Unexpected sawtooth artifact in beat-to-beat pulse transit time measured from patient monitor data

Object: It is increasingly popular to collect as much data as possible in the hospital setting from clinical monitors for research purposes. However, in this setup the data calibration issue is often not discussed and, rather, implicitly assumed, while the clinical monitors might not be designed for the data analysis purpose. We hypothesize that this calibration issue for a secondary analysis may become an important source of artifacts in patient monitor data. We test an off-the-shelf integrated photoplethysmography (PPG) and electrocardiogram (ECG) monitoring device for its ability to yield a reliable pulse transit time (PTT) signal. Approach: This is a retrospective clinical study using two databases: one containing 35 subjects who underwent laparoscopic cholecystectomy, another containing 22 subjects who underwent spontaneous breathing test in the intensive care unit. All data sets include recordings of PPG and ECG using a commonly deployed patient monitor. We calculated the PTT signal offline. Main Results: We report a novel constant oscillatory pattern in the PTT signal and identify this pattern as a sawtooth artifact. We apply an approach based on the de-shape method to visualize, quantify and validate this sawtooth artifact. Significance: The PPG and ECG signals not designed for the PTT evaluation may contain unwanted artifacts. The PTT signal should be calibrated before analysis to avoid erroneous interpretation of its physiological meaning.

💡 Research Summary

The authors investigated a previously undocumented artifact that appears when pulse transit time (PTT) is derived from electrocardiogram (ECG) and photoplethysmography (PPG) signals recorded by standard bedside patient monitors. Two retrospective datasets were used: (1) recordings from 33 patients undergoing laparoscopic cholecystectomy (average duration ~27 min) captured on Philips IntelliVue MP60/MX800 monitors, and (2) recordings from 22 intubated ICU patients during a spontaneous breathing test (5 min each) captured on the same monitor family. ECG was sampled at 500 Hz, PPG at 125 Hz; the latter was up‑sampled to 500 Hz by linear interpolation. For each cardiac cycle, the R‑peak of the ECG defined the start time, and the maximum of the first derivative of the PPG ascent defined the pulse arrival time. The interval between these two points constituted the beat‑to‑beat PTT, which was then resampled at 4 Hz using cubic spline interpolation.

Visual inspection of the resulting PTT traces revealed a regular, sawtooth‑shaped oscillation with a period of roughly 100 seconds (≈0.01 Hz). Power spectral density analyses confirmed dominant peaks at 0.01 Hz, 0.012 Hz, and occasionally 0.1 Hz across all subjects, regardless of surgical stage or physiological state. To rigorously assess whether this low‑frequency component reflected genuine physiology or a systematic measurement error, the authors applied the de‑shape short‑time Fourier transform (dsSTFT), a nonlinear time‑frequency method that isolates the fundamental instantaneous frequency of non‑sinusoidal signals while suppressing harmonics. dsSTFT maps consistently displayed a strong, persistent line at 0.01 Hz throughout each recording, corroborating the presence of a non‑physiological, device‑related oscillation.

The authors attribute the artifact to a lack of calibration for PTT measurement in the monitor’s firmware. The ECG and PPG channels are not guaranteed to be perfectly synchronized; differences in sampling rates, internal buffering, and the proprietary data‑dumping software likely introduce a fixed latency that repeats with the monitor’s internal timing cycle. A private communication with a Philips engineer confirmed that the device was never intended for precise inter‑channel timing analysis. Consequently, any research that extracts PTT directly from such monitors without accounting for this hidden latency risks drawing erroneous conclusions, for example, misinterpreting the sawtooth pattern as a physiological rhythm or using it as a predictor in clinical models.

The study underscores the broader issue of using clinical monitors as “big‑data” sources without thorough validation. While the authors did not test other monitor brands, they argue that similar hidden calibration problems may exist elsewhere, especially when multiple physiological streams are fused for secondary analyses. They recommend that investigators (1) obtain detailed timing specifications from manufacturers, (2) verify inter‑channel alignment using external reference devices, and (3) apply post‑hoc correction algorithms when raw timing offsets are identified.

In summary, this work provides the first quantitative description of a sawtooth‑type artifact in PTT derived from off‑the‑shelf patient monitors, demonstrates its detection using advanced time‑frequency analysis, and calls for systematic calibration procedures before employing bedside monitor data for research purposes.