2019/07/15 by Ioannis Bargiotas, Bargiotas, Ioannis, Argyris Kalogeratos +9
Health Professions · Medicine · #62H15 #Applications (stat.AP) #Balance, Gait, and Falls Prevention #Cerebral Palsy and Movement Disorders #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Parkinson's Disease Mechanisms and Treatments
paper · pdf · doi:10.48550/arxiv.1907.06614
openalex publication_date 2019/07/15 · openalex created_date 2022/10/01 · openalex updated_date 2026/07/28
Falling in Parkinsonian syndromes (PS) is associated with postural\ninstability and consists a common cause of disability among PS patients.\nCurrent posturographic practices record the body's center-of-pressure\ndisplacement (statokinesigram) while the patient stands on a force platform.\nStatokinesigrams, after appropriate signal processing, can offer numerous\nposturographic features, which however challenges the efforts for valid\nstatistics via standard univariate approaches. In this work, we present the\nts-AUC, a non-parametric multivariate two-sample test, which we employ to\nanalyze statokinesigram differences among PS patients that are fallers (PSf)\nand non-fallers (PSNF). We included 123 PS patients who were classified into\nPSF or PSNF based on clinical assessment and underwent simple Romberg Test\n(eyes open/eyes closed). We analyzed posturographic features using both\nmultiple testing with p-value adjustment and the ts-AUC. While the ts-AUC\nshowed significant difference between groups (p-value = 0.01), multiple testing\ndid not show any such difference. Interestingly, significant difference between\nthe two groups was found only using the open-eyes protocol. PSF showed\nsignificantly increased antero-posterior movements as well as increased\nposturographic area, compared to PSNF. Our study demonstrates the superiority\nof the ts-AUC test compared to standard statistical tools in distinguishing PSF\nand PSNF in the multidimensional feature space. This result highlights more\ngenerally the fact that machine learning-based statistical tests can be seen as\na natural extension of classical statistical approaches and should be\nconsidered, especially when dealing with multifactorial assessments.\n