TL;DR
This paper proposes a novel computer vision approach to estimate the durability of bioprosthetic heart valves (BHVs) by analyzing the visual symmetry of their leaflet motion. By correlating visual symmetry with simulated functional flow patterns, the authors demonstrate a promising pathway towards non-invasive, CV-based durability assessment, potentially leading to improved patient outcomes.
What Problem Does This Paper Address?
Bioprosthetic heart valves (BHVs) are crucial medical devices, but a significant challenge is their limited durability, with over 50% failing within 15 years, often due to structural tears caused by localized mechanical stress. While in-vitro testing and numerical simulations are used to assess BHV performance, directly predicting long-term durability remains difficult. This paper tackles this by investigating whether the visual symmetry of how BHV leaflets move during pulsatile flow tests can serve as a proxy for their functional symmetry and, consequently, their susceptibility to failure. The core idea is that asymmetric leaflet motion might lead to uneven stress distribution, accelerating wear and tear.
Methods
The authors propose a two-stream approach:
- Computer Vision (CV) Stream:
- Leaflet Curve Extraction: High-speed videos of BHVs in a pulsatile flow simulator are used. The valve orifice region is segmented, and its contour is then split into three individual leaflet curves based on identified anchor points and the valve center.
- Symmetry Scoring: Two novel pair-wise leaflet symmetry scores are introduced:
- DOPM-Based Symmetry: This method, adapted from contour matching techniques, uses edge fragments and their curvature to quantify symmetry. It's designed to be affine-invariant.
- DTW-Based Symmetry: Dynamic Time Warping (DTW), a sequence alignment algorithm, is adapted to measure the reflective symmetry between two leaflet curves. The leaflet curves are first reflected across their symmetry axis, and then DTW calculates the optimal alignment path to minimize the distance between corresponding points. A lower DTW score indicates higher symmetry.
- Numerical Modeling Stream:
- Flow Pattern Simulation: A computational fluid dynamics (CFD) model (using the MPARS solver) simulates the cardiac cycle. This model takes pressure and velocity data from the in-vitro tests as boundary conditions and uses the valve orifice geometry from the videos.
- Functional Symmetry Metrics: The simulations output velocity and vorticity values at the leaflet borders. These are used as indicators of functional symmetry, representing flow patterns and potential shear stresses.
The two streams converge by comparing the calculated visual symmetry scores (DOPM and DTW) with the simulated functional symmetry metrics (velocity and vorticity differences between leaflet pairs).
Datasets
The paper introduces a new dataset assembled from their own experimental setup. This dataset consists of high-quality videos and associated pressure/velocity data from pulsatile flow tests of three different tricuspid BHVs (two 25mm diameter, one 19mm diameter). Specifically, four experimental cases were used: "BHV-1", "BHV-2-50" (50 million cycles), "BHV-2-200" (200 million cycles), and "BHV-3". The videos were captured using a Photron SA3 high-speed digital camera at frame rates of 250-1000 fps.
Key Insights and Ideas
- Visual Symmetry as a Durability Predictor: The central hypothesis that visual leaflet motion symmetry correlates with functional symmetry and, by extension, durability, is strongly supported.
- DTW's Superiority: Dynamic Time Warping emerges as a more effective measure of visual symmetry for BHV leaflets compared to the DOPM-based approach. This is attributed to DTW's global nature, which better captures overall leaflet motion patterns than DOPM's local fragment-based analysis.
- Identifying Asymmetric Leaflets: The DTW scores can help pinpoint which specific leaflet pair exhibits the most asymmetry, offering insights into potential failure points.
- Bridging CV and CFD: The paper successfully links computer vision analysis of visual motion with fluid dynamics simulations, demonstrating how these disparate fields can be integrated for a more comprehensive understanding of BHV performance.
Key Results
- Qualitative Correlation: Visual inspection of the plots in Figure 8 shows that DTW symmetry score profiles generally follow the trends of simulated velocity and vorticity profiles more closely than DOPM scores.
- Quantitative Comparison (RMSE): Table 1 presents Root Mean Square Error (RMSE) values comparing symmetry scores to simulated flow parameters. DTW consistently yielded lower RMSE values than DOPM across most comparisons, indicating better coherence. The lowest RMSE achieved was 0.0899.
- Leaflet-Specific Insights: For BHV-1, the A&C leaflet pair showed higher symmetry (lower DTW scores) and was identified as more visually symmetric. In BHV-2-50, the B&C pair exhibited higher symmetry.
- Valve State and Symmetry: Figure 9 suggests that symmetry scores tend to be lower (more symmetric) when the valve is fully open and higher (less symmetric) during the opening and closing phases.
Limitations and Future Work
- Limited Dataset: The authors acknowledge that their dataset, while diverse in terms of BHV types and test sessions, is still limited. They plan to expand it to include more BHV designs.
- Unified Symmetry Metric: Developing a single, unified symmetry measurement parameter that can easily compare valves with varying video characteristics (frame rate, resolution) is a future goal.
- Real-World Durability Data: The current validation relies on simulated functional symmetry. Future work will focus on gathering actual, real-world durability data to directly correlate visual symmetry with failure modes.
- Deep Learning for Prediction: The authors propose using DTW scores and/or a new unified symmetry parameter to train deep learning networks for direct BHV design durability estimation based on visual symmetry.
- Simplification in Simulation: The numerical model uses a 2D simulation of the valve orifice, which, while expected to preserve overall flow symmetries, might affect the precise flow patterns.
Summary updated