A Comparative Severity Assessment of Impaired Balance due to Cerebellar Ataxia using Regression Models

Thang Ngo,Lahiru L Abeysekara,Pubudu N Pathirana,Malcolm Horne,Laura Power,David J Szmulewicz

DOI: https://doi.org/10.1109/EMBC44109.2020.9175705

Abstract:Cerebellar ataxia (CA) refers to the impaired balance and coordination resulting from injury or degeneration of the cerebellum. Testing balance is one of the simplest means of assessing CA. This study compares instrumented assessment and clinical assessment scales of the balance test called Romberg's test. Inertial Measurement Unit (IMU) data were collected from a sensor attached to their chest of 53 subjects while they performed the test. The corresponding clinical scores were also tabulated. Using this data, 99 features were extracted to quantify acceleration, tremor and displacement of body sway. These features were filtered to identify the subset that better characterize the distinctive behavior of CA subjects. Elastic Net Regression model resulted a greater agreement (0.70 Pearson coefficient) with the clinical SARA scores. The overall results indicated that data from a single IMU sensor is sufficient to accurately assess balance in CA. The significance of this study is that evaluation of balance using Recurrence Quantification Analysis produces a comprehensive framework for the assessment of CA.