (varied iPhone & Android models from 2018 onward), 50–90 cm from participant's head
  • Ground truth: PSG with synchronized audio channel; manual scoring by AASM-trained sleep technicians, blinded to SomniSense output
  • Comparison: per-event sensitivity, per-event precision, per-night BRI vs PSG AHI agreement (Bland–Altman analysis)

What we are not

SomniSense is a wellness and self-monitoring tool. It is not:

  • ❌ A medical device
  • ❌ A diagnostic for sleep apnea, OSA, or any other condition
  • ❌ A substitute for professional sleep evaluation
  • ❌ Validated for users under 18

If your BRI consistently exceeds 15 events per hour (the same threshold clinicians use for AHI), or if your partner observes prolonged breathing pauses, please consult a board-certified sleep specialist.

Want the full methodology paper?

A peer-reviewed publication and US patent application are in active preparation. Until they publish, the documentation here is the canonical accuracy reference.

Try Free for 7 Days

After trial: $7.99/mo or $49.99/yr. Cancel anytime in App Store / Google Play.