Revealing posturographic profile of patients with Parkinsonian syndromes through a novel hypothesis testing framework based on machine learning.

Falling in Parkinsonian syndromes (PS) is associated with postural instability and consists a common cause of disability among PS patients. Current posturographic practices record the body's center-of-pressure displacement (statokinesigram) while the patient stands on a force platform. Statokinesigrams, after appropriate processing, can offer numerous posturographic features. This fact, although beneficial, challenges the efforts for valid statistics via standard univariate approaches. In this work, 123 PS patients were classified into fallers (PSF) or non-faller (PSNF) based on the clinical assessment, and underwent simple Romberg Test (eyes open/eyes closed). We developed a non-parametric multivariate two-sample test (ts-AUC) based on machine learning, in order to examine statokinesigrams' differences between PSF and PSNF. We analyzed posturographic features using both multiple testing with p-value adjustment and ts-AUC. While ts-AUC showed significant difference between groups (p-value = 0.01), multiple testing did not agree with this result (eyes open). PSF showed significantly increased antero-posterior movements as well as increased posturographic area compared to PSNF. Our study highlights the superiority of ts-AUC compared to standard statistical tools in distinguishing PSF and PSNF in multidimensional space. Machine learning-based statistical tests can be seen as a natural extension of classical statistics and should be considered, especially when dealing with multifactorial assessments.