A reduced complexity ECG imaging model for regularized inversion optimization.

Area of Science:

• Biomedical Engineering
• Computational Cardiology
• Medical Imaging

Background:

• The inverse problem of electrocardiography is vital for diagnosing and treating cardiac arrhythmias.
• Current cardiac simulation models are often too complex or lack realistic outputs.
• Accurate cardiac electrical behavior simulation is needed for evaluating inversion methods.

Purpose of the Study:

• To design a low-resolution heart-torso model capable of generating realistic cardiac electrical mappings and electrocardiograms (ECGs).
• To evaluate the performance of different Tikhonov-based inversion methods for determining regularization parameters in various cardiac conditions.

Main Methods:

• Developed a simplified heart-torso geometry model using the finite element method and monodomain formalism.
• Implemented a model reduction technique via sensitivity analysis and evolutionary optimization to identify key parameters.
• Compared Tikhonov-based inversion methods using synthesized ECGs against reference data for healthy, ischemic, and ventricular tachycardia scenarios.

Main Results:

• Sensitivity analysis identified 25 influential parameters out of 58, with regional variations in influence.
• Synthesized ECGs achieved an 88% correlation with reference data, preserving characteristic shapes.
• Robust Generalized Cross Validation and Discrepancy Principle methods demonstrated superior performance in inverse problem resolution.

Conclusions:

• The proposed low-resolution heart-torso model effectively simulates realistic cardiac electrical activity for healthy and pathological cases.
• The model aids in evaluating inversion methods, with specific techniques showing high success rates and low errors.
• This simulation tool supports advancements in cardiac arrhythmia diagnosis and catheter-based therapy.