Differential contribution of sensorimotor cortex and subthalamic nucleus to unimanual and bimanual hand movements

Abstract

Why does unilateral deep brain stimulation improve motor function bilaterally? To address this clinical observation, we collected parallel neural recordings from sensorimotor cortex (SMC) and the subthalamic nucleus (STN) during repetitive ipsilateral, contralateral, and bilateral hand movements in patients with Parkinson's disease. We used a cross-validated electrode-wise encoding model to map electromyography data to the neural signals. Electrodes in the STN encoded movement at a comparable level for both hands, whereas SMC electrodes displayed a strong contralateral bias. To examine representational overlap across the two hands, we trained the model with data from one condition (contralateral hand) and used the trained weights to predict neural activity for movements produced with the other hand (ipsilateral hand). Overall, between-hand generalization was poor, and this limitation was evident in both regions. A similar method was used to probe representational overlap across different task contexts (unimanual vs. bimanual). Task context was more important for the STN compared to the SMC indicating that neural activity in the STN showed greater divergence between the unimanual and bimanual conditions. These results indicate that SMC activity is strongly lateralized and relatively context-free, whereas the STN integrates contextual information with the ongoing behavior.

Keywords: DBS; ECoG; bmanual movement; ipsilateral; movement encoding.

Fig. 1

Electrode-wise encoding model. Ridge regression was used to predict neural activity in individual electrodes from continuous EMG activity. A) Cross-validation. Nested 5-fold cross-validation was used to select the regularization hyperparameter () on inner validation sets. B) Prediction performance. The held-out EMG feature matrix was multiplied with the trained weights to create model predictions.

Fig. 2

Generalized encoding model configurations. A) Across hand model. To quantify between-hand generalization, we fit the model with data from the unimanual contralateral condition and used the trained weights to predict neural activity during the unimanual ipsilateral condition. B) Across context model. To quantify generalization across contexts we fit the model with data from the unimanual contralateral condition and used the trained weights to predict neural activity during the bimanual condition.

Fig. 3

Summary of predictive electrodes across conditions and brain regions. Each electrode was assigned to one of four categories based on predictive performance (R² > 0.01): Contralateral only; ipsilateral only; both; or none. During unimanual movement (top row), SMC electrodes are more likely to be predictive of contralateral movement whereas STN electrodes are equally predictive of contralateral and ipsilateral movement. Comparing across brain regions during bimanual movements (bottom row), significantly more electrodes were classified as none in the STN region compared to SMC.

Fig. 4

Stronger bilateral encoding in the STN region compared to SMC during unimanual movements. A) Unimanual prediction performance. Performance of all predictive electrodes, measured as the square of the Pearson correlation (R²), using EMG from either the contralateral (Y-axis) or ipsilateral (X-axis) hand during unimanual movements. Overall performance did not vary by brain region, but a significant interaction was found with electrodes in the STN region performing equally well for both hands whereas the SMC displayed a strong contralateral bias. Data are averaged prediction performance from the five held-out test sets. Upper right corner: Difference distribution for each brain region. B) Example traces. Held-out predictions of the LMP time series for one test set of E1 and E2 during unimanual movements.

Fig. 5

Stronger bilateral encoding in the STN region compared to SMC during bimanual movements. A) Bimanual prediction performance. Performance of all predictive electrodes, measured as the square of the Pearson correlation (R²), using EMG from either the contralateral (Y-axis) or ipsilateral (X-axis) hand when both hands were moving. A significant interaction was found between hand and brain region with electrodes in the STN region performing equally well for both hands whereas a strong contralateral bias was found in SMC. Data are averaged prediction performance from the five held-out test sets. Upper right corner: Difference distribution for each brain region. Insets: Time-locked average of LMP for contralateral and ipsilateral hand opening for E1 and E2. B) Example traces. Held-out predictions of the LMP time series for one test set of E1 and E2 when both hands were moving.

Fig. 6

Poor across hand generalization in SMC and STN. A) Hand generalization. Percent change from within hand R² to across hand R² was calculated for each electrode for SMC (left) and STN (right). No difference in hand generalization was found between the two brain regions. B) Across hand performance. Performance of all predictive electrodes, measured as the square of the Pearson correlation (R²), using EMG from either the same hand (Y-axis) or training and testing across hands (X-axis) during unimanual movements. Data are averaged prediction performance from the five held-out test sets. Upper right corner: Difference distribution for each brain region. Insets: Time-locked average of LMP for contralateral and ipsilateral hand opening for two electrodes (E1 and E2). C) Example traces. Held-out predictions of the LMP time series for one test set of E1 and E2, the first being within hand and the second being across hand predictions.

Fig. 7

Across context prediction performance. A) Context generalization. Percent change from within task R² to across task R² was calculated for each electrode for SMC (left) and STN (right). A significant difference was found in context generalization between the two brain regions, with electrodes within the SMC showing stronger context generalization. B) Across context performance. Performance of all predictive electrodes, measured as the square of the Pearson correlation (R²), using EMG from either the same task (unimanual; Y-axis) or training and testing across tasks (unimanual to bimanual; X-axis). Data are averaged prediction performance from the five held-out test sets. Upper right corner: Difference distribution for each brain region. Insets: Time-locked average of LMP for contralateral and ipsilateral hand opening for E1 and E2. C) Held-out predictions. Held-out predictions of the LMP time series for one test set of E1 and E2 when both hands were moving.