Next Steps

What we will be doing.

immediate goals

  • Use post-processing to and machine learning to quickly generate robust tip trajectories
  • Measure tip-tip trajectories and calculate the corresponding birth-death rates
  • Measure how a change of geometry affects birth-death rates
  • Measure how a change of physiology affects birth-death rates
  • Begin searching for methods to improve targeted ablation surgeries<

Posted by Timothy Tyree on July 17, 2019



well aligned with the goals of the QIB training grant

Our research is well matched with the NIH funded Quantitative Integrative Biology training program’s goal of supporting the study of quantitative physiology, since it uses computational and analytical methods to study the dynamics of atrial fibrillations experimentally observed in human hearts. More specifically, We are in the process of numerically calculating rate constants for master equations relevant to the dissipation of spiral waves driving atrial fibrillation. This will be done ideally in a geometry close to that of a human heart. There will also be analytical efforts to use field theory to collapse the 2+1 D field-field interactions to 2+1 D particle interactions, which would drastically decrease the necessary computation time for a given geometry. Using the same massively parallelized approach we are currently using, the resulting simulations should be good candidates for real-time patient-specific modeling. Since different atria take different shapes, our analytical effort might soon involve the use of an unsupervised machine learning algorithms that could also be used on individual patients or on hearts being prepared for transplant.

Posted by Timothy Tyree on July 12, 2019

Posted by Timothy Tyree on July 17, 2019



on coming features for the website

  • What is computationally nontrivial is taking low-dimensional measurements from these massively parallelized algorithms without slowing them down. High time resolution measurements still slow them down. Although after optimizing the simulations, we can still record data fast enough for research purposes– we can do better. To do better, I will remove the need for several buffering and rendering steps by unifying GPU and CPU memory through CUDA. This should make the data recording process slow down the simulation negligibly.
  • Another coming feature is improved spiral tip locating and extraction of the chirality of the spiral tips. I plan to do this using a convolutional neural network with a tensorflow 2.0 backend.

    • Posted by Timothy Tyree on July 17, 2019

    from the literature

    ”To date, most numerical studies of cardiac wave propagation have focused on simplified geometries including two-dimensional sheets and threedimensional slabs of tissue [...]. For these simplified geometries, it is straightforward to implement no flux boundary conditions, even when including tissue anisotropy. For example, in a slab of tissue, anisotropy can be incorporated easily when the tissue fibers are parallel to faces of the slab.In contrast, hearts have much more complicated geometries that include curved boundaries and complicated fiber orientations.”

    Taken from Fenton F., Cherry E., Karma A., Rappel W., Chaos. (15), 013502 (2005).

    further steps for the long term

  • Making the software practical for use in a surgical setting
  • Making robust extensions to ventricular fibrillation, the number one cause of sudden cardiac death
  • Making real time predictions useful to surgeons conducting cardiac ablation
  • Making real time prediction useful to internal medicine physicians perscribing antirhythmics

    • room for novel results for theoretical physics

  • For future simulations of ventricular fibrillation, exporting the filament lengths of scroll waves. Filament lengths may be used to study the effect of a string tension, or an energy per unit length, μ > 0. Such an interface may lend itself to representing scroll wave filaments as fermionic Jordan-Wigner strings, which can potentially see benefits from some of the results of string theory.
  • Studying quantum field theory on the geometry of a human heart is novel from a theorist’s perspective because most quantum field theories generate their equations of motion by integration by parts. Since the domain of such theories are typically assumed to be spatially extended, the relevant fields are taken to vanish at infinity, effectively setting the boundary terms resulting from integration by parts to zero. On the topologically nontrivial compact domain of the human heart, however, these boundary terms do not necessarily vanish and may have nontrivial or even significant effects on the equations of motion that govern the dynamics of spiral tips during atrial fibrillation, for example.