Session: 12-03-02: Data-Enabled Predictive Modeling, Scientific Machine Learning, and Uncertainty Quantification in Computational Mechanics

Paper Number: 120156

120156 - Computational Ultrasonic Neuromodulation 

Ultrasound neuromodulation (UNM) has recently received significant attention as a promising tool for neuroscience, where a region in the brain is targeted by focused ultrasound (FUS), which, in turn, causes excitation or inhibition of neural activity. Despite its great potential in neuroscience, several aspects of UNM are still unknown. One class of important questions pertains to the FUS-induced wave patterns in the brain, including the off-target effects of UNM and their dependency on stimulation frequency. Another class of problems concerns the patient-specificity inherent to the human brain, and the effects thereof in predictions. Computational ultrasonic neuromodulation aims to shed light on these issues that imped the progress of UNM procedures to clinical applications. 

 

I will first present some work on understanding the FUS-induced wave patterns, where we pursue a computational analysis of UNM in both humans and rodents with recourse to explicit finite element method, accounting for the intricate geometry and the viscoelastic mechanical behavior of individual tissues. We demonstrate that, upon subjecting a region on the skin above the skull to FUS pressure, the bone acts as a waveguide for ultrasound-induced shear waves, carrying them away from the FUS target. As we demonstrate in our human study, this phenomenon help explain the off-target auditory responses observed during neuromodulation experiments. Our findings could help explain the off-target responses observed during neuromodulation experiments and inform the development of mitigation and sham control strategies. 

 

Next, I will touch upon the patient-specificity of the viscoelastic properties of human brain and its importance in predictive computational modeling in medicine and neuroscience, where a particular outcome for a specific human subject is predicted, rather than a statistical prediction for a population of subjects. While non-invasive imaging techniques, such as MRI, presently allows for accurate representations of human anatomy down to ultra-fine features, the accurate and reliable material modeling of human tissue has proven challenging and remains a major predictive bottleneck. By way of contrast, recent advances in microscopy and elastography techniques, such as Magnetic Resonance Elastography (MRE), enables, for the first time, the in vivo characterization of the viscoelastic response, generating personalized atlases of viscoelastic properties. We present a class of model-free Data-Driven solvers that effectively enable the utilization of in situ and in vivo imaging data directly in full-scale calculations of the mechanical response of the human brain to ultrasound stimulation, entirely bypassing the need for analytical modeling or regression of the data. We demonstrate its application in UNM and its potential in addressing the challenges in designing personalized UNM procedures. 
 

Presenting Author: (Amir)Hossein Salahshoor Duke University

Presenting Author Biography: Dr. (Amir)Hossein Salahshoor is an Assistant Professor in Duke University since August 2023. Prior to that, he was a postdoctoral scholar at California Institute of Technology in the department of Aerospace Engineering. (Amir)Hossein also received his Ph.D. in Aerospace Engineering from Georgia Institute of Technology in 2018. During his Ph.D., he also received a master’s degree in Mathematics. His research interests lie broadly in mechanics and its intersection with material science, biology, and data science.

Authors:

(Amir)Hossein Salahshoor Duke University