Questions about how individual parts of claim 1 might be changed. A question may describe something different from what the claim language recites, or a version of it; no view is offered on which.
No entries.
Section 2 · machine-generated
Variations
Further arrangements of the same subject matter, described here as of the publication date at the foot of this page.
Other field of use
Virtual lead reconstruction for equine monitoring
A veterinary recorder uses skin electrodes at two torso positions to measure two equine electrocardiogram leads in successive intervals. A training archive contains simultaneous multilead equine recordings grouped by rhythm and electrode placement. Software extracts the two input leads from separate intervals of each training record and uses the simultaneous leads from the first interval as targets. At inference, the trained model estimates the virtual leads on the time grid of the first input lead.
Component substitution
Ridge regression for virtual lead reconstruction
A multivariate ridge regression model replaces the neural network. Software detects ventricular depolarization peaks in both sequential recordings, groups beats by rhythm and waveform shape, and resamples each representative beat to a fixed number of points. Training pairs concatenate representative beats from separate intervals as inputs and use simultaneous multilead beat samples from the first interval as targets. At inference, the fitted regression maps the concatenated input beats to a set of virtual lead templates on a shared beat grid.
Parameter or scale variation
Thirty second recordings for six lead reconstruction
A portable recorder samples one lead for 30 seconds at 500 samples per second, pauses for 5 seconds while the electrode position changes, and samples a second lead for 30 seconds. Training software extracts input pairs with the same timing from simultaneous six-lead recordings. The model receives 15,000 samples per input lead and generates six virtual lead traces of 15,000 samples each on the time grid of the first recording.
Combination with existing systems
Virtual leads in a treadmill stress test
A two-lead recorder exchanges timestamps and exercise stage identifiers with a treadmill stress testing workstation. During each constant-workload stage, the recorder acquires the two leads in successive windows, and software pairs windows from that stage with similar heart rates. A reconstruction model trained on simultaneous stress test recordings estimates a multilead trace on the first window's time grid. The workstation stores the measured and virtual traces with their acquisition times, reconstruction status, and exercise stage.
Automated or data-driven variant
Automatic selection of matched recording windows
An acquisition application divides both sequential lead recordings into overlapping windows and calculates heart rate, beat interval variability, baseline drift, and clipping for each window. It rejects windows that exceed configured drift or clipping limits and ranks the remaining pairs by differences in heart rate and beat interval variability. The application sends the highest-ranked pair that meets the matching thresholds to the trained reconstruction model. It stores the selected timestamps with the virtual leads and records a quality flag if no pair meets the thresholds.
Virtual leads with sample level uncertainty
An ensemble of reconstruction models trains on different patient subsets drawn from an archive of simultaneous multilead electrocardiograms. Each model receives the same two sequential input recordings and predicts virtual leads on the first recording's time grid. Software calculates the ensemble median and interquartile range for each lead at each sample. The review interface plots the median traces and displays the interquartile ranges as bands that show disagreement among the models.
Cite aspatent.fyi, US 2026/0354729 A1, claim 1, first published 2026-11-19T02:47:19-05:00, https://patent.fyi/US20260354729/, SHA-256 a96126e7e4fae7f8a0d88da2ea367a021d327276f46fd278e0302e9f0de616f5