Vito – A Generic Agent for Multi-physics Model Personalization: Application to Heart Modeling
Precise estimation of computational physiological model parameters from patient data is one of the main hurdles towards their clinical applicability. Designing robust estimation algorithms is often a tedious and model-specific process. We propose to use, for the first time to our knowledge, artificial intelligence (AI) concepts to learn how to personalize a computational model, inspired by how an expert manually personalizes. We reformulate the parameter estimation problem in terms of Markov decision process and reinforcement learning. In an off-line phase, the artificial agent, called Vito, automatically learns a representative state-action-state model through data-driven exploration of the computational model under consideration. In other words, Vito learns how the model behaves under change of parameters and how to personalize it. Vito then controls the on-line personalization by exploiting its automatically derived action policy. Because the algorithm is model-independent, personalizing a completely new model would require only adjusting some simple parameters of the agent and defining the observations to match, without the full knowledge of the model itself. Vito was evaluated on two challenging problems: the inverse problem of cardiac electrophysiology and the personalization of a lumped-parameter whole-body circulation model. Obtained results suggested that Vito could achieve equivalent goodness of fit than standard methods, while being more robust (up to 25% higher success rates) and with faster (up to three times) convergence rate. Our AI approach could thus make model personalization algorithms generalizable and self-adaptable to any patient, like a human operator.
Unable to display preview. Download preview PDF.
- 2.Nordsletten, D., Niederer, S., Nash, M., Hunter, P., Smith, N.: Coupling multi-physics models to cardiac mechanics. PBMB 104(1), 77–88 (2011)Google Scholar
- 4.Wong, K.C., Sermesant, M., Rhode, K., Ginks, M., Rinaldi, C.A., Razavi, R., Delingette, H., Ayache, N.: Velocity-based cardiac contractility personalization from images using derivative-free optimization. J. Mech. Behav. Biomed. 43 (2015)Google Scholar
- 5.Seegerer, P., et al.: Estimation of regional electrical properties of the heart from 12-lead ECG and images. In: Camara, O., Mansi, T., Pop, M., Rhode, K., Sermesant, M., Young, A. (eds.) STACOM 2014. LNCS, vol. 8896, pp. 204–212. Springer, Heidelberg (2015)Google Scholar
- 6.Sutton, R., Barto, A.: Reinforcement Learning: An Introduction. MIT Press (1998)Google Scholar
- 7.Kveton, B., Theocharous, G.: Kernel-based reinforcement learning on representative states. In: Association for the Advancement of Artificial Intelligence (2012)Google Scholar
- 9.Neumann, D., et al.: Robust image-based estimation of cardiac tissue parameters and their uncertainty from noisy data. In: Golland, P., Hata, N., Barillot, C., Hornegger, J., Howe, R. (eds.) MICCAI 2014, Part II. LNCS, vol. 8674, pp. 9–16. Springer, Heidelberg (2014)Google Scholar
- 11.Zettinig, O., Mansi, T., Neumann, D., Georgescu, B., Rapaka, S., Seegerer, P., Kayvanpour, E., Sedaghat-Hamedani, F., Amr, A., Haas, J., Steen, H., Meder, B., Navab, N., Kamen, A., Comaniciu, D.: Data-driven estimation of cardiac electrical diffusivity from 12-lead ECG signals. Med. Image Anal. 18(8), 1361–1376 (2014)CrossRefGoogle Scholar
- 12.Itu, L., Sharma, P., Georgescu, B., Kamen, A., Suciu, C., Comaniciu, D.: Model based non-invasive estimation of PV loop from echocardiography. In: Engineering in Medicine and Biology Society (EMBC), pp. 6774–6777 (2014)Google Scholar