This project is designed by Dr. Karlo Malaga, Assistant Professor of Biomedical Engineering at Bucknell University

Background

Parkinson Disease is a neurodegenerative disorder caused by the death of dopamine-secreting cells in a specific area of the brain called Substantia Nigra. People with Parkinson Disease may experience tremor, rigidity, and postural instability. 

Deep Brain Stimulation (DBS) is one of the popular treatments of Parkinson Disease. DBS involves chronically implanting electrodes in deep brain structures and delivering high-frequency electrical pulses to them. The target of DBS is often the Sub-thalamic nucleus. The generated pulses are configured to have frequency greater than 130 Hz, amplitude between 1-5 V, and last between 60-200 us.  

Tissue activation modeling is a neural engineering approach based on the belief that DBS's therapeutic benefit is strongly dependent on spatial distribution of stimulation-induced electric field relative to specific neuroanatomy. In tissue activation modeling, we closely look at the Volume of Tissue Activated (VTA) to simulate different clinical scenarios and make meaningful predictions.

Patient-specific tissue activation modeling is used to identify optimal DBS targets for Parkinson Disease because a truly patient-specific approach will result in more accurate predictions of stimulation spread into specific anatomical structures and clinical outcome. 

Overview

In this project, I use COMSOL to build the VTA in anisotropic and isotropic brain models to examine whether it is necessary to incorporate anisotropy in tissue activation model. I also investigated the difference between the VTA defined by electric field and the VTA defined by electric potential to find a better method of modeling. Lastly, I looked at the optimal location to implant the electrode in DBS. 

DESIGN & ANALYSIS

1. Anisotropic versus Isotropic modeling

From the two models above, I observed that the VTA in an anisotropic model is less spherical and smaller than the one in an isotropic model. Such difference occurs because in the isotropic model, we assume that the conductivity of the brain is the same throughout the tissue matter. This means that the electric potential can extend outward equally from the electrode position; that’s why the VTA is spherical. However, in the anisotropic model, there is some difference in conductivity, which makes the electric potential distribution less even. As a result, there will be some “missing” parts to make up the entire sphere, accounting for the smaller volume of the anisotropic model. 

I think it is necessary to incorporate anisotropy in tissue activation models because the brain is essentially anisotropic, consisting of many domains such as gray and white matter, cerebrospinal fluid, skull, scalp. These domains have different electrical properties, one of which is resistivity (table below), which is crucial in determining the electric potential around the implanted electrode. Therefore, an anisotropic model will better resemble the real characteristics of the brain. However, if we consider the utility effect between an anisotropic and isotropic model, an isotropic one is more commercially convenient, considering that we can compute many patients’ data in a shorter amount of time. Depending on our purpose to perform complicated research or to gain quick access to the electric potential in the activated tissue, we can choose anisotropic or isotropic model. In this class setting, since I want to learn more about the neural engineering technique, an anisotropic model will be more beneficial. 

2. The role of frequency in brain modeling

The model I built is purely resistive. However, since DBS operates on a range of frequencies, it is worth noticing the brain's capacitive component.

DBS typically operates at frequency > 130 Hz. When DBS operates between 100Hz and 1000Hz, the brain can be assumed to be purely resistive because the impedance's imaginary component is negligible. If we were to stimulate at a very high frequency such as 10,000 Hz, the brain model is no longer purely resistive. 

3. Electric-potential-defined versus Electric-field-defined model 

By comparing the Electric-potential-defined model and the Electric-field-defined model, I noticed that the electric-potential VTA is smoother and more concentrated whereas the Electric-field VTA is coarser and more spread out. The difference in smoothness between two models occurs because the electric field is the derivative of the electric potential. When we compute the derivative, we essentially take a difference between two points in a finite set of points, which suggests that we are reducing the data set. Since there are fewer points to be plotted in the electric field model, the plot generated is less continuous, and therefore is coarser than the plot in the electrical-potential case. 

Although the electric-field-defined VTA seems less desirable due to its coarseness, in real life, smoothness is not a critical factor to determine which definition of VTA is better. We should consider each definition in terms of clinical relevance. If the VTA defined by electric potential matches well with the patients clinically while the VTA defined by the electric field does not, we will choose the electric potential one, and vice versa. Many neural engineers define VTAs by second spatial difference of the electric potential because it lines up with the advanced activation model’s formula widely used in research.

4. Optimal location for electrode implantation

Using MATLAB, I combined the patient-specific Sub-thalamic nucleus (STN), VTA, DBS electrode locations, and the brain on the same plot to make it more convenient to visualize.

I compared electrode stimulation within STN with electrode stimulation just above the STN to find a more optimal location for DBS. In the figure on the left, electrode is right on above the STN. In the figure on the right, electrode is within the STN.