Photo Credit: Andrey Popov
Recently-published findings state that a combination of machine learning algorithms and gait analysis was found to effectively predict fear of falling in MS.
Findings published in Multiple Sclerosis indicate that the combination of machine learning (ML) algorithms and gait analysis could efficiently predict fear of falling (FOF) in individuals with MS (pwMS).
“The progressive impairment of gait is one of the most important pathognomic symptoms which are associated with falls and FOF in pwMS,” researchers wrote. “[Sixty percent] of pwMS show a FOF, which leads to restrictions in mobility as well as physical activity and reduces the quality of life in general.”
Study Design
The cross-sectional study aimed to ascertain the most appropriate clinical gait analysis methods to detect FOF in pwMS and identify the ideal ML algorithms to predict FOF.
Between November 2020 and September 2021, the researchers reviewed gait data from 1,240 pwMS. Participants underwent multidimensional gait analysis utilizing pressure and motion sensors.
Patient-reported outcomes (PROs) were employed to evaluate walking impairments based on the patient’s perspective, including “walking, running, and stair climbing ability, subjective effort, and the required concentration and walking speed, the frequency of tripping and the indication of avoidance of social activities.”
The study employed standardized gait analysis to collect data from pwMS, including parameters such as gait/walking speed, stride length, variability, and other measures linked to balance and stability. A feature selection ensemble (FS-Ensemble) was developed to improve the classification performance. ML algorithms, specifically support vector machines (SVM) and random forest (RF), were employed to examine and categorize the data to classify patients based on their likelihood of experiencing FoF.
Single and Dual Gait Tasks Detect FoF
The analysis revealed that 37% of pwMS had a FOF (n=458; average age, 51; 76% women; average Expanded Disability Status Scale [EDSS], 4.0). The FS-Ensemble improved classification performance in most cases, and the SVM exhibited the best performance of the four classification models in the detection of FOF. The PROs demonstrated the best F1 scores (Early Mobility Impairment Questionnaire F1=0.81; 12-item MS Scale F1=0.80).
The results indicate that the single (F1=0.60) and dual (F1=0.65) task gait tests are better at detecting FoF than the data sets of the postural control, according to investigators. “This confirms the assumption that cognition is associated with FOF and related to gait changes with dual-task challenges.”
