Trans-stenotic pressure gradient estimation using a modified Bernoulli equation

This study introduces and validates a modified Bernoulli equation that incorporates Reynolds-number-dependent loss coefficients to more accurately estimate trans-stenotic pressure gradients across varying flow regimes compared to traditional simplified and extended formulations, while also demonstrating that peak-velocity-based estimations are more robust to MRI pixel size limitations than bulk flow rate measurements.

The Gist

Imagine your bloodstream as a busy highway, and a stenosis (a narrowing in an artery) as a sudden, massive construction zone where the road shrinks from four lanes down to one.

When cars (blood cells) rush through this narrow gap, they speed up. But just like traffic jams, this speeding causes friction and chaos, leading to a drop in "pressure" (the force pushing the cars forward). Doctors need to measure this pressure drop to decide if the blockage is dangerous enough to require surgery.

Currently, doctors use a "shortcut" formula (the Simplified Bernoulli equation) to guess this pressure drop just by looking at how fast the cars are going at the narrowest point. This paper argues that this shortcut is often wrong, especially when the traffic is heavy and chaotic. Instead, the authors propose a smarter, updated formula and show how the quality of the "traffic camera" (MRI) changes the accuracy of the guess.

Here is the breakdown of their findings using simple analogies:

1. The Problem with the Old Shortcut

The old formula assumes that the relationship between speed and pressure drop is always the same, like a fixed rule: "If speed doubles, pressure drops quadruple."

• The Reality: It's more like driving through a tunnel.
  • Low Traffic (Laminar Flow): When cars move slowly and orderly, the friction is mostly just the tires rubbing against the road.
  • High Traffic (Turbulent Flow): When cars speed up, they start swerving, crashing into each other, and creating a chaotic mess. This "turbulence" creates extra friction that the old formula ignores.
• The Result: The old formula often overestimates the danger. It tells the doctor, "This blockage is a disaster!" when it's actually just a moderate annoyance, potentially leading to unnecessary surgery.

2. The New "Smart" Formula (Modified Bernoulli)

The authors created a Modified Bernoulli (MB) equation. Think of this as a GPS that doesn't just look at speed, but also checks the Reynolds Number (a fancy way of saying "how chaotic the flow is").

• How it works: It adds a "chaos factor" to the calculation.
  • If the flow is smooth, it calculates friction one way.
  • If the flow is turbulent (high speed), it adds a penalty for the extra energy lost to the chaos.
• The Outcome: In their lab tests (using a clear plastic pipe and water), this new formula was spot on, predicting the pressure drop within about 10% of the real measurement. The old formula was off by as much as 55%.

3. The "Camera Resolution" Problem (The MRI Issue)

To use these formulas, doctors need to measure the speed of the blood. They often use MRI, which is like taking a photo of the traffic.

• The Pixel Problem: An MRI image is made of tiny squares called pixels.
  • High Resolution (Small Pixels): Imagine a photo with millions of tiny dots. You can clearly see the fast cars in the middle and the slow cars near the walls.
  • Low Resolution (Large Pixels): Imagine a photo with only a few giant, blocky squares.
• The "Blur" Effect: If the squares are too big, a single square might cover both a fast car in the center and a slow car near the wall. The camera averages them out, making the fast car look slower than it really is. This is called Partial Volume Effect.
• The Consequence:
  • If you use the average speed (bulk velocity) from a blurry image, you severely underestimate how fast the blood is actually moving.
  • Because the new "Smart Formula" relies on this speed, a blurry image leads to a huge underestimation of the pressure drop (thinking the blockage is safe when it's actually dangerous).
  • The Fix: The authors found that if you measure the peak speed (the fastest car in the very center) instead of the average, the "blur" matters much less. The center of the stream is so fast that even a blurry square can't hide it.

4. The Big Takeaway for Patients

This research offers two major lessons for the future of heart care:

1. Stop using the old "one-size-fits-all" math. We need to use the new "Smart Formula" that accounts for whether the blood flow is smooth or chaotic. This will stop doctors from over-diagnosing blockages.
2. Don't trust a blurry picture. If an MRI scan doesn't have enough detail (pixels) to clearly see the narrowest part of the artery, the pressure calculation will be wrong.
   • Recommendation: Doctors need to ensure the MRI scan is "high definition" enough to capture the narrowest point clearly. Alternatively, they should focus on measuring the peak speed in the center of the flow, which is more forgiving of lower-quality images.

In summary: The authors built a better calculator for heart blockages that understands the difference between smooth driving and traffic chaos. They also warned that if your "traffic camera" (MRI) isn't sharp enough, even the best calculator will give you the wrong answer—unless you focus on the fastest car in the middle of the pack.

Technical Summary

1. Problem Statement

Accurate non-invasive estimation of trans-stenotic pressure gradients is critical for assessing the severity of arterial stenosis and guiding therapeutic decisions (e.g., valve intervention). Current clinical practice relies heavily on the Simplified Bernoulli (SB) equation, which estimates pressure drop ($\Delta p$) solely from peak velocity ($V_{peak}$).

• Limitations of Existing Models:
  • SB: Assumes inviscid flow and neglects pressure recovery and flow regime changes. It often overestimates pressure drops, particularly as flow rates increase.
  • Extended Bernoulli (EB): Incorporates geometric area ratios to account for pressure recovery but still fails to explicitly model the transition from viscous-dominated (laminar) to inertia-dominated (turbulent) losses.
  • MRI Limitations: When using Phase-Contrast MRI (PC-MRI) for velocity inputs, partial volume effects (PVE) caused by coarse in-plane pixel sizes lead to systematic underestimation of bulk velocities, which propagates into significant errors in pressure drop calculations.
• Gap: There is a lack of a unified model that accounts for Reynolds-number-dependent loss mechanisms (viscous vs. turbulent) and explicitly quantifies the impact of MRI spatial resolution on pressure estimation accuracy.

2. Methodology

The study utilized a combined experimental and analytical approach using an idealized stenosis model.

• Experimental Setup:

  • Model: A scaled-up FDA benchmark nozzle (PMMA) representing a canonical stenosis with a throat diameter of 6.6 mm and inlet diameter of 20 mm.
  • Fluid: Water at 22°C (Newtonian surrogate).
  • Flow Conditions: Steady flow rates ranging from 0.65 to 3.9 L/min, covering Reynolds numbers ($Re$) from ~1,000 to 4,000 (spanning laminar, transitional, and turbulent regimes).
  • Measurements:
    • Direct Pressure: Differential pressure sensors ($P_1, P_2$) provided ground-truth pressure drops.
    • Velocity (UIV): Ultrasound Imaging Velocimetry provided high-resolution velocity fields to determine bulk and peak velocities.
    • Velocity (PC-MRI): 3T MRI was used to simulate clinical imaging. Data was acquired at a high resolution (0.13 mm/px) and retrospectively downsampled to coarser resolutions (0.25, 0.50, 1.00, 1.33 mm/px) to study PVE.

• Analytical Framework:

  • Modified Bernoulli (MB) Equation: The authors derived a formulation where the pressure-loss coefficient ($K_{MB}$) is a function of the Reynolds number:
    $$\Delta p_{MB} = \frac{1}{2}\rho V_{bulk}^2 \left( \frac{k_v}{Re} + k_t \left(1 - \frac{A_A}{A_t}\right)^2 \right)$$
    Where $k_v$ represents viscous losses and $k_t$ represents turbulent expansion losses.
  • Calibration: Coefficients $k_v$ and $k_t$ were fitted to the experimental pressure drop data.
  • Comparison: The MB model was benchmarked against the SB and EB equations using both UIV and PC-MRI velocity inputs.

3. Key Contributions

1. Development of a Regime-Dependent MB Model: Introduced a pressure drop formulation that explicitly separates viscous and turbulent loss terms, making the loss coefficient dependent on the flow regime ($Re$) rather than a constant.
2. Quantification of MRI Pixel Size Effects: Systematically evaluated how in-plane pixel size affects velocity measurements and subsequent pressure drop estimations, providing a "pixel-size–error map."
3. Dual-Formulation Strategy: Demonstrated that the MB model can be effectively applied using either bulk flow rate (derived from MRI integration) or peak velocity (derived from Doppler or MRI peak detection), with distinct sensitivity profiles to pixel size.
4. Validation of a Fixed Velocity Ratio ($C$): Showed that a constant ratio between peak and bulk velocity ($C \approx 1.42$) is sufficient for accurate MB predictions across a wide range of flow conditions, simplifying clinical application.

4. Key Results

• Model Performance (Pressure Drop Estimation):

  • MB Model: Achieved the highest accuracy, agreeing with experimental measurements within ±10% (typically ~5% error) across the entire flow regime.
  • SB Model: Systematically overestimated pressure drops by 10–55%, with errors increasing significantly at higher Reynolds numbers (turbulent flow).
  • EB Model: Showed intermediate performance, underestimating at low $Re$ and overestimating at high $Re$, with errors ranging from -15% to +25%.
  • Clinical Relevance: In the clinically relevant range ($\Delta p = 10-15$ mmHg), the MB model maintained errors <5%, whereas SB and EB showed maximum errors of +55% and +25%, respectively.

• Impact of MRI Pixel Size:

  • Bulk Velocity ($V_{bulk}$): Highly sensitive to pixel size. Coarse sampling (3–4 pixels across the throat radius) led to systematic underestimation of bulk velocity by -34% to -44%. Consequently, MB pressure drops calculated from flow rate were underestimated by -52% to -62%.
  • Peak Velocity ($V_{peak}$): Significantly less sensitive to pixel size. Even at coarse resolutions, peak velocity errors remained small (-13% to -18.7%).
  • MB Pressure Drop Sensitivity:
    • When using Flow Rate inputs: Pressure drop errors mirrored bulk velocity errors (up to -65% underestimation at 1.33 mm/px).
    • When using Peak Velocity inputs: Pressure drop errors were substantially reduced, remaining within -3% to -18.7% even at coarse resolutions.

• Sampling Requirements: To keep pressure drop estimation errors within 10%, the study suggests a minimum of 15–20 pixels across the stenosis lumen (approx. 1/10th of the throat radius).

5. Significance and Clinical Implications

• Improved Diagnostic Accuracy: The MB model offers a more physiologically accurate method for estimating stenosis severity, potentially preventing misclassification of patients (e.g., avoiding under-treatment of severe stenosis due to SB overestimation or underestimation due to poor MRI resolution).
• MRI Protocol Optimization: The study provides concrete guidelines for MRI protocol design. It highlights that while peak velocity-based MB estimations are robust to lower resolutions, flow-rate-based estimations require high-resolution imaging to avoid massive underestimation of pressure gradients.
• Pathway to Clinical Translation: The authors propose a "library" approach where pre-calibrated coefficients ($k_v, k_t, C$) are mapped to specific lesion geometries (eccentricity, length, contraction ratio). This would allow clinicians to apply the MB model to patient-specific imaging without needing invasive calibration for every case.
• Limitations & Future Work: The study used a rigid, idealized model with steady flow. Future work must address arterial elasticity (compliance) and pulsatile flow effects, though previous literature suggests steady-flow approximations are valid for peak-systolic pressure gradients in severe stenosis.

In conclusion, this paper bridges the gap between fluid dynamics theory and clinical imaging by introducing a Reynolds-dependent pressure drop model and quantifying the critical role of spatial resolution in MRI-based hemodynamic assessment.