simtk.org

Please post your questions and comments about our webinar In silico Design of Novel Surgical Methods for Children with Single Ventricle Hearts: From Computation to Clinic here, also see webinar announcement at:

http://wiki.simtk.org/cpms/Events/20150 ... calMethods

Best,

Ahmet

We had a number of good questions from participants which we were unable to address during the webinar. I am posting them here in the hopes that we will continue this discussion both about the research and the process of translating the results to the clinic.

Joy


Questions

1. I'll skip listing the many things that I liked in your presentation & skip to a general M&S methods question. I understand that, early on, you elected to draw on an existing "hammer" (model). Each of the specific questions that you posed (use cases) seemed to be coarse grain, yet your simulations are very fine grain. In making them fine grain and continuous you had to abstract away a variety of multilevel uncertainties. I infer that because of limits on wet-lab data, many fine grain, individualized, simulation details cannot be validated. In retrospect, for such exploratory simulation, why not pursue a more parsimonious approach, consistent with requirements, that employs more flexible, coarse grain simulation mechanisms that can be made finer grain when needed?
2. Do you think a mechanistic validation is necessary if you can illustrate clinical usefulness?
3. You mentioned multiscale modeling. What exactly do mean by multiscale? What scales are you including in the simulations?
4. Could you please explain how to compute oxygen level?
5. What simulation software package did you use?

Thank you for your interest in the webinar. I've done my best to answer everyone's questions below.

1. Thank you for the positive feedback on the presentation. This is an excellent question. Given the high computational cost associated with running CFD simulations, it is often beneficial to use lower order models when possible. We sometimes use lumped parameter (circuit) models for this purpose since they can be run very fast on a desktop computer and provide global hemodynamics information. Of course, the drawback to LPN models is that they do not provide any local hemodynamics information. The use of fine grain vs. coarse grain models really depends on the application. For example, if one is interested in a global view on cardiac work load or changes in PV loops in a single ventricle patient, an LPN model alone is often sufficient. But when you get into clinical questions related to flow distribution or thrombotic risk, you often need the hemodynamic details to get a full enough picture.

2. This is also an excellent question. I think it's important to perform validation studies to ensure that models have been implemented correctly. However, I do not believe that validation will necessarily ensure clinical relevance, and it is possible to demonstrate clinical relevance or perform correlations with clinical outcomes data in absence of validation. Validation in in vitro models is necessary, but not sufficient for demonstrating that in vivo flow conditions have been properly reproduced. Given the difficulties of validating against patient specific in vivo data, I believe it is important that we proceed with high quality studies that directly correlate model predictions with clinical outcomes.

3. Great question. In some sense, one could argue that our models are not truly multiscale. However, I use this term to mean that we simulate local hemodynamics in a 3D CFD model, and we simulate the rest of the circulation in a 0D model. In this reduced order 0D model, we simulate global flow and pressure waveforms without "fine scale" hemodynamics. In the 3D model, we compute fine scale detailed hemodynamics, including quantities like wall shear stress and residence times that are not available from the 0D model. In addition, the 0D model can be viewed as an "organ" level model, with blocks representing the heart and other organs.

4. Yes, we have described in detail our oxygen delivery computation, which does require several assumptions, in our publications. See for example: Esmaily Moghadam, M., Migliavacca, F., Vignon-Clementel, I.E., Hsia, T.-Y., and Marsden, A.L., “Optimization of shunt placement for the Norwood surgery using multi-domain modeling,” J. Biomech. Eng., 134(5), (2012).

5. Indeed, I forgot to mention this in the presentation but would like to make sure people are aware of available software resources. We are using the SimVascular open source package, available for download on simtk.org, with documentation provided at simvascular.org. We have an active NSF-funded project to support software development and we would strongly encourage interested users to join the project by registering on simtk.org. Binary executables are currently available on all three major platforms (windows, linux, mac) and we plan to provide a source code repository within the next couple weeks.

The question that i wanted to raise here is about the need for EHR in this. Do you think that using an ehr system in this environment would be profitable and useful or would it be a burden in the workflow?







---------------------------
Corbin Gravely
EHR Software
---------------------------