Overview
Recent developments in neuroscience and biomedical engineering face bottlenecks and challenges in the fundamental understanding of mechanical mechanisms, among other phenomena, in brain tissue across different length and time scales, and various loading conditions. Ultrasound Neuromodulation (UNM), where non-invasive low intensity focused ultrasound is utilized for brain research and therapeutic applications, is an example of a crucial area where a further expansion of the technology hinges on the accurate understanding of the coupling between the mechanical wave and the brain material, both at the tissue and neuron level. More generally, advancing the treatments of brain-related diseases and injuries heavily relies on predictive computational models of brain mechanics where it could be adopted in a patient-specific manner. Our lab at Duke aims to expand on our previous work and establish a comprehensive, multi-scale, and predictive model of the mechanical response of the brain for a broad range of therapeutic applications. Below is an example project that showcases our endeavors in this research thrust:
A Data-Driven patient-specific paradigm for prediction of acoustic wave patterns in brain:
Predictive modeling of the mechanical response of the human brain to harmonic stimulation is an important part of many preventive, diagnostic, and therapeutic biomedical procedures. Computational models are increasingly incorporated to inform or guide biomedical research such as transcranial ultrasound stimulation therapies. However, the efficacy of such models depends crucially on precise patient-specific representations of both the geometry and the material response of the brain in vivo. But developing material models for brain matter has proven challenging and remains a major predictive bottleneck. These modeling challenges arise from the ultra-compliant, complex, heterogeneous, and patient-specific nature of the mechanical properties of brain tissue. By way of contrast, recent advances in microscopy and elastography techniques, such as Magnetic Resonance Elastography (MRE), have made possible the in vivo characterization of the viscoelastic response of brain tissue in individual patients, which, in turn, have radically transformed brain biomechanics. Leveraging these advancements, we have developed a novel patient-specific computational paradigm, where we entirely bypass the intermediate material modeling step with recourse to a model-free data-driven approach that directly incorporates in situ personalized spatially-varying viscoelastic data to predict acoustic wave patterns in brain.
In this paradigm, we use anatomical data, in the form of MRI images, in conjunction with a co-registered atlas of viscoelastic properties of the brain tissue (borrowed from Hiscox L. V., et. al., Hum. Brain Mapp. (2020)), to generate a finite-element model of the brain. We then solve in the frequency domain for the steady-state wave pattern induced by, e. g., low-intensity focus ultrasound (LIFUS), applied to a small region on the frontal lobe. We have demonstrated the framework by performing full-scale spatially-resolved calculations based on public domain viscoelastic data from MRE measurements.