Activity recognition in patients with lower limb impairments: Do we need training data from each patient?

Lonini, Luca; Gupta, Aakash; Kording, Konrad; Jayaraman, Aran

doi:10.1109/embc.2016.7591425

Cited by 18 publications

(27 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our results for the stroke population align with previous work validating AR reliability for persons with neurological injury when using training data from a young, healthy cohort [14-16,18]. We have extended this analysis to stroke by comparing AR across levels of gait impairment, finding that the healthy-trained model increasingly underestimated ambulation with impairment severity.…”

Section: Discussionsupporting

confidence: 84%

“…Recent work has demonstrated that AR classifiers trained using in-lab data from young, able-bodied adults do not generalize to older adults [14], persons with Parkinson’s disease [15], or patients with lower limb impairments [16]. Rather, using training data from the neurological population of interest notably improved AR accuracy, likely because such groups exhibit different movement patterns than a young, healthy cohort [17].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Activity Recognition for Persons With Stroke Using Mobile Phone Technology: Toward Improved Performance in a Home Setting

O’Brien¹,

Shawen²,

Mummidisetty³

et al. 2017

J Med Internet Res

Self Cite

View full text Add to dashboard Cite

BackgroundSmartphones contain sensors that measure movement-related data, making them promising tools for monitoring physical activity after a stroke. Activity recognition (AR) systems are typically trained on movement data from healthy individuals collected in a laboratory setting. However, movement patterns change after a stroke (eg, gait impairment), and activities may be performed differently at home than in a lab. Thus, it is important to validate AR for gait-impaired stroke patients in a home setting for accurate clinical predictions.ObjectiveIn this study, we sought to evaluate AR performance in a home setting for individuals who had suffered a stroke, by using different sets of training activities. Specifically, we compared AR performance for persons with stroke while varying the origin of training data, based on either population (healthy persons or persons with stoke) or environment (laboratory or home setting).MethodsThirty individuals with stroke and fifteen healthy subjects performed a series of mobility-related activities, either in a laboratory or at home, while wearing a smartphone. A custom-built app collected signals from the phone’s accelerometer, gyroscope, and barometer sensors, and subjects self-labeled the mobility activities. We trained a random forest AR model using either healthy or stroke activity data. Primary measures of AR performance were (1) the mean recall of activities and (2) the misclassification of stationary and ambulatory activities.ResultsA classifier trained on stroke activity data performed better than one trained on healthy activity data, improving average recall from 53% to 75%. The healthy-trained classifier performance declined with gait impairment severity, more often misclassifying ambulatory activities as stationary ones. The classifier trained on in-lab activities had a lower average recall for at-home activities (56%) than for in-lab activities collected on a different day (77%).ConclusionsStroke-based training data is needed for high quality AR among gait-impaired individuals with stroke. Additionally, AR systems for home and community monitoring would likely benefit from including at-home activities in the training data.

show abstract

Section: Discussionsupporting

confidence: 84%

Section: Introductionmentioning

confidence: 99%

Activity Recognition for Persons With Stroke Using Mobile Phone Technology: Toward Improved Performance in a Home Setting

O’Brien¹,

Shawen²,

Mummidisetty³

et al. 2017

J Med Internet Res

Self Cite

View full text Add to dashboard Cite

show abstract

“…On the other hand, global models are arguably easier to deploy, as they do not require collecting data on each and every new patient [ 14 ]. Interestingly, in our scenario, personal device-specific models surpassed global models only for identifying stair-climbing activities, while being equally accurate at detecting walking.…”

Section: Discussionmentioning

confidence: 99%

“…B. Experimental setup, data processing, and activity recognition steps (adapted with permission from [ 14 ]). A patient performed a set of activities while wearing a KAFO and a triaxial accelerometer.…”

Section: Methodsmentioning

confidence: 99%

“…The majority of wearable- and mobile phone–based AR studies have been conducted using healthy individuals, whereas relatively fewer studies are focused on people with disabilities [ 6 ], such as those resulting from stroke [ 7 - 9 ] or Parkinson disease [ 10 , 11 ]. Some of these studies showed that a model trained on young healthy individuals will yield poor performance when used with a different population [ 9 , 11 - 13 ], including those who need an assistive device for walking [ 14 ]. These differences arise due to the fact that movements are unique to individuals, and movements in people with a disability are different from that of able-bodied individuals [ 15 ].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Activity Recognition in Individuals Walking With Assistive Devices: The Benefits of Device-Specific Models

Lonini

Gupta

Deems-Dluhy

et al. 2017

JMIR Rehabil Assist Technol

Self Cite

View full text Add to dashboard Cite

BackgroundWearable sensors gather data that machine-learning models can convert into an identification of physical activities, a clinically relevant outcome measure. However, when individuals with disabilities upgrade to a new walking assistive device, their gait patterns can change, which could affect the accuracy of activity recognition.ObjectiveThe objective of this study was to assess whether we need to train an activity recognition model with labeled data from activities performed with the new assistive device, rather than data from the original device or from healthy individuals.MethodsData were collected from 11 healthy controls as well as from 11 age-matched individuals with disabilities who used a standard stance control knee-ankle-foot orthosis (KAFO), and then a computer-controlled adaptive KAFO (Ottobock C-Brace). All subjects performed a structured set of functional activities while wearing an accelerometer on their waist, and random forest classifiers were used as activity classification models. We examined both global models, which are trained on other subjects (healthy or disabled individuals), and personal models, which are trained and tested on the same subject.ResultsMedian accuracies of global and personal models trained with data from the new KAFO were significantly higher (61% and 76%, respectively) than those of models that use data from the original KAFO (55% and 66%, respectively) (Wilcoxon signed-rank test, P=.006 and P=.01). These models also massively outperformed a global model trained on healthy subjects, which only achieved a median accuracy of 53%. Device-specific models conferred a major advantage for activity recognition.ConclusionsOur results suggest that when patients use a new assistive device, labeled data from activities performed with the specific device are needed for maximal precision activity recognition. Personal device-specific models yield the highest accuracy in such scenarios, whereas models trained on healthy individuals perform poorly and should not be used in patient populations.

show abstract

Wearable sensors for Parkinson’s disease: which data are worth collecting for training symptom detection models

et al. 2018

Self Cite

View full text Add to dashboard Cite

Machine learning algorithms that use data streams captured from soft wearable sensors have the potential to automatically detect PD symptoms and inform clinicians about the progression of disease. However, these algorithms must be trained with annotated data from clinical experts who can recognize symptoms, and collecting such data are costly. Understanding how many sensors and how much labeled data are required is key to successfully deploying these models outside of the clinic. Here we recorded movement data using 6 flexible wearable sensors in 20 individuals with PD over the course of multiple clinical assessments conducted on 1 day and repeated 2 weeks later. Participants performed 13 common tasks, such as walking or typing, and a clinician rated the severity of symptoms (bradykinesia and tremor). We then trained convolutional neural networks and statistical ensembles to detect whether a segment of movement showed signs of bradykinesia or tremor based on data from tasks performed by other individuals. Our results show that a single wearable sensor on the back of the hand is sufficient for detecting bradykinesia and tremor in the upper extremities, whereas using sensors on both sides does not improve performance. Increasing the amount of training data by adding other individuals can lead to improved performance, but repeating assessments with the same individuals—even at different medication states—does not substantially improve detection across days. Our results suggest that PD symptoms can be detected during a variety of activities and are best modeled by a dataset incorporating many individuals.

show abstract

Activity recognition in patients with lower limb impairments: Do we need training data from each patient?

Cited by 18 publications

References 15 publications

Activity Recognition for Persons With Stroke Using Mobile Phone Technology: Toward Improved Performance in a Home Setting

Activity Recognition for Persons With Stroke Using Mobile Phone Technology: Toward Improved Performance in a Home Setting

Activity Recognition in Individuals Walking With Assistive Devices: The Benefits of Device-Specific Models

Wearable sensors for Parkinson’s disease: which data are worth collecting for training symptom detection models

Contact Info

Product

Resources

About