A graph-based approach to study motor coordination in Parkinson’s Disease gait: a longitudinal study to assess the effectiveness of Deep Brain Stimulation neurosurgery

Introduction. Graph theory is emerging as a promising technique in different contexts, and it can be used to extract a network of muscles based on their coordinated activity during gait. This work aims to investigate the motor control strategies of Parkinson’s Disease (PD) patients through graph theory and Louvain clustering and to evaluate the successfulness of Deep Brain Stimulation (DBS) in alleviating PD motor symptoms. Methods. Gait analysis, inclusive of surface electromyography (EMG) of the main muscles involved in locomotion, was carried out on 30 PD patients and 30 controls. PD patients were longitudinally followed-up, with assessments at 3 time points: pre-DBS implant (T0), 3-month post-DBS implant (T1), and 12 months post-DBS implant (T2). Intermuscular adjacency matrices computed from EMG data of 12 lower-limb and trunk muscles were used to extract graph networks. Each graph network consists of nodes (i.e., muscles) and edges (i.e., weighted connections between muscles). The graph “modularity” was extracted from each graph. A 1-way ANOVA with Bonferroni correction for multiple comparisons was performed to discriminate statistically significant differences in graph modularity among PD patients (at the 3 time points) and controls. Results. Muscle-network graph of a representative PD patient is shown before (Fig. 1A) and 12 months after DBS (Fig. 1B). The graph modularity increased from 0.28 to 0.49 during the follow-up of this specific patient. The modularity of the PD population at T0, T1, and T2 vs. controls are shown in Fig. 1C. The PD modularity at T0 was significantly smaller than that of controls (PD at T0: 0.35±0.01 (mean±SE); controls: 0.40±0.01; p=0.019), becoming not different from that of controls at T1 (0.36±0.01; p=0.18) and T2 (0.38±0.01; p=1.00). Discussion. Modularity is a metric that reflects the separability of muscle groups that activate synergistically. Lower graph modularity may indicate reduced independence among the muscle groups and decreased motor control complexity. Graph modularity proved a sensitive measure to assess short- and long-term motor improvements in PD patients following DBS.