360° turning in individuals with mild to moderate Parkinson’s disease: a clustering analysis of segmental coordination strategies

Turning is a complex motor task frequently impaired in Parkinson’s disease (PD), contributing to fall risk and reduced mobility. While previous studies have documented increased turn duration and en-bloc segmental coordination during 90° and 180° turns in people with PD, repeated 360° turning-in-place is embedded in standard balance scales, imposes greater demands on axial sequencing and balance than smaller turns, and yields several consecutive turns under standardized conditions. Segment-specific temporal turning strategies and their variability across repeated turns have not yet been characterized during this task. To characterize axial segmental coordination during 360° turning in people with mild to moderate Parkinson’s disease and in age- and sex-matched healthy controls (HC). Fifty-four participants with idiopathic PD and 22 age- and sex-matched HC completed a one-minute 360° turning task, alternating between left and right turns. Inertial measurement units (IMUs) on the Head, Sternum, and Lumbar spine captured rotational movement data. A double-stage, unsupervised K-means, clustering algorithm was used to detect segment-specific turn onsets and delay times as a measure of en-bloc coordination. Segmental initiation rates were analyzed, and unsupervised clustering identified dominant movement strategies. Compared with HCs, participants with PD completed fewer turns ( p = 0.011), rotated through a smaller angle per turn, defined as the total lumbar rotation achieved in each turn ( p = 0.002), and took longer to complete each turn ( p = 0.006 and 0.002 for the slower and faster side). They more often initiated turns with the Head, whereas HCs showed a more distributed Head and Lumbar onset pattern, confirmed by turn-level mixed-effects logistic regression [all false discovery rate (FDR)-adjusted p < 0.001]. The Sternum and Lumbar onset delays were longer in PD ( p < 0.001 and p = 0.028), and the intersegmental delay areas were larger in PD for all three pairs ( p = 0.037, 0.037, and 0.040). After adjustment for turn duration, the Sternum onset delay remained longer in PD, whereas the delay-area differences did not reach significance. The net intersegmental lead, the signed delay area, showed the Head ahead of both trunk segments in both groups ( p < 0.001), with no group difference. Within PD, the Head-led cluster showed a longer Head-Sternum onset delay than the Lumbar-led cluster (FDR-adjusted p = 0.009). During 360° turns, neither HC nor PD used a single axial coordination pattern. Participants with PD favored a Head-first strategy and showed greater temporal separation between the Head and the trunk, partly accounted for by their slower turning, while the Head led the trunk in both groups and the Sternum-Lumbar pair remained the most tightly coupled. Full-circle turning in PD therefore does not reduce to a uniform rigid en-bloc strategy. Whether this reflects a task-specific adaptation or compensatory visual anchoring remains to be tested, as only one turn amplitude was studied and gaze was not recorded. These patterns may help refine assessment tools and interventions.