Érica Fontana Paiva

Advisors
Malte Ollenschläger (M.Sc.), Nils Roth (M.Sc.), Dr. med. Martin Regensburger (Molekulare Neurologie Uniklinikum Erlangen), PD Dr. phil. Heiko Gaßner (Molekulare Neurologie Uniklinikum Erlangen), Dr.-Ing. Felix Kluge, Prof. Dr. Björn Eskofier

Duration
11 / 2021 – 05 / 2022

Abstract
Hereditary Spastic Paraplegia (HSP) is a neurodegenerative disease which affects neurons in the spinal cord. This disorder weakens lower limbs leading to a progressive gait disorder [1]. The disease severity can be assessed by the Spastic Paraplegia Rating Scale (SPRS), a 13-item scale designed to rate functional impairment of spastic paraplegia. It is rated by clinicians without any assistive devices and is thus rater-dependent and subjective [2].

Therefore, instrumented gait analysis systems have been developed, which can help to differentiate HSP from other diseases, for instance, spastic diplegia [3] and cerebral palsy [4]. For data acquisition, standardized walking tests are performed and recorded using a sensor system. In order to analyse the acquired data, it is segmented into single strides and spatio-temporal gait parameters, and then stride time or length are calculated.

Methods for stride segmentation include peak detection [9], template-based [12], wavelet-based fractional analysis [10], dynamic time warping (DTW) [11], hidden Markov models (HMM) [5, 7] and local cyclicity estimation [6, 8]. For several cohorts of different neurodegenerative diseases it was found that HMMs are the preferable choice for stride segmentation [5, 6]. However, the HMMs differ in the stride definition they employ. While Martindale et al. [6] used the foot’s initial and terminal contact to predict stance and swing phases, Roth et al. made use of the stride segmentation suggested by Barth et al. [13], which detects the beginning and the end of a complete stride instead of the stance and swing phase separately.

On one hand, Martindale’s predictions can be used to directly assess temporal gait parameters, whereas Roth’s approach would need to be extended by a gait event detection. On the other hand, Martindale’s approach requires personalized models, whereas Roth’s HMM uses a general model. A further difference in these approaches is that Martindale et al. used data of HSP patients recorded at the lab while Roth et al. used free-living gait data of patients with Parkinson’s disease.

To date, it is unclear to what extent these two approaches differ in performance in a cohort of HSP patients. Therefore, these approaches will be compared in this thesis. Additionally to the HSP at-lab dataset, the student will make use of a home-monitoring dataset acquired by the department of molecular neurology, UK Erlangen. It contains data of five HSP patients, who performed standardized walking test three times a day for 14 days.

References:
[1] Salinas, S. et al. Hereditary spastic paraplegia: clinical features and pathogenetic mechanisms. The Lancet Neurology, 2008.
[2] Schüle, R. et al. The Spastic Paraplegia Rating Scale (SPRS). Neurology, 2006.
[3] Piccinini, L. et al 3D gait analysis in patients with hereditary spastic paraparesis and spastic diplegia: A kinematic, kinetic and EMG comparison. European Journal of Paediatric Neurology, 2011.
[4] Sebastian I. Wolf, Frank Braatz, Dimitrios Metaxiotis, Petra Armbrust, Thomas Dreher, Leonhard Doderlein, Ralf Mikut. Gait analysis may help to distinguish hereditary spastic paraplegia from cerebral palsy. Gait & Posture, vol 33 (4) 2011.
[5] Haji Ghassemi, N.; Hannink, J.; Martindale, C.F.; Gaßner, H.; Müller, M.; Klucken, J.; Eskofier, B.M. Segmentation of Gait Sequences in Sensor-Based Movement Analysis: A Comparison of Methods in Parkinson’s Disease. Sensors, vol 18 (1), 2018.
[6] Martindale, C. F. et al. Technical Validation of an Automated Mobile Gait Analysis System for Hereditary Spastic Paraplegia Patients. IEEE Journal of Biomedical and Health Informatics, vol 24 (5), 2019.
[7] Roth, N., Küderle, A., Ullrich, M. et al. Hidden Markov Model based stride segmentation on unsupervised free-living gait data in Parkinson’s disease patients. J NeuroEngineering Rehabil vol 18, (93), 2021.
[8] Sprager S., Juric M.B. Robust Stride Segmentation of Inertial Signals Based on Local Cyclicity Estimation. Sensors vol 18, 2018.
[9] Ying, H., Silex, C., Schnitzer, A., Leonhardt, S. and Schiek, M. Automatic step detection in the accelerometer signal. Proceedings of the 4th International Workshop on Wearable and Implantable Body Sensor Networks, 2007.
[10] Sekine, M., Tamura, T., Akay, M., Fujimoto, T., Togawa, T. and Fukui, Y. Discrimination of walking patterns using wavelet-based fractal analysis. IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol 10, 2002.
[11] Prateek, G. V., Mazzoni, P., Earhart, G. M. and Nehorai, A. Gait Cycle Validation and Segmentation Using Inertial Sensors in IEEE Transactions on Biomedical Engineering, 2020.
[12] Vienne-Jumeau A., Oudre L., Moreau A., et al. Personalized Template-Based Step Detection From Inertial Measurement Units Signals in Multiple Sclerosis Frontiers in Neurology, 2020.
[13] Barth, J., Oberndorfer, C., Pasluosta, C, Schülein, S., Gaßner, H., Reinfelder, S., Kugler, P., Schuldhaus, D., Winkler, J., Klucken, J., Eskofier, B. Stride Segmentation during Free Walk Movements Using Multi-Dimensional Subsequence Dynamic Time Warping on Inertial Sensor Data. Sensors vol 15, 2015.