Innovative Approaches to Sleep Monitoring in Stroke Patients
Stroke survivors frequently grapple with sleep disorders, a challenge that significantly affects their overall health and quality of life. Understanding the nuances of sleep stages such as deep sleep and rapid eye movement (REM) sleep is crucial, as each stage contributes uniquely to brain health. Continuous, high-resolution sleep assessments could revolutionize the early detection and management of sleep disorders during recovery. However, the traditional method of polysomnography (PSG), while being the gold standard, presents challenges in inpatient settings due to its cost, complexity, and the burden it places on patients.
Exploring Wearable Technology for Sleep Analysis
Wearable sensors coupled with advanced machine learning models present a promising, scalable alternative to PSG. Nonetheless, the physiological changes following a stroke can diminish the accuracy of these models, and there is a scarcity of post-stroke PSG data to refine them. To bridge these gaps, researchers have explored sleep stage classification models using diverse datasets from inpatient, chronic, and non-stroke populations.
Methods and Analysis
In the study, 14 sleep stage classifiers were developed using data from three groups: 8 inpatients undergoing stroke rehabilitation, 131 individuals with chronic stroke, and 145 control subjects without stroke. The research employed ordinal logistic regression models to categorize sleep into two, three, or four stages using heart rate and RR interval data derived from ECG or blood oxygen saturation (SpO2/SaO2).
Key Findings
The use of leave-one-subject-out models trained on control and/or chronic stroke ECG data enhanced the detection of steady-state sleep stages compared to models trained solely on steady-state data. Specifically, the Cohen’s Kappa values were 0.31, 0.24, and 0.17 for the 2-, 3-, and 4-stage classifiers, respectively. Incorporating SpO2/SaO2 data boosted performance in control and chronic stroke populations by approximately 20% per stage. However, it decreased performance by about 50% per stage in hospitalized patients, highlighting the distinct differences among these populations.
Conclusion
The findings indicate that training data from mixed populations can enhance sleep stage classification post-stroke. This research supports the potential of scalable, wearable-based sleep monitoring to facilitate personalized interventions during stroke recovery.
For more detailed information, you can access the full study Here.
“`
