Patrick Höfner

Advisors:
Malte Ollenschläger (M.Sc.), Martin Ullrich (M.Sc.),  Prof. Dr. Björn Eskofier

Duration:
04/2021 – 10/2021

Abstract:
Hereditary Spastic Paraplegia (HSP) is a neurodegenerative disease that causes spasticity predominantly in adductors and calves. This is phenotypically expressed as scissor gait and / or foot drop, which in turn lead to gait impairment and eventually loss of ambulation. Thus, the increasing loss of ambulation becomes the highest perceived burden for patients as the disease progresses [1]. Disease progression and severity are commonly assessed using the Spastic Paraplegia Rating Scale (SPRS), which is rater dependent and therefore alternative methods need to be developed.
Previous studies showed that optical motion capture systems can be used to detect significant differences between participants with drop foot condition as compared to control participants as well as between subgroups of HSP patients [2, 3]. The study results show a reduction of range of motion in hip, knee, and foot joints. However, the mere assessment of significant group differences does not allow for a disease severity classification based on sensor signals. Thus, no decision support system is available, which assists the clinician to assess patient status objectively and remotely. Moreover, optical motion capture systems cannot be used for remote assessment as for example in a home environment, although these scenarios become increasingly important in treatment and clinical trials [4]. Therefore, there is a need for patient classification based on range of motion measured with wearable sensors.
Automated classification of foot drop based on range of motion signals from wearable sensors is fundamentally possible, as shown in a previous study by Bibadabi et al. [5]. They recorded signals of inertial measurement units (IMUs) of participants with ankle-dorsi exion weakness with L5 radiculopathy origins before and after a lumbar spine surgery. Afterwards, they successfully classified the sensor signals to being recorded before or after surgery. However, the classification was not performed with respect to a clinical annotation of foot drop but only with respect to pre-/post-surgery, which limits its generalizability. Furthermore, the results of this study cannot be applied to HSP patients because L5 radiculopathy results in accid paresis while HSP results in spastic paresis. The available data set for this thesis contains IMU-sensor data of approximately 30 HSP patients with different levels of spasticity. Gait trials were labeled by two medical experts on a three-level rating scale according to the severity of foot drop. Furthermore, SPRS annotations are available.
In this work, a measure for foot drop in HSP patients will be developed based on the described data set to serve as primer to calf spasticity. After segmenting data into single strides as suggested by Martindale et. al [6], stride-based features will be calculated. Features suggested in literature [5], as well as generic and handcrafted features will be used. Subsequently, patients will be classified using state of the art classification techniques. As optional part of this thesis, the student can assess the relation between the developed foot drop classification with stride length estimation or with SPRS.
The master thesis must contain a literature research on characteristics of range of motion signals for foot drop patients and algorithms for estimation of foot drop using inertial measurement units. This will be performed with a focus on – but not limited to – the domain of spastic gait. Afterwards, features must be calculated based on previously segmented strides and the below mentioned classifiers have to be implemented. The evaluation of their performances as well as feature importance are the main outcomes of this thesis. The thesis must contain a detailed description of all developed and used algorithms as well as a profound result evaluation and discussion. The implemented code has to be documented, tested and provided. An extended research on literature, existing patents and related work in the corresponding areas has to be performed.

References:
[1] van Lith, B. J. H. et al.: Experienced complaints, activity limitations and loss of motor capacities in patients with pure hereditary spastic paraplegia: a web-based survey in the Netherlands. Orphant Journal of Rare Diseases, vol 15, 2020.
[2] Wiszomirska, Ida et al. Eect of Drop Foot on Spatiotemporal, Kinematic, and Kinetic Parameters during Gait. Applied bionics and biomechanics vol. 2017.
[3] Serrao, M. et al. Gait Patterns in Patients with Hereditary Spastic Paraparesis. PLoS One, vol 11 (10) 2016.
[4] Eggers, Carsten et al. Care of patients with Parkinson’s disease in Germany: status quo and perspectives as re ected in the digital transition. Der Nervenarzt, 1{8, 2020.
[5] Bibadabi, S.S. et al. Tracking Foot Drop Recovery Following Lumbar-Spine Surgery, Applying Multiclass Gait Classification Using Machine Learning Techniques. MDPI Sensors, vol 19 (11), 2019.
[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), 2020.
[7] Martindale, C. F. et al.: Mobile Gait Analysis using Personalised Hidden Markov Models for Hereditary Spastic Paraplegia Patients. IEEE Engineering in Medicine and Biology Society, 2018.