REFERENCE REVIEW · 8 SEPTEMBER 2026
Closer to a healthy heartbeat.
The ECG now uses a healthy reference fit synchronized to contraction. Apical tracers circulate more visibly, and colour Doppler has a focused, less uniform appearance. This is still an approximate educational simulation.
Before
Broad experimental ECG; whole-sector colour; short tracer lifetimes.
Download recordingRevised
Healthy reference ECG; focused mitral colour region; longer tracer transit.
Download recordingEach clip contains two beats at the source rate of 64.3 bpm. The actual WebGPU rendering code and ECG drawing function were captured offline at fixed simulation times and encoded at 30 fps. These are not browser screen recordings or performance benchmarks.
ECG: a reference fit, with clear timing
The original forward model produced an atypical trace: a weak, broad ventricular complex and a dominant T wave. Its inferred ventricular activation spanned about 188 ms. The new default fits compact waveform components to recorded healthy Lead II and keeps the original calculation under Geometry source · experimental.
Intervals apply at the original playback rate. Changing the speed scales the entire authored cycle; it does not simulate physiological rate adaptation.
PTB healthy control 104 is the fit reference. Its R-aligned waveform correlation with the fit is 0.995, which describes fit quality, not diagnostic accuracy. Independent examples 105 and 116 give correlations of 0.904 and 0.674; their Q/S depths, R amplitudes and T waves differ. Three examples are not population validation.
See variation between the three healthy recordings
| PTB record | Median rate | Median R peak | Role |
|---|---|---|---|
| 104 / s0306lre | 61.5 bpm | 1.161 mV | Fit reference |
| 105 / s0303lre | 78.4 bpm | 0.899 mV | Independent example |
| 116 / s0302lre | 62.0 bpm | 0.685 mV | Independent example |
All three headers identify healthy controls. Lead II was read at 1000 Hz and 2000 ADC units/mV, filtered with a third-order 0.5–40 Hz forward-backward Butterworth filter, R-aligned and median averaged. The pre-P baseline was removed. These plots show processed signals, not unfiltered recordings.
The R anchor follows the maximum of the fitted LV volume curve, and the T wave falls during late ejection. Healthy-mode electrical activation uses the same phase clock with remapped atrial and ventricular timings. There is no measured ECG in the FBX: this synchronization is a model assumption, and the healthy waveform is not calculated from the atlas’s electrical sources.
Doppler: better visual resemblance, limited quantitative fidelity
Real normal mitral and five-chamber Doppler clips were inspected in Aquilina et al. The revised view uses a selectable colour region, power grain/dropout and the same moving velocity field as the tracers. Red-toward/blue-away direction, wall filtering and Nyquist aliasing remain explicit.
| Feature | Comparison | Remaining gap |
|---|---|---|
| Flow direction | Mitral inflow is predominantly toward an apical probe; outflow has an away component. Direction changes and aliasing appear in both the model and real clips. | Local jets depend on an approximate flow field and estimated plane. |
| Colour appearance | Focused colour acquisition and mottled power resemble the real clips more closely than uniform chamber fills. | Grain is a display approximation, without RF/IQ pulse ensembles or a clinical estimator. |
| Cycle timing | The fitted motion produces two filling peaks and main systolic outflow, but also an unintended secondary outflow pulse between filling peaks. | Velocities, E/A measurements and valve flow profiles have no matched clinical ground truth. |
| Anatomy and image quality | The chamber/valve anatomy drives each slice. | RV geometry, tissue texture and fine valve echoes differ; there is no chest wall, ribs, breathing or complete acoustic artefact model. |
The real Doppler clips and PTB ECGs are separate examples, not synchronized measurements of the same person or this atlas. No pixelwise similarity score, velocity-accuracy percentage or clinical validation is claimed. The external clips are linked at their source and are not redistributed in this app.
Apex: more circulation, fewer stalled steps
The baseline showed slow motion rather than permanently fixed particles. The revised model adds an assumed ventricular recirculation field, driven by filling with a 120 ms decay. Tracers start farther inside the cavities, slide along boundaries and live longer before being reseeded.
| Measured in the recordings | Before | Revised |
|---|---|---|
| Median apical movement per frame | 0.279 mm | 0.741 mm |
| Median apical movement / time | 8.38 mm/s | 22.24 mm/s |
| Apical steps below 0.02 mm | 11.49% | 1.75% |
These measurements cover two cycles at fixed 1/30-second steps after one warmup cycle. Apical samples are LV y < −65 mm and RV y < −60 mm in reference coordinates. Reseeds are excluded. The interior seed distribution changed, so these are conditional visualization statistics, not paired particles or measured blood washout.
The recirculation field is projected to remove divergence and through-wall face flow on the reference voxel grid. Interpolation and moving-domain transport remain approximate. Slower apical flow and near-wall motion are expected; the aim is coherent circulation, not a minimum speed everywhere. This is not CFD, and reseeded tracers cannot establish true blood residence time.
Sources and scope
PTB Diagnostic ECG Database — Bousseljot, Kreiseler and Schnabel (1995), Physikalisch-Technische Bundesanstalt, via PhysioNet. Data used under the Open Data Commons Attribution License v1.0; the displayed medians are processed derivatives.
Aquilina et al. (2007), Normal adult echocardiography — apical views — real mitral and five-chamber Doppler video references, Figures 11 and 14. © Images in Paediatric Cardiology, CC BY-NC-SA 3.0.
Elbaz et al. (2014), Vortex flow during early and late LV filling in normal subjects — physiological basis for recirculation during filling. Our assumed vortex is not a reproduction of their MRI measurements.
ECGSYN — precedent for parameterized synthetic ECG morphology. This app uses an independent fit, without embedding ECGSYN code.
Numerical checks and native rendering checks cover the declared models, pause/scrubbing, loop continuity and colour regions. No browser end-to-end test, real-GPU performance benchmark or clinical review was performed. The source geometry, textures and baked heartbeat are unchanged.