Selective Photobiomodulation for Emotion Regulation: A Model-Based Dosimetry Study
The transcranial photobiomodulation (t-PBM) technique is a promising approach for the treatment of a wide range of neuropsychiatric disorders, including disorders characterized by poor regulation of emotion such as major depressive disorder (MDD). In this study, we examined various approaches to deliver red and near-infrared light to the dorsolateral (dlPFC) and ventromedial prefrontal (vmPFC) cortex in the human brain, both of which have shown strong relevance to the treatment of MDD. We applied our hardware-accelerated Monte Carlo simulations to systematically investigate the light penetration profiles using a standard adult brain atlas. To better deliver light to these regions-of-interest, we studied, in particular, intranasal and transcranial illumination approaches. We found that transcranial illumination at the F3-F4 location (based on 10-20 system) provided excellent light delivery to the dlPFC, while a light source located in close proximity to the cribriform plate was well suited for reaching the vmPFC, despite the fact that accessing the latter location may require a minimally invasive approach. Alternative non-invasive illumination strategies for reaching vmPFC were also studied and both transcranial illumination at the Fp1-FpZ-Fp2 location and intranasal illumination in the mid-nose region were shown to be valid. Different illumination wavelengths, ranging from 670-1064 nm, were studied and the amounts of light energy deposited to a wide range of brain regions were quantitatively compared. We found that 810 nm provided the overall highest energy delivery to the targeted regions. Although our simulations carried out on locations and wavelengths were not designed to be exhaustive, the proposed illumination strategies should inform the design of t-PBM systems likely to improve brain emotion regulation, both in clinical research and practice.
Dual-grid mesh-based Monte Carlo simulations for photon transport
The mesh-based Monte Carlo (MMC) method is an efficient algorithm to model light propagation inside tissues with complex boundaries, but choosing appropriate mesh density can be challenging. A fine mesh improves the spatial resolution of the output but requires more computation. In this paper, we propose an improved MMC dual-grid MMC or DMMC to accelerate photon simulations by using a coarsely tessellated tetrahedral mesh for ray-tracing computation and an independent voxelated grid for output data storage. The decoupling between the ray-tracing and data storage grids allows us to simultaneously achieve faster simulations and improved output spatial accuracy. Furthermore, we developed an optimized ray-tracing technique to eliminate unnecessary ray-tetrahedron intersection tests in optically-thick mesh elements. We validate the proposed algorithms using a complex heterogeneous domain and compare the solutions with those from MMC and voxel-based Monte Carlo. We found that DMMC with an unrefined constrained Delaunay tessellation of the boundary nodes yielded the highest speedup, ranging from 1.3X to 2.9X for various scattering settings, with nearly no loss in accuracy. In addition, the optimized ray-tracing technique offers excellent acceleration in high scattering media, reducing the ray-tetrahedron test count by over 100-fold. Our DMMC software can be downloaded at http://mcx.space/mmc
Developing surrogate models for industrial chemical processes
Industrial chemical processes, especially in the presence of recycle loops, exhibit highly nonlinear behaviors that oftentimes do not have a closed-form expression. A recent trend in the area is to develop surrogate models, also known as metamodels, that are approximate representations of the underlying process. These models, mostly polynomial expressions, have a more tractable form that can be more easily used to optimize and to understand the interactions between various process units and manipulated variables. In practice, the size of the input parameters can be extremely large, which makes the fitting of these surrogate models computationally expensive.
In this work, we show how high-dimensional steady-state surrogate models can be developed with high accuracy by combining the use of design of experiment techniques to alleviate computational costs and regularization techniques to avoid overfitting given the large number of terms involved. Through examples of three chemical process simulations (the Tennessee Eastman problem, an ethyl benzene, and a mono-isopropyl amine plant), we demonstrate the accuracy of the models and show applications of these surrogate models.