Freezing of gait (FOG) is a debilitating symptom of Parkinson's disease increases the risk of falling. Despite being a common symptom, it is still difficult to evaluate freezing of gait quickly and accurately. Currently, the gold-standard method to determine the severity of FOG is a manual analysis of video footage by an experienced assessor, collected during standardized FOG-provoking walking tests. Because this is a very time-intensive process, where different assessors sometimes obtain different results, our team at KU Leuven have developed an artificial-intelligent (AI) algorithm trained to identify FOG episodes based on wearable inertial measurement unit (IMU) sensor data. The AI algorithm has already undergone initial validation during laboratory testing, yielding promising results. The aim of this study is to investigate whether the AI algorithm can accurately detect FOG episodes in a less controlled environment, namely the home environment. In a second phase, the investigators will also use the collected data to improve the AI algorithm for automated FOG detection in the home. Finally, the investigators want to explore whether the AI algorithm can detect FOG in real-time.
Voluntary written informed consent of the participant has been obtained prior to any study-related procedures, except the non-recorded pre-screening questions;
At least 18 years of age at the time of signing the Informed Consent Form (ICF);
Person is cognitively able to follow and understand instructions and provide voluntary written informed consent;
Person is able to walk for short distances (± 10 meters) independently, with- or without use of a walking aid;
Occurrence of any of the following within 3 months prior to informed consent: myocardial infarction, hospitalization for unstable angina, stroke, coronary artery bypass graft (CABG), percutaneous coronary intervention (PCI), implantation of a cardiac resynchronization therapy device (CRTD), active treatment for cancer or other malignant disease, uncontrolled congestive heart disease (NYHA class >3), acute psychosis or major psychiatric disorders or continued substance abuse, other neurological (than PD) or orthopaedic impairment that significantly impacts on gait;
Participant self-reports daily falls;
Participation in another interventional study, with or without an investigational medicinal product (IMP) or device (IMD)
People with PD and healthy age-matched controls will be recruited from three primary sites, namely KU Leuven, Hamburg Medical Center, and Tel-Aviv Sourasky Medical Center.
Cohort
Prospective
Not listed
Comparing the agreement between AID-FOG and gold-standard expert annotation to detect the percentage of time spent with freezing of gait (FOG) in relation to total time duration (%TF).
The primary outcome (percentage of time spent with FOG in relation to total task duration = %TF) will be established by manual annotations of video footage by an experienced assessor (=gold-standard reference) and by the automated AID-FOG algorithm v1.0 applied post-hoc (i.e. offline) to IMU data collected during the same walking tasks. This will be calculated for standardized walking tasks on which the AID-FOG algorithm has been trained, standardized walking tasks on which the AID-FOG algorithm was not trained, and a free-living walking condition on which the AID-FOG algorithm was not trained.
F1-score
Same as primary outcome, but now for the F1-score (rather than percent TF). F1 scores range between 0 and 1, the higher the score the better.
Number of FOG episodes
Same as primary outcome, but now for the absolute number of FOG episodes.
The performance of the AID-FOG algorithm to differentiate between the FOG manifestations.
Freezing of gait (FOG) manifests in multiple forms, including akinetic and kinetic subtypes, which may be associated with trembling or occur without it. This study investigates the performance of AID-FOG in discriminating between these manifestations, using expert annotations as the reference standard.
Comparing performance of AID-FOG to detect freezing in OFF and ON medication states.
The FOG outcomes as obtained by the human expert and the offline AID-FOG algorithm v1.0 will be calculated for both the OFF and ON medication states. These scores will be compared to evaluate the change in algorithm performance depending on medication status. The FOG outcomes will be the percentage TF which ranges between 0-100 percent. The higher the percentage the more freezing the patient has. But also the F1-score which ranges between 0-1. The higher the score the more overlap there is between the expert and the algorithm.
AID-FOG version 2.0 (percentage TF)
The data obtained in the study will be used to train the AID-FOG algorithm v1.0. This trained AID-FOG algorithm v2.0 will be evaluated using the same listed outcome measures, using a leave-one-subject-out approach. Percentage of time spent with FOG in relation to total task duration (= %TF) will be established by manual annotations of video footage by an experienced assessor (=gold-standard reference) and by the automated AID-FOG algorithm v2.0 applied post-hoc (i.e. offline) to IMU data collected during the same walking tasks. This will be calculated for standardized walking tasks on which the AID-FOG algorithm has been trained, standardized walking tasks on which the AID-FOG algorithm was not trained, and a free-living walking condition on which the AID-FOG algorithm was not trained.
AID-FOG version 2.0 (F1-score)
The data obtained in the study will be used to train the AID-FOG algorithm v1.0. This trained AID-FOG algorithm v2.0 will be evaluated using the same listed outcome measures, using a leave-one-subject-out approach. Same as percentage TF, but now for the F1-score (rather than percent TF). F1 scores range between 0 and 1, the higher the score the better.
AID-FOG version 2.0 (Number of FOG episodes)
The data obtained in the study will be used to train the AID-FOG algorithm v1.0. This trained AID-FOG algorithm v2.0 will be evaluated using the same listed outcome measures, using a leave-one-subject-out approach. Same as percentage TF, but now for the absolute number of FOG episodes.
AID-FOG version 2.0 (FOG manifestations)
The data obtained in the study will be used to train the AID-FOG algorithm v1.0. This trained AID-FOG algorithm v2.0 will be evaluated using the same listed outcome measures, using a leave-one-subject-out approach. The performance of the AID-FOG algorithm v2.0 to discriminate between the FOG manifestations. Freezing of gait (FOG) manifests in multiple forms, including akinetic and kinetic subtypes, which may be associated with trembling or occur without it. This study investigates the performance of AID-FOG in discriminating between these manifestations, using expert annotations as the reference standard.
Click on the lead sponsor to view all of their trials.
KU Leuven
Lead sponsor
Michael J. Fox Foundation for Parkinson's Research
Collaborator
Tel Aviv Medical Center
Collaborator
Medical School Hamburg
Collaborator
