A Computational Model of Deep Brain Stimulation for Parkinson's Disease Tremor and Bradykinesia.

Parkinson's disease (PD) is a progressive neurological disorder that is typically characterized by a range of motor dysfunctions, and its impact extends beyond physical abnormalities into emotional well-being and cognitive symptoms. The loss of dopaminergic neurons in the substantia nigra pars compacta (SNc) leads to an array of dysfunctions in the functioning of the basal ganglia (BG) circuitry that manifests into PD. While active research is being carried out to find the root cause of SNc cell death, various therapeutic techniques are used to manage the symptoms of PD. The most common approach in managing the symptoms is replenishing the lost dopamine in the form of taking dopaminergic medications such as levodopa, despite its long-term complications. Another commonly used intervention for PD is deep brain stimulation (DBS). DBS is most commonly used when levodopa medication efficacy is reduced, and, in combination with levodopa medication, it helps reduce the required dosage of medication, prolonging the therapeutic effect. DBS is also a first choice option when motor complications such as dyskinesia emerge as a side effect of medication. Several studies have also reported that though DBS is found to be effective in suppressing severe motor symptoms such as tremors and rigidity, it has an adverse effect on cognitive capabilities. Henceforth, it is important to understand the exact mechanism of DBS in alleviating motor symptoms. A computational model of DBS stimulation for motor symptoms will offer great insights into understanding the mechanisms underlying DBS, and, along this line, in our current study, we modeled a cortico-basal ganglia circuitry of arm reaching, where we simulated healthy control (HC) and PD symptoms as well as the DBS effect on PD tremor and bradykinesia. Our modeling results reveal that PD tremors are more correlated with the theta band, while bradykinesia is more correlated with the beta band of the frequency spectrum of the local field potential (LFP) of the subthalamic nucleus (STN) neurons. With a DBS current of 220 pA, 130 Hz, and a 100 microsecond pulse-width, we could found the maximum therapeutic effect for the pathological dynamics simulated using our model using a set of parameter values. However, the exact DBS characteristics vary from patient to patient, and this can be further studied by exploring the model parameter space. This model can be extended to study different DBS targets and accommodate cognitive dynamics in the future to study the impact of DBS on cognitive symptoms and thereby optimize the parameters to produce optimal performance effects across modalities. Combining DBS with rehabilitation is another frontier where DBS can reduce symptoms such as tremors and rigidity, enabling patients to participate in their therapy. With DBS providing instant relief to patients, a combination of DBS and rehabilitation can enhance neural plasticity. One of the key motivations behind combining DBS with rehabilitation is to expect comparable results in motor performance even with milder DBS currents.

A generalized reinforcement learning based deep neural network agent model for diverse cognitive constructs.

Human cognition is characterized by a wide range of capabilities including goal-oriented selective attention, distractor suppression, decision making, response inhibition, and working memory. Much research has focused on studying these individual components of cognition in isolation, whereas in several translational applications for cognitive impairment, multiple cognitive functions are altered in a given individual. Hence it is important to study multiple cognitive abilities in the same subject or, in computational terms, model them using a single model. To this end, we propose a unified, reinforcement learning-based agent model comprising of systems for representation, memory, value computation and exploration. We successfully modeled the aforementioned cognitive tasks and show how individual performance can be mapped to model meta-parameters. This model has the potential to serve as a proxy for cognitively impaired conditions, and can be used as a clinical testbench on which therapeutic interventions can be simulated first before delivering to human subjects.

A Multiscale, Systems-Level, Neuropharmacological Model of Cortico-Basal Ganglia System for Arm Reaching Under Normal, Parkinsonian, and Levodopa Medication Conditions.

In order to understand the link between substantia nigra pars compacta (SNc) cell loss and Parkinson's disease (PD) symptoms, we developed a multiscale computational model that can replicate the symptoms at the behavioural level by incorporating the key cellular and molecular mechanisms underlying PD pathology. There is a modelling tradition that links dopamine to reward and uses reinforcement learning (RL) concepts to model the basal ganglia. In our model, we replace the abstract representations of reward with the realistic variable of extracellular DA released by a network of SNc cells and incorporate it in the RL-based behavioural model, which simulates the arm reaching task. Our results successfully replicated the impact of SNc cell loss and levodopa (L-DOPA) medication on reaching performance. It also shows the side effects of medication, such as wearing off and peak dosage dyskinesias. The model demonstrates how differential dopaminergic axonal degeneration in basal ganglia results in various cardinal symptoms of PD. It was able to predict the optimum L-DOPA medication dosage for varying degrees of cell loss. The proposed model has a potential clinical application where drug dosage can be optimised as per patient characteristics.