A machine learning approach to support deep brain stimulation programming

Abstract

Adjusting the stimulation parameters is a challenge in deep brain stimulation (DBS) therapy due to the vast number of different configurations available. As a result, systems based on the visualization of the volume of tissue activated (VTA) produced by a particular stimulation setting have been developed. However, the medical specialist still has to search, by trial and error, for a DBS set-up that generates the desired VTA. Therefore, our goal is developing a DBS parameter tuning strategy for current clinical devices that allows defining a target VTA under biophysically viable constraints. We propose a machine learning approach that allows estimating the DBS parameter values for a given VTA, which comprises two main stages: i) A K-nearest neighbors-based deformation to define a target VTA preserving biophysically viable constraints. ii) A parameter estimation stage that consists of a data projection using metric learning to highlight relevant VTA properties, and a regression/classification algorithm to estimate the DBS parameters that generate the target VTA. Our methodology allows setting a biophysically compliant target VTA and accurately predicts the required configuration of stimulation parameters. Also, the performance of our approach is stable for both isotropic and anisotropic tissue conductivities. Furthermore, the computational cost of the trained system is acceptable for real-world implementations.

Artificial Intelligence in Deep Brain Stimulation: A Brief Review

“…Many factors affect the success of DBS, including candidate selection, accurate placement of electrodes, and pro-gramming process. The selection of patients depends on age, disease stage, and responsiveness to common medication, such as levodopa (20). On the other hand, there is no specific guideline for programming DBS, and it generally is performed by movement disorder neurologists, neurosurgeons, nurses, etc.…”

Closed-Loop Deep Brain Stimulation with Reinforcement Learning and Neural Simulation

“…This approach is beneficial for various aspects, including computeraided diagnosis/detection (CAD) for DBS candidate selection, optimization of cl-DBS algorithms, surgical targeting, etc [8], [9]. Our primary focus lies in optimizing the cl-DBS algorithm, addressing the crucial post-surgical challenge of DBS device programming for therapy [10].…”

“…⃝ synaptic current from STN to GPi (I ST N →GP i ). Inhibitory inputs are indicated by gray arrows, namely 7 ⃝ synaptic current from GPi to TH (I GP i→T H ), 8 ⃝ synaptic current from GPe to STN (I ST N ), 9 ⃝ synaptic current from GPe to GPi (I GP e→GP i ), and 10 ⃝ synaptic current from GPe to itself (I GP e→GP e ). Refer to Equation ( 1)-(3).…”

“…Excitatory inputs are represented by black arrows, including 1 ⃝ input from the sensorimotor cortex (I SM ), 2 ⃝ constant bias current, I app(ST N ) , to STN, 3 ⃝ constant bias current, I app(GP e) , from Stiatum to GPe, 4 ⃝ constant bias current, I app(GP i) , from Stiatum to GPi, 5 ⃝ synaptic current from STN to GPe (I ST N →GP e ), and 6 ⃝ synaptic current from STN to GPi (I ST N →GP i ). Inhibitory inputs are indicated by gray arrows, namely 7 ⃝ synaptic current from GPi to TH (I GP i→T H ), 8 ⃝ synaptic current from GPe to STN (I ST N ), 9⃝ synaptic current from GPe to GPi (I GP e→GP i ), and10 ⃝ synaptic current from GPe to itself (I GP e→GP e ). Refer to Equation (1)-(3).…”