Postural control analysis in multiple sclerosis with perceptive computing based on Microsoft's Kinect

Abstract

Background: Postural control is an important key factor for secure and independent mobility. It is restricted in the vast majority of patients with multiple sclerosis (MS). Even if a patient does not yet suffer from functional disability - deficits of balance do already appear in the early phase of MS, but are not detected by standard examination and scoring procedures. Perceptive computing using data from consumer grade 3D sensors like the Microsoft Kinect allows motion tracking in everyday settings and promises exact and relatively low cost motor diagnostics. Objectives: To develop a reliable and observer-independent method for quantification of postural control in patients with multiple sclerosis using perceptive computing. Methods: We examined 99 patients with Clinically Isolated Syndrome (CIS) or MS (57 female, 42 male, mean(plus or minus)SD age 44(plus or minus)11 years, median EDSS 2.5) and 60 healthy controls (HC) with a custom built system using Microsoft's Kinect. Subjects performed three measurements with comfortable, closed and tandem stance, each with eyes open the first 15 seconds and eyes closed the last 15 seconds. Roll speed along the medio-lateral axis, pitch speed along the anterior-posterior axis and sum of 3D movements were calculated. Patients were examined based on the Expanded Disability Status Scale (EDSS) and the Multiple Sclerosis Functional Composite (MSFC). Self-reported walking disability was assessed using the MS walking scale questionnaire (MSWS-12). Results: Already in comfortable stance with open eyes, MS patients showed significantly higher sum movements than HC (7.52(plus or minus)3.54 cm vs. 5.79(plus or minus)1.95 cm, p< 0.001), pitch speed (0.41(plus or minus)0.22 cm/sec vs. 0.30(plus or minus)0.10 cm/sec, p=< 0.001) and roll speed (0.26(plus or minus)0.15 cm/sec vs. 0.20(plus or minus)0.09 cm/sec, p=0.005). On average, MS patients worsened both with closed eyes as well as with increasing stance difficulty. Best correlations with disease severity and self-reported walking disability were achieved with closed stance and closed eyes for summed 3D movement (EDSS Spearman's Rho=0.500, p< 0.001; MSWS-12 Rho=0.455, p< 0.001), pitch speed (EDSS Rho=0.478, p< 0.001, MSWS-12 Rho=0.426) and roll speed (EDSS Rho=0.455, p< 0.001; MSWS-12 Rho=0.420, p< 0.001). Conclusions: Perceptive computing based assessment of postural control was fast, well tolerated and effective in detecting sway differences between HC and MS patients