Digital health technologies and machine learning augment patient reported outcomes to remotely characterise rheumatoid arthritis

Creagh, Andrew P.; Hamy, Valentin; Yuan, Hang; Mertes, Gert; Tomlinson, Ryan E.; Chen, Wen-Hung; Williams, R.; Llop, Christopher; Yee, Christopher; Duh, Mei Sheng; Garcia‐Gancedo, Luis; Clifton, David A.

doi:10.1101/2022.11.18.22282305

Cited by 5 publications

(2 citation statements)

References 94 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Multi-task selfsupervision automatically extracts the features relevant to motion by learning to discriminate different spatiotemporal transformations applied to the unlabelled 700,000 person-days of data. Self-supervised pre-training has been shown to help classify human activity recognition not just in healthy but clinical populations 35 . See Supplementary Section 1.2 for further details of the model development.…”

Section: Deep Learning Analysis Of Sleep Stages From Wrist-worn Accel...mentioning

confidence: 99%

Self-supervised learning of accelerometer data provides new insights for sleep and its association with mortality

Yuan,

Plekhanova,

Walmsley

et al. 2024

npj Digit. Med.

View full text Add to dashboard Cite

Sleep is essential to life. Accurate measurement and classification of sleep/wake and sleep stages is important in clinical studies for sleep disorder diagnoses and in the interpretation of data from consumer devices for monitoring physical and mental well-being. Existing non-polysomnography sleep classification techniques mainly rely on heuristic methods developed in relatively small cohorts. Thus, we aimed to establish the accuracy of wrist-worn accelerometers for sleep stage classification and subsequently describe the association between sleep duration and efficiency (proportion of total time asleep when in bed) with mortality outcomes. We developed a self-supervised deep neural network for sleep stage classification using concurrent laboratory-based polysomnography and accelerometry. After exclusion, 1448 participant nights of data were used for training. The difference between polysomnography and the model classifications on the external validation was 34.7 min (95% limits of agreement (LoA): −37.8–107.2 min) for total sleep duration, 2.6 min for REM duration (95% LoA: −68.4–73.4 min) and 32.1 min (95% LoA: −54.4–118.5 min) for NREM duration. The sleep classifier was deployed in the UK Biobank with 100,000 participants to study the association of sleep duration and sleep efficiency with all-cause mortality. Among 66,214 UK Biobank participants, 1642 mortality events were observed. Short sleepers (<6 h) had a higher risk of mortality compared to participants with normal sleep duration of 6–7.9 h, regardless of whether they had low sleep efficiency (Hazard ratios (HRs): 1.58; 95% confidence intervals (CIs): 1.19–2.11) or high sleep efficiency (HRs: 1.45; 95% CIs: 1.16–1.81). Deep-learning-based sleep classification using accelerometers has a fair to moderate agreement with polysomnography. Our findings suggest that having short overnight sleep confers mortality risk irrespective of sleep continuity.

show abstract

Section: Deep Learning Analysis Of Sleep Stages From Wrist-worn Accel...mentioning

confidence: 99%

Self-supervised learning of accelerometer data provides new insights for sleep and its association with mortality

Yuan,

Plekhanova,

Walmsley

et al. 2024

npj Digit. Med.

View full text Add to dashboard Cite

show abstract

“…However, traditional methods of measuring disease activity in RA depend on regular physical assessments by rheumatology clinicians, requiring frequent hospital clinic visits. As such, there is a need to remotely monitor individuals with RA in real-world settings rather than confining assessments solely to intermittent physician-administered assessments [4].…”

Section: Introductionmentioning

confidence: 99%

Exploring the Effect of Sampling Frequency on Real-World Mobility, Sedentary Behaviour, Physical Activity and Sleep Outcomes Measured with Wearable Devices in Rheumatoid Arthritis: Feasibility, Usability and Practical Considerations. (Preprint)

Sarvestan,

Baker,

Del Din

2024

Preprint

View full text Add to dashboard Cite

BACKGROUND Modern treat-to-target management of rheumatoid arthritis (RA) involves titration of drug therapy to achieve remission, requiring close monitoring of disease activity through frequent clinical assessments. Accelerometry offers a novel method for continuous remote monitoring of RA activity by capturing fluctuations in mobility, sedentary behaviours, physical activity and sleep patterns over prolonged periods, without the expense, inconvenience and environmental impact of extra hospital visits. OBJECTIVE We aimed to: (a) assess the feasibility, usability and acceptability of wearable devices in patients with active RA; (b) investigate the multivariate relationships within the dataset; and (c) explore robustness of accelerometry outcomes to downsampling to facilitate future prolonged monitoring. METHODS Eleven patients with active RA newly starting an arthritis drug completed clinical assessments at 4-week intervals for 12 weeks. Participants wore an Axivity AX6 wrist device (sampling frequency 100Hz) for 7 days after each clinical assessment. Measures of macro gait (volume, pattern and variability), micro gait (pace, rhythm, variability, asymmetry and postural control of walking), sedentary behaviour (standing, sitting and lying) and physical activity (moderate to vigorous physical activity [MVPA], sustained inactive bouts [SIBs]) and sleep outcomes (sleep duration, wake up after sleep onset, number of awakenings) were recorded. Feasibility, usability and acceptability of wearable devices was assessed using Rabinovich’s questionnaire, principal component (PC) analysis was used to investigate the multivariate relationships within the dataset, and Bland-Altman plots (bias and Limits of Agreement) and Intraclass Correlation Coefficient (ICC) were used to test robustness of outcomes sampled at 100Hz versus downsampled at 50Hz and 25Hz. RESULTS Wearable devices obtained high feasibility, usability and acceptability scores among participants. Macro gait outcomes and MVPA (first PC), and micro gait outcomes and number of SIBs (second PC) exhibited the strongest loadings, with these first two PCs accounting for 40% of the variance of the dataset. Furthermore, these devices metrics were robust to downsampling, showing good to excellent agreements (ICC≥0.75). CONCLUSIONS We identified two main domains of mobility, physical activity and sleep outcomes of people with RA: micro gait outcomes plus MVPA, and micro gait outcomes plus number of SIBs. Combined with high usability and acceptability of wearable devices, and robustness of outcomes to downsampling, our real-world data supports the feasibility of accelerometry for prolonged remote monitoring of RA disease activity.

show abstract

Self-supervised learning for human activity recognition using 700,000 person-days of wearable data

Yuan,

Chan,

Creagh

et al. 2024

npj Digit. Med.

View full text Add to dashboard Cite

Accurate physical activity monitoring is essential to understand the impact of physical activity on one’s physical health and overall well-being. However, advances in human activity recognition algorithms have been constrained by the limited availability of large labelled datasets. This study aims to leverage recent advances in self-supervised learning to exploit the large-scale UK Biobank accelerometer dataset—a 700,000 person-days unlabelled dataset—in order to build models with vastly improved generalisability and accuracy. Our resulting models consistently outperform strong baselines across eight benchmark datasets, with an F1 relative improvement of 2.5–130.9% (median 24.4%). More importantly, in contrast to previous reports, our results generalise across external datasets, cohorts, living environments, and sensor devices. Our open-sourced pre-trained models will be valuable in domains with limited labelled data or where good sampling coverage (across devices, populations, and activities) is hard to achieve.

show abstract

Digital health technologies and machine learning augment patient reported outcomes to remotely characterise rheumatoid arthritis

Cited by 5 publications

References 94 publications

Self-supervised learning of accelerometer data provides new insights for sleep and its association with mortality

Self-supervised learning of accelerometer data provides new insights for sleep and its association with mortality

Exploring the Effect of Sampling Frequency on Real-World Mobility, Sedentary Behaviour, Physical Activity and Sleep Outcomes Measured with Wearable Devices in Rheumatoid Arthritis: Feasibility, Usability and Practical Considerations. (Preprint)

Self-supervised learning for human activity recognition using 700,000 person-days of wearable data

Contact Info

Product

Resources

About