Wearable gait Analysis LAB as a biomarker of Parkinson’s disease motor stages and Quality of life: a preliminary study

Abstract

Bradykinetic, shuffling and shorter steps are prototypical signs of gait in Parkinson's Disease (PD) and important indicators of Quality of Life (QoL). Advances in wearable technologies enabled their use to objectively evaluate these gait fluctuations complementing the subjective categorical clinical scales usually used by clinicians. This paper aims to study the ability of a wearable gait analysis lab, developed by our team, to serve as a biomarker of PD motor stages and an indicator of patients' QoL. We accomplished experimental tests which involved repeated measurements of walking trials from a crosssection study with eighteen PD patients and thirteen healthy subjects. We measured gait spatiotemporal parameters and cross these data with commonly PD-scales to assess motor symptoms (UPDRS-III) and quantify patients' QoL (PDQ39). Patients presented a bradykinetic gait with shorter steps, variability and asymmetry in spatiotemporal parameters. These prototypical signs were also observed along the disease progression considering UPDRS-III and PDQ39 levels. Positive outcomes demonstrated the feasibility and applicability of our objective wearable sensorbased gait analysis in PD to measure typical parkinsonian gait and a (bio)marker of PD motor stages and patients' quality of life level.

Objective assessment of postural instability in Parkinson’s disease under dynamic conditions

“…Indeed, the Spearman product-moment correlation analysis showed this tendency by verifying positive/negative values of correlation factor. In turn, while walking, patients with increased postural instability show slower and shorter gait cycles and more locomotion asymmetry and variability [21]. The Spearman correlation analysis showed indeed this effect.…”

“…Gait-associated metrics were estimated based on a pre-identi cation of initial/ nal contact detection using the gait events detection algorithm validated by the authors team in [19]. In addition, the computational method used to estimate most of these metrics was previously validated in [19] showing an acceptable error of ≤ 13,45%, and tested in [21] describing the expected parkinsonian prototypical motor signs. Given patients with higher postural instability tend to present a more cautious behaviour in performing the tasks, which leads to a slower and smaller steps, it is expected a decrease in step/stride length, velocity, cadence and swing time and, in turn, an increase in step/stride time and stance and double support time [8, 9,11,19].…”

Deep learning model for doors detection: A contribution for context-awareness recognition of patients with Parkinson’s disease

“…Patients with PD tend to walk slowly, as observed in (Branquinho et al, 2021), where, e.g., 20 patients performed a mean step duration of ~0.601sec when walking 10 m. When testing the model in the RPi integrated on the wearable device, it was observed that by using TF-Lite environment, in real-time detection, we achieved a ~2.87 fps to capture and identify doors, which is a time-effective performance to detect step changes over real-life scenarios Fig. 6.…”

“…6. Indeed, considering the outcomes measured in (Branquinho et al, 2021), with our solution we have ~2frames in a step time of 0.601sec to identify that the patient is in front of a door and deliver on-demand the proprioceptive cue. Also, the DL model was accurate and efficient in detecting doors in real-life scenarios, as observed in Fig.…”