Artificial intelligence is being used by researchers at the Department of Physical Education’s Human Movement Laboratory (Movi-Lab) at São Paulo State University (UNESP) in Bauru, Brazil, to assist in the diagnosis and prediction of the progression of Parkinson's disease.
The results of the study, in which the disease was detected using machine learning algorithms by examining spatial and temporal gait parameters, are detailed in an article published in the journal Gait & Posture.
Step length, velocity, width, and consistency (or width variability) were discovered by the researchers to be the four gait features most crucial for the diagnosis of Parkinson’s.
Step width variability and double support time (during which both feet are in contact with the ground) were the most important variables to consider when determining the disease’s severity.
Our study innovated in comparison with the scientific literature by using a larger database than usual for diagnostic purposes. We chose gait parameters as the key criteria because gait impairments appear early in Parkinson’s and get worse over time, and also because they don’t correlate with physiological parameters like age, height and weight.
Fabio Augusto Barbieri, Study Co-Author and Professor, Department of Physical Education, School of Sciences, Universidade Estadual Paulista
FAPESP provided funding for the study through three projects (14/20549-0, 17/19516-8, and 20/01250-4).
A total of 63 patients with Parkinson's disease and 63 healthy controls made up the study sample. The patients were enrolled in FC-UNESP’s multidisciplinary program Ativa Parkinson. Each volunteer was older than 50. For seven years, data was gathered and fed into the repository that was used in the machine learning processes.
By examining gait parameters for healthy controls and contrasting them with levels predicted for this age group, a baseline assessment was created. This involved measuring each person's strides for length, width, duration, velocity, cadence, and single and double support time, as well as step variability and asymmetry, using a specialized motion capture camera.
The data was used by the researchers to develop two distinct machine-learning models, one for disease diagnosis and the other for estimating the severity of the disease in the patient under consideration. In this portion of the study, researchers from the School of Engineering at the University of Porto in Portugal worked together.
They subjected the data to six algorithms: Logistic Regression (LR), Naïve Bayes (NB), Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), and Multilayer Perceptron (MLP). NB had a diagnostic accuracy of 84.6%, and NB and RF were the best at determining severity.
Barbieri added, “Typical accuracy for clinical assessments is around 80%. We could significantly reduce the probability of diagnostic error by combining clinical assessment with artificial intelligence.”
Parkinson’s disease is caused, at least in part, by the degeneration of nerve cells in the parts of the brain that control movement due to inadequate dopamine production. A neurotransmitter called dopamine is responsible for sending signals to the limbs.
Movement is hampered by low dopamine levels, which also affect speech and writing and cause tremors, a sluggish gait, rigidity, and poor balance.
The patient’s clinical history and a neurological examination are currently used to make the diagnosis; no specific tests are performed. There is no precise data available, but Parkinson’s disease is thought to affect 3–4% of people over 65.
The study’s findings will be helpful to improve diagnostic assessment in the future, but the cost could be a deterrent, according to another co-author, Ph.D. candidate Tiago Penedo, whose research is under Barbieri’s supervision.
We made progress with the tool and contributed to expansion of the database, but we used expensive equipment that is hard to find in clinics and doctor’s offices.
Tiago Penedo, Study Co-Author and PhD Student, Universidade Estadual Paulista
About $100,000 worth of equipment was used in the study.
“It is possible to analyze gait with cheaper techniques, using a chronometer, force plate and so on, but the results are not precise,” Penedo stated.
The methods used in the study, particularly the gait patterns, could aid in a better understanding of the disease’s underlying mechanisms, according to the researchers.
An earlier study showed that Parkinson’s patients had 53% shorter step-length synergy while navigating obstacles than healthy subjects of the same age and weight, according to a 2021 article with Barbieri as the last author.
Synergy in this context refers to the ability of the locomotor (or musculoskeletal) system to adapt movement by fusing elements like speed and foot position, for example, when stepping off a curb.
Another study, also published in Gait & Posture, revealed that Parkinson’s patients were less able than their neurologically healthy peers to maintain postural control and rambling-trembling stability. The findings, according to the authors, offered fresh explanations for the larger, swifter, and more variable sway observed in Parkinson’s patients.
Ferreira, M. I. A. S. N., et al. (2022) Machine learning models for Parkinson’s disease detection and stage classification based on spatial-temporal gait parameters. Gait & Posture. doi:10.1016/j.gaitpost.2022.08.014.