Algorithm for Wearable Sensors Ably Measures Tremor Severity as Patients Go About Daily Life, Study Says

Catarina Silva, MSc avatar

by Catarina Silva, MSc |

Share this article:

Share article via email
Stada, triple combo therapy

Researchers have developed algorithms that work with wearable sensors to continuously monitor tremor, and estimate total tremor, in Parkinson’s patients as they go about their daily routines.

Analyses of sensor results using one algorithm, in particular, were similar to an established test assessing tremor without being dependent on the time the test is given.

The study, “Wearable Sensors for Estimation of Parkinsonian Tremor Severity during Free Body Movements,” was published in Sensors.

Resting tremor, or the rhythmic shaking of muscles while relaxed, is among the motor symptoms of Parkinson’s disease (PD), and some patients also have active tremor, or shaking while engaged in voluntary muscle movement. Others motor symptoms are slowness of movement (bradykinesia), rigidity, and problems with posture, balance, and gait.

Currently, Parkinson’s motor symptoms are assessed using the Unified Parkinson’s Disease Rating Scale (UPDRS) Part III, scores of whose tests (like the finger-to-nose test) are evaluated by doctors. This test requires office visits where the tasks are performed, providing essentially only a snapshot of a person’s tremor experience in day-to-day life.

“A single, clinical examination in a doctor’s office often fails to capture a patient’s complete continuum of tremors in his or her routine daily life,” Behnaz Ghoraani, PhD, an assistant professor at Florida Atlantic University’s (FAU) Institute for Sensing and Embedded Network Systems (I-SENSE) and FAU’s Brain Institute (I-BRAIN), and lead author of the study, said in a press release.

“Wearable sensors, combined with machine-learning algorithms, can be used at home or elsewhere to estimate a patient’s severity rating of tremors based on the way that it manifests itself in movement patterns,” Ghoraani added.

Investigators developed two distinct machine-learning algorithms that, when combined with wearable sensors, could estimate total Parkinsonian tremor as patients performed a variety of free body movements.

In a collaboration between FAU, the Icahn School of Medicine at Mount Sinai and the University of Rochester Medical Center, researchers developed two algorithms: gradient tree boosting and long short-term memory (LSTM)-based deep learning. These tools can estimate tremor severity both in a resting and action state.  

A total of 24 Parkinson’s patients (10 women and 14 men; mean age, 58.9) had data on their movements recorded in two studies.

“In both protocols, the subjects stopped their medication the night before the experiment, and the experiment started in the morning. Yet, if the subjects were unable to withdraw their medication overnight, they came to the laboratory near the time of a scheduled dose of their PD medication,” the researchers noted.

For the experiment, doctors placed one motion sensor (consisting of a gyroscope and an accelerometer) on patients’ wrist and another on the ankle of the most disease-affected body side. Movement was recorded as they went about daily life activities.

Fifteen individuals were instructed to perform four rounds of specific activities, like walking, resting, eating, drinking, dressing, combing hair, putting groceries on a table, and cutting food. Motion data were recorded only while performing these activities.

Patients then took their routine “morning dose” of Parkinson’s medications. Later, they repeated the same activities at the start of every hour for up to four hours. For comparison purposes, standard UPDRS-based assessment were also given preformed every hour of the testing period before each round of daily life activities.

The other nine patients had their motion recorded continuously for the entire experiment. These people were instructed to cycle through six stations in a home-like setting, performing tasks like personal hygiene, dressing, eating, desk work, entertainment, and laundry. They took their medications after finishing a first round of activities, and when the treatment kicked in, they repeated the previous exercises. This set of activities lasted up to two hours, and the motor part of UPDRS (Part III) was assessed before and after the experiment.

“The data from the 15 subjects who performed rounds of specific [activities of daily living] was used to train the [artificial intelligence] models, and the data from the remaining nine subjects who performed continuous [daily life activities] were held out for testing the models,” the researchers wrote.

Results revealed the gradient tree boosting method estimated total tremor as well as resting tremor with high accuracy, and in most cases, with the same results found by doctors scoring UPDRS Part III.

Importantly, gradient tree boosting-based sensors were able to detect decreases in tremors after patients took their medication, even in cases where results did not match total tremor sub-scores from the UPDRS assessments. The LSTM-based algorithm was less effective in doing the same.

“These results indicate that our approach holds great promise in providing a full spectrum of the patients’ tremor from continuous monitoring of the subjects’ movement in their natural environment,” the researchers wrote.

“It is especially interesting that the method we developed successfully detected hand and leg tremors using only one sensor on the wrist and ankle, respectively,” said Murtadha Hssayeni, a study co-author and a PhD student at FAU’s Department of Computer and Electrical Engineering and Computer Science.

The new gradient tree boosting algorithm combined with wearable sensor technology resulted  “in the highest correlation … reported in the literature when using unconstrained body movements’ data,” the researchers wrote.

“This finding is important because our method is able to provide a better temporal resolution to estimate tremors to provide a measure of the full spectrum of tremor changes over time,” Ghoraani added.