Shoulder physiotherapy exercise recognition: machine learning the inertial signals from a smartwatch

Burns, David M.; Leung, Nathan; Hardisty, Michael; Whyne, Cari; Henry, Patrick; McLachlin, Stewart D.

doi:10.1088/1361-6579/aacfd9

Cited by 92 publications

(91 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…This study found that machine learning methods could accurately classify a variety of upper extremity exercises using biomechanical data from an IMU-based device. The highest accuracy (98.6%) was attained using a random forest method-a level of accuracy surpassing the study goal of 90% and similar to the accuracies (96.85% to 97.5%) attained with other IMU devices [23], [24] and to the accuracies (95.2% to 98.3% accuracy) attained with RGB-D cameras [10], [25]. The greater performance in this study compared to previous studies could be explained by the additional IMUs allowing for additional joint kinematic data that is not ascertainable with simpler sensor designs.…”

Section: Discussionsupporting

confidence: 61%

“…The objectives of this study are: 1) to evaluate machine learning models for classifying nine different upper extremity exercises, based upon biomechanics captured from the IMU-based device, and 2) to determine the effect of various train-test splits and sampling frequencies (64 Hz, 11 Hz, and 5 Hz) of the IMU device on classifier performance. As high accuracy (>89%) has been reported when using a smartwatch to classify shoulder exercises [23], we sought to exceed a 90% classification accuracy as our device incorporated multiple sensors.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Evaluation of Machine Learning Models for Classifying Upper Extremity Exercises Using Inertial Measurement Unit-Based Kinematic Data

Hua

Chaudhari

Johnson

et al. 2020

IEEE J. Biomed. Health Inform.

View full text Add to dashboard Cite

The amount of home-based exercise prescribed by a physical therapist is difficult to monitor. However, the integration of wearable inertial measurement unit (IMU) devices can aid in monitoring home exercise by analyzing exercise biomechanics. The objective of this study is to evaluate machine learning models for classifying nine different upper extremity exercises, based upon kinematic data captured from an IMU-based device. Fifty participants performed one compound and eight isolation exercises with their right arm. Each exercise was performed ten times for a total of 4500 trials. Joint angles were calculated using IMUs that were placed on the hand, forearm, upper arm, and torso. Various machine learning models were developed with different algorithms and train-test splits. Random forest models with flattened kinematic data as a feature had the greatest accuracy (98.6%). Using triaxial joint range of motion as the feature set resulted in decreased accuracy (91.9%) with faster speeds. Accuracy did not decrease below 90% until training size was decreased to 5% from 50%. Accuracy decreased (88.7%) when splitting data by participant. Upper extremity exercises can be classified accurately using kinematic data from a wearable IMU device. A random forest classification model was developed that quickly and accurately classified exercises. Sampling frequency and lower training splits had a modest effect on Manuscript

show abstract

Section: Discussionsupporting

confidence: 61%

Section: Introductionmentioning

confidence: 99%

Evaluation of Machine Learning Models for Classifying Upper Extremity Exercises Using Inertial Measurement Unit-Based Kinematic Data

Hua

Chaudhari

Johnson

et al. 2020

IEEE J. Biomed. Health Inform.

View full text Add to dashboard Cite

show abstract

“…Our search found 13 systems have been developed since the year 2000, which appears to be a low number, considering the importance of TEs in Physiotherapy rehabilitation and the advances in technology over the last two decades. A multitude of research studies exists surrounding the creation of digital systems for targeted exercise detection and classification using a variety of motion sensors [32,[54][55][56][57][58]. Challenges to creating such systems are noted in this body of literature and may explain why so few systems were identified.…”

Section: Systems Identifiedmentioning

confidence: 99%

Feedback Design in Targeted Exercise Digital Biofeedback Systems for Home Rehabilitation: A Scoping Review

Brennan

Dorronzoro

Caulfield

2019

Sensors

View full text Add to dashboard Cite

Digital biofeedback systems (DBSs) are used in physical rehabilitation to improve outcomes by engaging and educating patients and have the potential to support patients while doing targeted exercises during home rehabilitation. The components of feedback (mode, content, frequency and timing) can influence motor learning and engagement in various ways. The feedback design used in DBSs for targeted exercise home rehabilitation, as well as the evidence underpinning the feedback and how it is evaluated, is not clearly known. To explore these concepts, we conducted a scoping review where an electronic search of PUBMED, PEDro and ACM digital libraries was conducted from January 2000 to July 2019. The main inclusion criteria included DBSs for targeted exercises, in a home rehabilitation setting, which have been tested on a clinical population. Nineteen papers were reviewed, detailing thirteen different DBSs. Feedback was mainly visual, concurrent and descriptive, frequently providing knowledge of results. Three systems provided clear rationale for the use of feedback. Four studies conducted specific evaluations of the feedback, and seven studies evaluated feedback in a less detailed or indirect manner. Future studies should describe in detail the feedback design in DBSs and consider a robust evaluation of the feedback element of the intervention to determine its efficacy.Sensors 2020, 20, 181 2 of 20 to improve motor learning, engagement with and adherence to rehabilitation in musculoskeletal, neurological and orthopaedic populations [3][4][5]. They can improve awareness of exercise technique by providing information regarding movement that was previously unavailable. DBSs can generate objective measurements which act as rehabilitation goals, such as knee range of movement (ROM). With the appropriate data analytics and visualisations, the data collected can be compiled to create progress reports and allow remote monitoring by physiotherapists [6,7].Feedback consists of four main components (mode, content, frequency and timing), which each may be applied in a multitude of ways. Different combinations of these components will result in different feedback designs, and, theoretically, different effects on the user [8]. We have summarised the components of feedback that are observed in the relevant literature and presented them in Figure 1 as a framework within which to describe feedback interventions in detail. The first of these components is feedback mode, this may be visual, audio, haptic or multimodal (more than one mode represents the same variable at the same time). Through these modes, the feedback may be presented in a direct manner, e.g., a measure of ROM [9], or in an indirect manner, e.g., abstract graphical displays or gamified interfaces [10,11]. Feedback timing can be concurrent (also known as 'real-time' or 'simultaneous') or terminal, meaning it can be delivered during or after exercise execution, respectively. The frequency of feedback can be constant, reduced, or fading (decreasing over time). Feedback con...

show abstract

“…Sin embargo, el seguimiento de estos protocolos es a menudo deficiente, y más aún en los programas de ejercicio en casa, ya que las herramientas para dicha tarea son muy limitadas y carecen de supervisión. Una opción, es el desarrollo de un sistema de monitoreo de la terapia de hombro usando un reloj inteligente como dispositivo de adquisición de datos (Burns et al, 2018). Los algoritmos empleados para clasificar los ejercicios de hombro son: k-vecinos más cercanos, bosques aleatorios, SVM y con el mejor desempeño Redes Neuronales Recurrentes (CRNN, por sus siglas en ingles), logrando una exactitud del 99.4%.…”

Section: Aplicaciones Del Aprendizaje Automático En La Medicina Físicunclassified

Aprendizaje Automático en Aplicaciones Fisioterapéuticas

González-Islas

2019

ICBI

View full text Add to dashboard Cite

El aprendizaje automático (AA) o aprendizaje máquina se centra en el desarrollo de sistemas de análisis que aprenden de datos existentes para predecir o agrupar datos futuros. El AA se ha desarrollado sustancialmente en años recientes en muchas áreas de la ciencia y de la ingeniería, siendo una de las de mayor interés el área de la medicina. En los ambientes médicos reales existe una gran cantidad de datos de naturaleza multivariable y multidimensional, que requieren analizarse para diagnóstico y tratamiento médico, lo que representa un área de oportunidad importante para el aprendizaje máquina. Este trabajo presenta una revisión de algunas de las aplicaciones más relevantes del aprendizaje automático, así como las tendencias y expectativas emergentes del mismo. Posteriormente, se hace una descripción de los trabajos más significativos del aprendizaje máquina aplicados a la medicina, haciendo especial énfasis en aplicaciones fisioterapéuticas. Finalmente, un tema de particular interés en el contexto de la medicina física es el análisis de la marcha, por lo que se hace una revisión de los trabajos realizados para dicho propósito.

show abstract

Shoulder physiotherapy exercise recognition: machine learning the inertial signals from a smartwatch

Abstract: This proof of concept study demonstrates the technical feasibility of a smartwatch device and supervised machine learning approach to more easily monitor and assess the at-home adherence of shoulder physiotherapy exercise protocols.

Cited by 92 publications

References 29 publications

Evaluation of Machine Learning Models for Classifying Upper Extremity Exercises Using Inertial Measurement Unit-Based Kinematic Data

Evaluation of Machine Learning Models for Classifying Upper Extremity Exercises Using Inertial Measurement Unit-Based Kinematic Data

Feedback Design in Targeted Exercise Digital Biofeedback Systems for Home Rehabilitation: A Scoping Review

Aprendizaje Automático en Aplicaciones Fisioterapéuticas

Contact Info

Product

Resources

About