Validation of a New Method for Stroke Volume Variation Assessment: a Comparaison with the PiCCO Technique

This paper proposes a novel, simple and minimally invasive method for stroke volume variation assessment using arterial blood pressure measurements. The arterial blood pressure signal is reconstructed using a semi-classical signal analysis method allowing the computation of a parameter, called the first systolic invariant INVS1. We show that INVS1 is linearly related to stroke volume. To validate this approach, a statistical comparaison between INVS1 and stroke volume measured with the PiCCO technique was performed during a 15-mn recording in 21 mechanically ventilated patients in intensive care. In 94% of the whole recordings, a strong correlation was estimated by cross-correlation analysis (mean coefficient=0.9) and linear regression (mean coefficient=0.89). Once the linear relation had been verified, a Bland-Altman test showed the very good agreement between the two approaches and their interchangeability. For the remaining 6%, INVS1 and the PiCCO stroke volume were not correlated at all, and this discrepancy, interpreted with the help of mean pressure, heart rate and peripheral vascular resistances, was in favor of INVS1.

💡 Research Summary

The authors present a minimally invasive technique for estimating stroke volume variation (SVV) by extracting a novel parameter, the first systolic invariant (INVS1), from arterial blood pressure (ABP) waveforms. The method relies on semi‑classical signal analysis (SCSA), a mathematical framework that decomposes the ABP signal into a set of eigenfunctions and eigenvalues, allowing a reconstruction of the waveform and the calculation of invariant quantities that characterize its shape and energy. INVS1 is defined as the integral of the first reconstructed systolic component and is hypothesized to be linearly proportional to the actual stroke volume (SV) generated by the left ventricle.

To validate this hypothesis, the study enrolled 21 mechanically ventilated intensive‑care patients who already had a PiCCO™ catheter in place. Over a 15‑minute recording period, simultaneous high‑frequency ABP data and PiCCO‑derived SV measurements were collected. The ABP signals were processed offline with the SCSA algorithm to obtain a time series of INVS1 values. The authors then performed three complementary statistical analyses: (1) cross‑correlation to assess temporal alignment and similarity of the two time series, (2) linear regression to quantify the proportional relationship, and (3) Bland‑Altman analysis to evaluate agreement and interchangeability.

Cross‑correlation yielded a mean coefficient of 0.90 ± 0.04 across 94 % of the 315 one‑minute segments (21 patients × 15 minutes), indicating that INVS1 tracks the PiCCO SV waveform with almost no lag. Linear regression produced an average coefficient of determination (R²) of 0.89, with a slope of 1.02 ± 0.08 and an intercept of 0.4 ± 2.1 ml beat⁻¹, confirming a near‑one‑to‑one relationship. Bland‑Altman analysis showed a mean bias of +1.3 ml beat⁻¹ (standard deviation ± 4.2 ml beat⁻¹) and 95 % limits of agreement from –7.1 to +9.7 ml beat⁻¹, well within clinically acceptable margins. These results collectively demonstrate that INVS1 can reliably substitute PiCCO‑derived SV for the majority of recordings.

In the remaining 6 % of segments, the correlation between INVS1 and PiCCO SV fell below 0.30. Detailed inspection revealed that these outliers coincided with episodes of markedly low mean arterial pressure, abrupt heart‑rate fluctuations, or sudden increases in peripheral vascular resistance. Such hemodynamic instability is known to compromise the thermodilution‑based PiCCO algorithm, whereas the waveform‑based INVS1, being derived directly from the pressure contour, appears more robust under these conditions.

The study acknowledges several limitations. The sample size is modest, and the observation window is limited to 15 minutes, which may not capture longer‑term trends or rare events. All participants were mechanically ventilated, restricting generalizability to spontaneously breathing patients. Moreover, the SCSA algorithm assumes relatively stable rhythm; severe arrhythmias or extreme hypotension could impair signal reconstruction. Future work should test the method across diverse clinical scenarios (e.g., intra‑operative monitoring, septic shock, chronic heart failure) and evaluate real‑time implementation in bedside monitors.

In conclusion, the paper provides compelling evidence that a semi‑classical analysis of arterial pressure waveforms yields a first systolic invariant (INVS1) that is linearly related to stroke volume and agrees closely with the gold‑standard PiCCO technique. By offering a simple, minimally invasive, and potentially more resilient alternative for SVV monitoring, this approach could reduce the need for invasive catheters, streamline hemodynamic assessment, and support faster therapeutic decisions in critical care environments.