Characteristics of the Babinski-Weil test in people with Multiple Sclerosis

Abstract— The Babinski-Weil test is commonly used to assess gait and balance abnormalities, but traditional methods for evaluating its outcomes often rely on qualitative and subjective measures. This study explores the use of the Intel Realsense T265 Tracking Camera to quantitatively assess the Babinski-Weil test. The Realsense sensor was validated against a gold-standard motion capture system, showing a median position error of 36.4 mm and a median angle error of 0.49°. A preliminary study suggests that the sensor can reliably obtain key clinical outcome such as deviation angles and lateral oscillations, offering a promising tool for enhancing the accuracy of neurological assessments. Keywords—Multiple Sclerosis, gait and balance assessment, tracking sensor I. INTRODUCTION Multiple Sclerosis (MS) is a neurodegenerative demyelinating disease that affects the central nervous system. It is the most common non-traumatic disabling disease in young adults between 20 and 40 years old [1] and it affects women twice more often than men. The main clinical subtypes are the following [2]: Relapsing-Remitting MS (RRMS), Secondary Progressive MS (SPMS), Primary Progressive MS (PPMS). Each of these subtypes can be further distinguished based on disease activity: the "active" phenotype describes people with MS (pwMS) who have an acute attack or MRI activity, i.e., contrast enhancing lesions or new/enlarging lesions (assessed at least annually) [2]. An additional descriptor of the disease course is progression: it refers to the clinical evidence of disease progression, independent from relapses [2]. In some cases, the disease may be "not active", i.e., with no acute relapses or MRI activity, but "progressive". The identification of progression in Multiple Sclerosis is based on clinical grounds and it is retrospective; however, there are currently no universally accepted criteria, and the classification of a pwMS as progressive is often delayed by months or even years [3]. Traditional clinical neurological tests and scales (e.g., EmNSA [4], EDSS [5], T25WT, Babinski-Weil test, Romberg test) are widely used in the monitoring of MS symptoms, but they have some limitations, including qualitative outcome measures derived from visual observations, dependence on clinician expertise, insensitivity to subtle impairment changes, and susceptibility to floor/ceiling effects. For this reason, there is a need to design and implement tools that are able to detect MS progression early and monitor the evolution of sensorimotor deficits [3]. With this goal in mind, the work presented in this abstract shows the steps of the validation of the Intel Realsense T265 Tracking Camera (Intel Realsense, Intel, Santa Clara, California) and its use in a study aimed at quantitively measuring the outcome of the Babinski-Weil test. II. INTEL REALSENSE T265 TRACKING SENSOR MEASUREMENT ERROR ASSESSMENT A. Experimental setup To validate the Intel Realsense T265 Tracking Camera (from now referred to as "Realsense"), its measurements were compared with those extracted by a system of eight motion capture infrared cameras (Vicon, Oxford Metrics plc, Yarnton, Oxfordshire, United Kingdom). This motion capture system tracks the position of reflective markers and is considered a gold standard in the field. Each of the trials included a single participant. The following setups were used to conduct the experiments: - Setup 1: the Realsense with a marker placed on top was worn at the waist level of the participant, who walked in a straight line for about 15 s after performing a knee flexion (this initial movement was used to synchronize the two acquisition systems). - Setup 2: the Realsense was placed onto a TIAgo robot controlled through a joystick, with one marker on the sensor and three markers on the upper side of the robot. The robot was used in order to have a more repeatable trajectory. - Setup 3: the Realsense with a marker placed on top was handheld while the participant walked along predefined trajectories that identified a range of angles from 0° to 90° with a 5° step. The Realsense has a sampling rate of 30 Hz, while the Vicon System cameras have a sampling rate of 100 Hz. B. Data analysis In the first and second setups, the outcome measure was the position measurement error. The data analysis pipeline for the first two experimental setups was the following: the Realsense and Vicon signals were synchronized by identifying the minimum value on the z-axis and making it the origin of both signals; then, the reference system of each acquisition was aligned. This was done firstly by subtracting the mean value of each spatial component of the data (x, y, z) from the trajectories, and secondly by determining the rotation matrix that minimizes the error between the recordings. The rotation matrix was then applied to the Realsense T265 data. Finally, the Vicon acquisition was resampled to match the number of the samples of the Realsense signal previously pre-processed. The position measurement error was computed as the median of the Euclidean distance between corresponding samples of the two recorded trajectories. In the third experimental setup, the outcome measure was the angle measurement error, computed as the difference between the angle identified by the two segments of the trajectory measured by the two sensors. As in the pipeline presented above, the data was pre-processed by applying a low pass 20 Hz uniform filter. However, in this instance, the reference frames did not need to be aligned to compare the absolute angle measurements, which are independent from the reference system. A linear regression model was applied to determine the angular coefficient of the two segments in the trajectory. The angle measurement was then computed from these angular coefficients. C. Results The median position measurement error was 36.4 mm. The median angle measurement error was 0.49°. These measurement errors are compatible with the precision required to evaluate and assess the characteristics of the Babinski-Weil test, since both the ranges of values of the total path length and of the deviation angle are significantly larger than the measurement error. III. PRELIMINARY STUDY ON THE BABINSKI-WEIL OUTCOMES IN PEOPLE WITH MULTIPLE SCLEROSIS In this paragraph, we present the study currently being conducted in the neurology clinic of the San Martino Hospital in Genoa. A. Participants We are recruiting people with Multiple Sclerosis with different EDSS scores, who are willing and able to perform the task described in the next paragraph. In case they require an assistive device for everyday walking, its use is allowed during the test. In the second part of the experimental phase of the study, we are going to recruit unimpaired controls, in order to compare the two populations. B. Experimental protocol The participants are asked to perform the Babinski-Weil test. It consists of walking three half-steps forward and three half-steps backwards, for a total of four times. Meanwhile, they are required to wear a blindfold, noise reducing headphones, and the Realsense fixed to an elastic band. They are asked to maintain the initial direction they are facing. Each participant performs a total of three trials. The outcome measures are the deviation angle from the initial direction and the lateral oscillation, which are derived from the trajectory tracked by the Realsense. C. Preliminary results The trajectories derived from one trial in the current study and from one unimpaired volunteer previously recorded are shown in Fig. 1 as an example of the potential discriminatory power of the Realsense instrumented task. At the end of the study, we expect to be able to use the quantitative measurements of deviation angle and lateral oscillation to discriminate between unimpaired controls and people with Multiple Sclerosis and to find thresholds between two or more disability groups, defined according to their EDSS scores.