Parkinson's disease (PD) is the second most common neurodegenerative disorders and is characterised by motor and non-motor symptoms. Freezing of gait (FOG) is one of the motor symptoms of PD in which patients have an interruption in walking by feeling their feet stuck to the floor. The main goal of the work was to classify PD patients with and without FOG by means of a short sway to analyse their postural stability. Forty-two PD patients (sixteen patients with FOG and twenty-six without FOG) performed the stabilometric analysis by standing on a force platform with their eyes open for 5-6 seconds. Univariate statistical analysis was conducted to compare PD with and without FOG. Findings suggested that postural instability was higher in patients with FOG as revealed by the values of minimum and maximum oscillations of the centre of pressure and the length as function of surface ratio index. Then, a machine learning (ML) analysis was performed using sway variables as input to different classifiers: Decision Tree (DT), Naïve Bayes (NB) and Random Forest (RF). All classifiers obtained high evaluation metrics. In particular, DT achieved the best accuracy (0.80) and sensitivity (0.89) while RF reached the highest value of AUCROC (0.84). The ML analysis was able to distinguish between PD patients with and without FOG based on sway variables. Finally, the sway data revealed postural instability severity in patients with FOG.
Measurements of Postural Sway to Classify Freezing of Gait in Parkinson's Disease
Calabrese, Maria Consiglia;Di Filippo, Federico;Barone, Paolo;Amboni, Marianna;
2024
Abstract
Parkinson's disease (PD) is the second most common neurodegenerative disorders and is characterised by motor and non-motor symptoms. Freezing of gait (FOG) is one of the motor symptoms of PD in which patients have an interruption in walking by feeling their feet stuck to the floor. The main goal of the work was to classify PD patients with and without FOG by means of a short sway to analyse their postural stability. Forty-two PD patients (sixteen patients with FOG and twenty-six without FOG) performed the stabilometric analysis by standing on a force platform with their eyes open for 5-6 seconds. Univariate statistical analysis was conducted to compare PD with and without FOG. Findings suggested that postural instability was higher in patients with FOG as revealed by the values of minimum and maximum oscillations of the centre of pressure and the length as function of surface ratio index. Then, a machine learning (ML) analysis was performed using sway variables as input to different classifiers: Decision Tree (DT), Naïve Bayes (NB) and Random Forest (RF). All classifiers obtained high evaluation metrics. In particular, DT achieved the best accuracy (0.80) and sensitivity (0.89) while RF reached the highest value of AUCROC (0.84). The ML analysis was able to distinguish between PD patients with and without FOG based on sway variables. Finally, the sway data revealed postural instability severity in patients with FOG.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.