Detection of Unilateral Arm Paresis after Stroke by Wearable Accelerometers and Machine Learning

The use of accelerometer bracelets to evaluate arm motor function over a stroke rehabilitation period - an explorative observational study

BACKGROUND

Assessments of arm motor function are usually based on clinical examinations or self-reported rating scales. Wrist-worn accelerometers can be a good complement to measure movement patterns after stroke. Currently there is limited knowledge of how accelerometry correlate to clinically used scales. The purpose of this study was therefore to evaluate the relationship between intermittent measurements of wrist-worn accelerometers and the patient's progression of arm motor function assessed by routine clinical outcome measures during a rehabilitation period.

METHODS

Patients enrolled in in-hospital rehabilitation following a stroke were invited. Included patients were asked to wear wrist accelerometers for 24 h at the start (T1) and end (T2) of their rehabilitation period. On both occasions arm motor function was assessed by the modified Motor Assessment Scale (M_MAS) and the Motor Activity Log (MAL). The recorded accelerometry was compared to M_MAS and MAL.

RESULTS

20 patients were included, of which 18 completed all measurements and were therefore included in the final analysis. The resulting Spearman's rank correlation coefficient showed a strong positive correlation between measured wrist acceleration in the affected arm and M-MAS and MAL values at T1, 0.94 (p < 0.05) for M_MAS and 0.74 (p < 0.05) for the MAL values, and a slightly weaker positive correlation at T2, 0.57 (p < 0.05) for M_MAS and 0.46 - 0.45 (p = 0.06) for the MAL values. However, no correlation was seen for the difference between the two sessions.

CONCLUSIONS

The results confirm that the wrist acceleration can differentiate between the affected and non-affected arm, and that there is a positive correlation between accelerometry and clinical measures. Many of the patients did not change their M-MAS or MAL scores during the rehabilitation period, which may explain why no correlation was seen for the difference between measurements during the rehabilitation period. Further studies should include continuous accelerometry throughout the rehabilitation period to reduce the impact of day-to-day variability.

Artificial intelligence to enhance prehospital stroke diagnosis and triage: a perspective

As health systems organize to deliver the highest quality stroke care to their patients, there is increasing emphasis being placed on prehospital stroke recognition, accurate diagnosis, and efficient triage to improve outcomes after stroke. Emergency medical services (EMS) personnel currently rely heavily on dispatch accuracy, stroke screening tools, bypass protocols and prehospital notification to care for patients with suspected stroke, but novel tools including mobile stroke units and telemedicine-enabled ambulances are already changing the landscape of prehospital stroke care. Herein, the authors provide our perspective on the current state of prehospital stroke diagnosis and triage including several of these emerging trends. Then, we provide commentary to highlight potential artificial intelligence (AI) applications to improve stroke detection, improve accurate and timely dispatch, enhance EMS training and performance, and develop novel stroke diagnostic tools for prehospital use.

Machine learning to detect, stage and classify diseases and their symptoms based on inertial sensor data: a mapping review

This article presents a systematic review aimed at mapping the literature published in the last decade on the use of machine learning (ML) for clinical decision-making through wearable inertial sensors. The review aims to analyze the trends, perspectives, strengths, and limitations of current literature in integrating ML and inertial measurements for clinical applications. The review process involved defining four research questions and applying four relevance assessment indicators to filter the search results, providing insights into the pathologies studied, technologies and setups used, data processing schemes, ML techniques applied, and their clinical impact. When combined with ML techniques, inertial measurement units (IMUs) have primarily been utilized to detect and classify diseases and their associated motor symptoms. They have also been used to monitor changes in movement patterns associated with the presence, severity, and progression of pathology across a diverse range of clinical conditions. ML models trained with IMU data have shown potential in improving patient care by objectively classifying and predicting motor symptoms, often with a minimally encumbering setup. The findings contribute to understanding the current state of ML integration with wearable inertial sensors in clinical practice and identify future research directions. Despite the widespread adoption of these technologies and techniques in clinical applications, there is still a need to translate them into routine clinical practice. This underscores the importance of fostering a closer collaboration between technological experts and professionals in the medical field.

Prehospital technologies for early stroke detection - A review

The rate of neural circuitry loss in a typical large vessel occlusion well emphasizes that 'Time is Brain'. Every untreated minute in a large vessel ischaemic stroke results in loss of 1.9 million neurons and 13.8 billion synapses. As such, it is essential to optimize the flow-limiting steps in delivering the current standard of care. The current diagnostic model involves recognition of symptoms by patients, followed by access to Emergency Medical Services and subsequent physical examination and neuroimaging in the Emergency Department. With more than 50% of stroke patients using Emergency Medical Services as the first point of care contact, it can be deduced that the outcome of the 'stroke chain of survival' can be improved by addressing the bottleneck of prehospital stroke diagnosis. Here we present a review of the existing technologies.

Safety and usability of wearable accelerometers for stroke detection the STROKE ALARM PRO 1 study

OBJECTIVES

The introduction of time-dependent reperfusion therapies in acute ischemic stroke has increased the need for early identification. We explore the safety and feasibility of STROKE ALARM which detects sudden arm paresis, the most frequent symptom in stroke.

MATERIALS AND METHODS

Consecutive patients admitted with a stroke or TIA at Skåne University Hospital were screened according to inclusion and exclusion criteria, and included in the STROKE ALARM PRO 1 Study aiming to explore the feasibility of prolonged use (30 days) of the system in the community. STROKE ALARM consists of paired arm bracelets with accelerometers, coupled with a stroke test in a smartphone application. In case of an imbalance in arm movements, the user is prompted to perform an app-based stroke test. Failure to respond or to complete the stroke test correctly, triggers notification by SMS to predefined emergency contacts. Patients were followed up by telephone after completion.

RESULTS

Thirty patients were included and 28 completed follow-up. Median age was 68 years and 36.7% were female. No stroke events were recorded during follow-up. False indications occurred in all but one patient, and 22 (78.6%) experienced alarms to their emergency contacts. Despite a high level of false alarms, general user experience was rated in a positive or neutral manner by almost 90%. Very frequent alarms were probably due to mild arm paresis not detected in routine clinical assessment.

CONCLUSIONS

Use of STROKE ALARM for 30 days after stroke/TIA was well tolerated warranting further study for early automated detection of stroke recurrence.