During the final year of my Master’s, I was working at Siemens Corporate Technology in Princeton, NJ, USA, on a very interesting project dealing with various aspects of electrophysiology and biomechanics models of the human heart muscle. My time there was one of my most productive phases so far, with not only the successful completion of my MSc thesis titled “Efficient and Robust Patient-Specific Model of the Heart Function based on MRI Images” but also a great publication outcome (see below).
In my work, fast and robust patient-specific parameter estimation for a biomechanic model of the human heart from clinical and imaging data is investigated. Of course, my results are based on available models of heart anatomy and electrophysiology, and – working in a great team at Siemens – I could heavily benefit from extensive experience in heart segmentation and model generation.
My thesis has two major contributions: First, an integrated framework to compute cardiac motion using a finite element setup is presented, in particular including an efficient strategy to parallelize the evaluation of stress and mechanical boundary conditions for high-performance implementations. Second, a novel, data-driven approach to calibrate electrophysiology (EP) parameters from clinically available 12-lead electrocardiograms (ECGs) is introduced, as illustrated in the figure.
Forward workflow to compute ECG parameters from electrophysiology (EP) model derived from segmented myocardium, and backward workflow to estimate EP model parameters from measured ECG features.
And this is what the final result looks like:
Related Publications
2014
A Framework for the Pre-clinical Validation of LBM-EP for the Planning and Guidance of Ventricular Tachycardia Ablation
Tommaso Mansi, Roy Beinart, Oliver Zettinig, Saikiran Rapaka, Bogdan Georgescu, Ali Kamen, Yoav Dori, M. Muz Zviman, Daniel A. Herzka, Henry R. Halperin, and Dorin Comaniciu
In Statistical Atlases and Computational Models of the Heart. Imaging and Modelling Challenges, Jan 2014
This manuscript presents a framework for the pre-clinical validation of LBM-EP, a fast cardiac electrophysiology model based on the lattice-Boltzmann method (LBM). The overarching goal is to assess whether the model is able to predict ventricular tachycardia (VT) induction given lead location and stimulation protocol. First, the random-walk algorithm is used to interactively segment the heart ventricles from delayed-enhancement magnetic resonance images (DE-MRI). Scar and border zone are visually delineated using image thresholding. Then, a detailed anatomical model is generated, comprising fiber architecture and spatial distribution of action potential duration. That information is rasterized to a Cartesian grid, and the cardiac potentials are computed. The framework is illustrated on one swine data, for which two different pacing protocols at four different sites were tested. Each of the protocols were then virtually tested by computing seven seconds of heart beat. Model predictions in terms of VT induction were compared with what was observed in the animal. Our parallel implementation on graphics processing units required a total computation time of about two minutes at an isotropic grid resolution of 0.8 mm (21s at a resolution of 1.5 mm), thus enabling interactive VT testing.
Patent
Method and System for Interactive Computation of Cardiac Electromechanics
Tommaso Mansi, Oliver Zettinig, Bogdan Georgescu, Ali Kamen, Dorin Comaniciu, and Saikiran Rapaka
Data-driven estimation of cardiac electrical diffusivity from 12-lead ECG signals
Oliver Zettinig, Tommaso Mansi, Dominik Neumann, Bogdan Georgescu, Saikiran Rapaka, Philipp Seegerer, Elham Kayvanpour, Farbod Sedaghat-Hamedani, Ali Amr, Jan Haas, Henning Steen, Hugo Katus, Benjamin Meder, Nassir Navab, Ali Kamen, and Dorin Comaniciu
Fast Data-Driven Calibration of a Cardiac Electrophysiology Model from Images and ECG
Oliver Zettinig, Tommaso Mansi, Bogdan Georgescu, Elham Kayvanpour, Farbod Sedaghat-Hamedani, Ali Amr, Jan Haas, Henning Steen, Benjamin Meder, Hugo Katus, Nassir Navab, Ali Kamen, and Dorin Comaniciu
In Medical Image Computing and Computer-Assisted Intervention–MICCAI 2013, Sep 2013
From Medical Images to Fast Computational Models of Heart Electromechanics: An Integrated Framework towards Clinical Use
Oliver Zettinig, Tommaso Mansi, Bogdan Georgescu, Saikiran Rapaka, Ali Kamen, Jan Haas, Karen S Frese, Farbod Sedaghat-Hamedani, Elham Kayvanpour, Ali Amr, Stefan Hardt, Derliz Mereles, Henning Steen, Andreas Keller, Hugo A Katus, Benjamin Meder, Nassir Navab, and Dorin Comaniciu
In Functional Imaging and Modeling of the Heart, Jun 2013
Towards Real-Time Cardiac Electrophysiology Computations Using GP-GPU Lattice-Boltzmann MethodIn MICCAI Workshop on High Performance Computing for Biomedical Image Analysis–HPC-MICCAI, Sep 2013
Fast Data-Driven Calibration of a Cardiac Electrophysiology Model from Images and ECGIn Medical Image Computing and Computer-Assisted Intervention–MICCAI 2013, Sep 2013
