WAVE
Wobble Amplitude & Variability Evaluation
Files stay on this device. Works offline.
Bring in your nights
ResMed EDFPLD gives device minute ventilation. Add BRP for airflow checks and breath-level zoom.
Analysis settings 20–120 s waves · 120 s envelope
Five minutes at each session end are excluded. Leak and invalid-value exclusions include a ±120-second buffer. Changing settings requires reprocessing; displayed results retain their original settings.
Import details
How the wubs change
● Device MV◆ Airflow-derivedFaint dots: individual periods · Curves: centered, distance-weighted trends
Hover for exact values; click to open a period. Smoothing uses nearby dates on both sides and keeps sources and long gaps separate. It changes only these charts.
Wubscape
One row per period. Hours since that period’s first session; gaps stay empty. Click a row to open it.
Compare periods
Compare date ranges within one ventilation source. Each period gets equal weight. These are within-person descriptions, not causal treatment estimates.
Recording periods
Grouped into months, newest firstSelected period
Labels are your annotations. Fixed/variable support describes recorded pressure behavior and does not identify a therapy mode.
Underlying airflow · zoom to individual breaths
Requires a matching BRP file. Drag or click this chart to zoom; minimum window five seconds.
Candidate packets
Threshold candidates, not scored clinical events. Select a row to inspect context. The table shows at most 300; exports include all.
Sensitivity and airflow agreement
Is the slow structure more than randomized waves?
Test uninterrupted 90-minute blocks. Phase controls preserve the wave spectrum. IAAFT also preserves the value distribution. They do not establish a physiological oscillator. All-block IAAFT runs can take a while; the page stays responsive and cancellation retains completed results.
How WAVE works
Strength, duration, timing
The amplitude envelope measures how the size of faster ventilation waves changes. Packet duration and repeat spacing are separate questions. A high quiet-to-active ratio can occur even when absolute waves are small.
Same signal pipeline
Third-order Butterworth bandpass, forward/backward filtering, reflected-padding Hilbert magnitude divided by √2, and centered smoothing. Default 120-second smoothing uses 61 two-second samples (122 s). The local baseline uses 302 s.
Data and quality
Device MV is preferred when PLD is present. BRP-only data use positive airflow averaged in two-second bins and smoothed over 14 s. The two sources have different amplitudes and remain labeled. Five-minute edge trims and leak/invalid buffers never bridge sessions.
Longitudinal summaries
Sessions within the chosen gap are grouped, with an 18-hour maximum span and separate ventilation sources. Groups under three recorded hours are labeled Nap. Period amplitudes are usable-time-weighted session medians, not pooled medians. Calendar trends use centered Gaussian weighting over the selected 7-, 14-, or 28-day window (standard deviation one quarter of the window). They use the displayed periods, keep sources separate, and break at gaps over 10 days.
Controls and interpretation
Controls use a 90-minute block with three minutes trimmed from each envelope end. Slow power covers 3–30-minute periods; repeat peaks search 5–25 minutes. IAAFT uses up to 100 iterations and reports spectral mismatch. Benjamini–Hochberg q values correct each metric across the current run’s blocks separately by method and source. Quick runs have coarse p-value resolution.
Privacy and limits
No analytics, accounts, servers, uploads, external fonts or third-party libraries. Data remain in memory until cleared; saving is explicit. Recording time is not scored sleep. No EEG, CO₂, effort or confirmed arousals are measured. Experimental, not a medical device.
WAT v0.5 workflow reference: main fcd24e08 (August 8, 2026). Earlier WAT versions are not used. WAVE 0.1.0 · MIT © 2026 existentialblu. Numerical methods follow EDF calibration, SciPy-style SOS filtering and Hilbert envelopes. See the companion README for validation.