Related Experiment Videos
A new method for regularization parameter determination in the inverse problem of electrocardiography
1Research Center, Hôpital du Sacré-Coeur, Montreal, P.Q., Canada.
IEEE Transactions on Bio-Medical Engineering
|January 1, 1997
Summary
This study introduces a simpler zero-crossing method for determining the regularization parameter in electrocardiography inverse problems. The new method performs comparably to existing techniques, especially when measurement noise dominates.
Area of Science:
- Biomedical Engineering
- Computational Electrophysiology
- Medical Imaging
Background:
- The inverse problem of electrocardiography (ECG) aims to compute epicardial potentials from body surface potentials.
- This problem is ill-posed, often requiring regularization techniques like zero-order Tikhonov regularization for stable solutions.
- Determining the appropriate regularization parameter (t) is crucial for accurate inverse solutions.
Purpose of the Study:
- To evaluate existing methods (CRESO, L-curve) for selecting the regularization parameter (t) in ECG inverse problems.
- To propose and validate a novel, computationally simpler zero-crossing method for determining 't'.
- To compare the performance of the proposed method against established techniques under various noise conditions and complexities.
Main Methods:
- Zero-order Tikhonov regularization was applied to the ECG inverse problem.
- The Composite Residual and Smoothing Operator (CRESO) and L-curve methods were used as benchmarks.
- A new zero-crossing method was proposed, selecting 't' where the squared residual norm equals 't' times the squared solution norm.
- Simulations involved concentric spheres and realistic geometries, with dipole sources and varying noise (geometry and measurement).
Main Results:
- The zero-crossing method showed comparable performance to CRESO and L-curve when measurement noise dominated.
- In scenarios with dominant correlated geometry noise, only CRESO successfully identified a suitable 't'.
- Under low measurement noise conditions, none of the evaluated methods consistently yielded optimal solutions.
Conclusions:
- The proposed zero-crossing method offers a simpler alternative for regularization parameter selection in ECG inverse problems.
- The choice of regularization parameter selection method is highly dependent on the noise characteristics of the specific application.
- Further research is needed to optimize regularization parameter selection, particularly under low noise conditions.