What causes gait disorders?
Machine learning helps to identify gait problems in people with multiple sclerosis
Anna Yudina, "Scientific Russia"
Scientists use machine learning to understand which problems are associated with multiple sclerosis and which are caused by other age-related diseases, according to a press release from the University of Illinois at Urbana-Champaign Machine learning helps spot gait problems in individuals with multiple sclerosis.
Article by Kaur et al. Predicting Multiple Sclerosis from Gait Dynamics Using an Instrumented Treadmill – A Machine Learning Approach is published in the journal IEEE Transactions on Biomedical Engineering – VM.
Monitoring the progression of gait problems associated with multiple sclerosis can be challenging in adults over the age of 50, because the clinician must distinguish between problems associated with multiple sclerosis and other age-related changes. Now researchers are combining gait data and machine learning to refine the tools used to monitor and predict disease progression.
A new study of this approach, conducted by Rachnit Kaur, a graduate student at the University of Illinois at Urbana Champaign, Manuel Hernandez, professor of kinesiology and public Health, and Richard Sowers, Professor of Industrial and entrepreneurial Engineering and mathematics.
Multiple sclerosis can manifest itself in different ways in about 2 million people worldwide, and walking problems are a common symptom. According to the study, about half of patients need help walking within 15 years after the onset of the disease.
"We wanted to understand the relationship between aging and concomitant changes associated with MS, and to find out if we can also distinguish between these two phenomena in older people with MS," Hernandez said. – Machine learning methods seem to work especially well when detecting complex hidden changes. We hypothesized that these analysis methods may also be useful for predicting sudden gait changes in people with MS."
Using an instrumental treadmill, the team collected gait data – normalized by body size and demographic characteristics – from 20 adults with MS and 20 elderly people without MS, corresponding to age, weight, height and gender. The participants walked at a comfortable pace for up to 75 seconds, while specialized software recorded gait events corresponding to the reaction forces of the earth and the position of the pressure center during each walk. The team extracted the characteristic spatial, temporal and kinetic features of each participant in their steps to study gait variations during each trial.
Changes in various gait characteristics, including a data feature called the butterfly diagram, helped the team detect differences in gait patterns between participants. The diagram gets its name from a butterfly-shaped curve created from the repetitive trajectory of the pressure center for several continuous steps while the subject is walking, and is associated with critical neurological functions, the study says.
"We are studying the effectiveness of a machine learning system based on gait dynamics in order to classify the steps of elderly people with MS from healthy control to generalization for various walking tasks and new subjects," Kaur said. – The proposed methodology is a step forward towards the development of an evaluation marker for medical professionals that allows predicting elderly people with MS, who are likely to have worsening symptoms in the near future."
According to Sauers, in future studies, more thorough studies can be carried out to cope with the problem of the small size of the study cohort.
"Biomechanical systems like walking are poorly modeled, making it difficult to identify problems in a clinical setting," Sauers said. "In this study, we are trying to draw conclusions from datasets that include many measurements of each person, but a small number of people. The results of this study have made significant progress in the field of clinical disease prediction strategies based on machine learning."
Portal "Eternal youth" http://vechnayamolodost.ru