TL;DR
This paper reviews and categorizes various computational techniques for automatically determining when an electrical impulse activates a specific point in the heart (Local Activation Time, LAT) and how fast and in what direction that impulse travels (Conduction Velocity, CV). It provides a valuable resource for researchers and clinicians by assessing the suitability of these methods for different types of cardiac mapping data, aiming to improve the diagnosis and treatment of cardiac arrhythmias.
What Problem Does This Paper Address?
Understanding how electrical signals propagate through the heart is crucial for diagnosing and treating arrhythmias, which are often caused by disruptions in this electrical activity. Cardiac conduction velocity (CV) and local activation times (LATs) are key metrics that reveal the health of the heart's electrical system. However, accurately measuring these parameters is challenging due to the complexity and noise in cardiac electrograms, the sparsity of data in clinical settings, and the intricate nature of wavefront propagation. Existing methods often require manual intervention, are computationally expensive, or are limited to specific data types. This paper addresses the need for a comprehensive overview and comparative analysis of automated techniques that can reliably estimate LATs and CVs across various cardiac mapping modalities.
Methods
The paper surveys a wide range of computational techniques for LAT and CV estimation, broadly categorizing them:
- LAT Estimation Techniques:
- Morphological Approaches: These methods analyze the shape of the electrogram signal, often by identifying specific points like zero-crossings or areas that divide the signal's modulus.
- Non-linear Energy Operator (NLEO): This technique measures signal energy to identify active regions and activation times.
- Time-Delay Cross-Correlation: This approach uses the time difference between signals from nearby electrodes, which is more robust than absolute timing, especially for noisy or fractionated signals.
- Wavelet Decomposition: This method uses wavelets to decompose signals, helping to separate noise and identify activation onset.
- Deconvolution: Framing electrogram generation as a convolution, this technique attempts to reverse the process to find localized activation.
- Template Matching: This involves comparing segments of an electrogram to a library of known morphologies.
- Multi-signal Spatial Methods: These leverage spatial voltage gradients and surface Laplacians for improved LAT estimation.
- Wavefront-Tracking Methods: While not strictly LAT estimators, these methods track wavefronts, which can indirectly inform LAT.
- CV Estimation Techniques:
- Triangulation: Uses sets of three non-collinear electrodes to estimate velocity vectors, suitable for clinical data with arbitrary point arrangements.
- Finite Difference Techniques: Approximates derivatives of activation times on a grid to estimate velocity, best suited for regularly spaced data like optical mapping.
- Polynomial Surface Fitting: Fits polynomial surfaces to space-time coordinates of activation, offering robustness to noise and arbitrary point arrangements.
- Cosine-Fit Techniques: Analyzes activation times across a circular arrangement of electrodes, useful for single-catheter analysis.
- Vector Loops and Ensembles: Infers activation direction from the relative amplitudes of bipolar electrograms from orthogonal electrode pairs.
- Radial Basis Function Interpolation: Interpolates LATs across surfaces to detect wavefronts and estimate CV.
- Isopotential Lines: Estimates CV by tracking the distance traveled by an isopotential line over a fixed time.
- Arbitrary Scalar Fields: Generalizes the isopotential line method using spatial gradients of various scalar fields.
- Time Delays: Estimates propagation velocity by analyzing time differences between electrodes in a local neighborhood.
- Analytic Expressions: Derives expressions for wavefront curvature, speed, and direction from a fixed stencil of points.
- Maximum Likelihood Estimation: A statistical approach that estimates time delays and angles of incidence, tolerant of LAT measurement errors.
The paper also discusses the importance of spatial resolution and the challenges of three-dimensional propagation.
Datasets
The paper discusses the application of these techniques to various datasets, including:
- Simulated electrograms: Used for testing and validating algorithms when ground truth is known.
- Laboratory recordings: From micro-electrode arrays and optical mapping of cardiac tissue (often using voltage-sensitive dyes).
- Clinical electrograms: From unipolar and bipolar extracellular recordings, often acquired using electroanatomic mapping systems.
The paper does not introduce any new datasets or benchmarks but rather reviews existing methods applied to these common data modalities.
Key Insights and Ideas
- No Single "Best" Method: The choice of the most appropriate technique for LAT and CV estimation is highly dependent on the data modality (electrogram vs. optical), spatial distribution of recording points, and the complexity of the wavefronts.
- Clinical vs. Lab Data: Clinical electrograms are often noisier and have lower spatial resolution than laboratory optical mapping data, necessitating different algorithmic approaches.
- LAT is the Bottleneck: Accurate LAT annotation is a critical prerequisite for reliable CV estimation, and many of the challenges in CV estimation stem from difficulties in precisely determining LATs, especially in fractionated or noisy signals.
- Triangulation and Cosine-Fit for Clinical Use: For clinical applications where data points may be irregularly distributed, triangulation and cosine-fit techniques are highlighted for their adaptability.
- Finite Difference and Polynomial Fitting for High-Resolution Data: These methods are well-suited for regularly sampled data, such as from optical mapping, but require careful handling of noise and missing data.
- The Importance of 3D: Cardiac propagation is inherently three-dimensional, and most current techniques operate on 2D surfaces, potentially leading to overestimation of CV for transmural propagation.
Key Results
The paper provides a comprehensive review and comparative assessment of numerous techniques. While it doesn't present new experimental results, it synthesizes findings from various studies:
- Morphological approaches are found to outperform traditional peak/slope methods for complex signals.
- Time-delay cross-correlation offers more precise LAT measurements than absolute timing methods.
- Triangulation has been successfully applied in clinical studies for global activation mapping.
- Finite difference techniques are effective for optical mapping data but can be sensitive to noise.
- Polynomial surface fitting demonstrates robustness to noise and missing data.
- The paper highlights that conduction velocity is a valuable metric that correlates with structural factors like fibrosis and can be used to infer properties of the myocardial substrate.
Limitations and Future Work
- Three-Dimensionality: A significant limitation is that most techniques operate on 2D data, while cardiac propagation is 3D. Future work should focus on robust 3D CV estimation.
- Real-time Mapping: The paper notes that current clinical systems often lack real-time CV mapping capabilities, which would greatly aid diagnosis and treatment during procedures.
- Uncertainty Quantification: Estimating the level of uncertainty in CV calculations, especially from 2D surface measurements of 3D phenomena, remains an open challenge.
- Atrial Fibrillation: Mapping CV during atrial fibrillation is still challenging but crucial for understanding its mechanisms.
- Complex Fractionated Signals: Identifying local activations in complex, fractionated signals remains a significant hurdle, requiring further advancements in electrogram decomposition.
- Integration with Imaging: Further integration of CV data with structural imaging (like LGE-MRI) is an active area of research.
Summary updated