Next, we have Dr. Jermakowicz speaking to us on analysis of thermal damage estimates for the prediction of mesiotemporal laser ablation zone. Thank you. So laser interstitial thermal therapy, or LITS, has become a viable minimally invasive treatment alternative for mesiotemporal epilepsy. Its recent resurgence is really a result of advances in MRI thermometry, which now allow for the monitoring of ablation extent in real time. To the right here, we see, on the right column, we see the thermal damage estimate, or TDE maps, which are essentially Arrhenius-derived estimates of irreversible cell damage that are overlaid on the patient MRI and then used by the surgeon to make intraoperative decisions. These estimates are generally thought of as being very accurate, however, inaccuracies do exist, and that's largely due to the fact that tissue responses to laser energy are variable. However, the mesiotemporal lobe provides a very intriguing opportunity to study LITS because unlike its use for brain tumors, in the case of chronic epilepsy, we use the same anatomic substrates each time, pretty much. So therefore, the aim of this talk was to use data from these TDE maps to help study the temporal progression of laser ablations and also to measure what the impact of variations in human anatomy or variations in the operative procedure are on these dynamics. So the questions we're addressing are, what patient or procedure-specific variables influence intraoperative tissue ablation dynamics, and how do TDEs spatially compare to immediate and six-month-delayed LIT ablations? So what we did is we took all pre-, intra-, and post-operative MRIs and co-registered them using cranial suites from Neurotargeting in Asheville, Tennessee. We then co-registered the TDE maps to all these images using the intraoperative MRIs since these were from the same imaging session. Pixel counts were then measured from TDE videos using custom-built MATLAB programs, and here's an example from one patient where you see different frames of a TDE video, and then to the right you see the pixel counts with the black line showing total pixel counts, the blue and yellow lines showing pixel counts lateral to the laser and mesial to the laser. All counts were fit by first-order linear dynamics, and 30 patients total are used in this study. I used 16 different independent variables to try to account for the variance of the data. Most are self-explanatory. CSF above is CSF volume above the hippocampus, CSF lateral is CSF volume lateral to the hippocampus. Those two and medullohippocampal volume were measured with serial manual tracings. What you see at the bottom there, ML position and SI position, essentially measures the position of the laser relative to the axis of the hippocampal head. So moving on to the results, when we looked at, in this graph here, we are looking at p-values of the multivariate regression statistics relating the thermal ablation time constant from these best-fit exponential curves to the 16 independent variables. What you'll notice is that the only two significant associations we have are for T2 signal and axial laser angle, and more specifically, increased T2 signal and a more mesially pointing laser appear to be correlated with slower ablation. Also, laser ablations appear to be asymmetric. Here on the left, we're looking at axial TDEs, and we're quantifying how fast tissues burn lateral versus medial to the laser, and you'll notice that past one minute, we actually got statistically significant differences, and also interestingly, these differences were correlated with, once again, increased T2 signal and also T1 with contrast signal. When we looked at the sagittal TDEs, we did not find such asymmetry. Next, we looked at the spatial overlap of TDE maps with post-operative ablation zones. The main point to get from this slide is that whereas the TDE maps very closely actually match the immediate ablation volume boundaries, they were much more variable in how they correlated with the six-month delayed post-op ablation boundaries. In this patient, for example, you see the TDEs on the top and then in the yellow tracings on the bottom, and you'll notice that in the immediate phase and in the delayed phase, the yellow trace, the TDEs are very well correlated with these boundaries, with ablation boundaries, with, here you see DICE similarity coefficients measured, but they're generally high, whereas let's look at this patient. In this patient, the immediate zones are correlated with the TDEs, but the six-month ablation zones, not as much. We then looked at this further by averaging, by normalizing all TDE maps and ablation zones into a common reference space and aligning them by the laser trajectory, and so what you're looking at here is for all 30 patients, we're looking at the averages of the TDEs and the averages of the immediate ablation boundaries, and you'll notice, consistent with the previous slide, that the TDEs are similar to or slightly underestimate immediate ablation effects, whereas with the delayed zones, you'll notice that the TDEs generally overestimate the effects, and these effects are particularly great at mesial structures. One thing that's important for me to note here, and I have this data for these sagittal maps, not as exciting, but one thing that's important to note here is we don't know whether these effects are because our estimates are not accurate in the long term or whether tissues have varying degrees of contraction that causes these measures. My hunch is that it's a mixture of both, but interestingly, these discrepancies between the TDE maps and the delayed ablations are correlated with volume of surrounding CSF spaces. So our conclusions are that mesiotemporal responses to thermal energy are diverse, TDE maps are accurate at predicting lit effects, and we believe that by incorporating preoperative imaging data into our models, we may eventually improve our predictions of lit effects. Dr. Schwab? So I think that's a really interesting question. I agree with your assessment as well. It's very hard for me to know if this is just contraction of dead tissue, and you might want to look at this in terms of estimates of the percentage of the hippocampus and amygdala ablated at the various time points and see if that remains constant. The other thing I would suggest as you move forward with this work is you may want to add a couple other variables with your preop variable list to look at, like flare signal, flare intensity over the course of the hippocampus, and other things like fractional anisotropy and MD and other diffusion-related characteristics if you have the data. Yeah. On a tumor data set, I'm actually doing that, and flare is related to ablation dynamics, and as expected, perfusion data also is. John? Maybe an insight, maybe it's not the same, but we've done a similar work with cavernous malformations, and what you see when you ablate a cavernous malformation, which isn't sitting in a ventricle, so then you're not confounded by the change of shape of a structure inside a ventricle, if you ablate a cavernous malformation, the cavernous malformation will shrink over time and you're left with a little zone of encephalomalacia. So suggesting that, indeed, what you're getting is tissue retraction, and when we actually have the opportunity to resect one of those, what you're left with is just a small piece of dead tissue. So maybe it's an insight that does suggest that probably what we're seeing for the most part is just shrinkage of the dead hippocampus tissue or bladed tissue, and it's very hard to match that up to preoperative and immediate postoperative. Yeah, it's really hard in the immediate post-op imaging because with all the swelling, you can't tell what's viable hippocampus or not. I mean, it's... But yeah, glad to hear we have the same concerns. Yeah, I'm naive to this, I have not performed this procedure, but is it possible to then segment out the tissue adjacent to the hippocampus to see if there is migration of that tissue to try to infer if there is migration or ablation? I mean, all these segmentations using the software I'm using become much more difficult after the ablation is there. You have less delineation of the gray-white boundaries, and that's just been difficult for us to do. That's definitely true. Yeah. Thank you.

thermal damage estimates

mesiotemporal laser ablation zone

LITT

TDE maps

preoperative imaging data