Visualization-Driven Time-Series Extraction from Wearable Systems Can Facilitate Differentiation of Passive ADL Characteristics among Stroke and Healthy Older Adults

A comprehensive review of machine learning algorithms and their application in geriatric medicine: present and future

The increasing access to health data worldwide is driving a resurgence in machine learning research, including data-hungry deep learning algorithms. More computationally efficient algorithms now offer unique opportunities to enhance diagnosis, risk stratification, and individualised approaches to patient management. Such opportunities are particularly relevant for the management of older patients, a group that is characterised by complex multimorbidity patterns and significant interindividual variability in homeostatic capacity, organ function, and response to treatment. Clinical tools that utilise machine learning algorithms to determine the optimal choice of treatment are slowly gaining the necessary approval from governing bodies and being implemented into healthcare, with significant implications for virtually all medical disciplines during the next phase of digital medicine. Beyond obtaining regulatory approval, a crucial element in implementing these tools is the trust and support of the people that use them. In this context, an increased understanding by clinicians of artificial intelligence and machine learning algorithms provides an appreciation of the possible benefits, risks, and uncertainties, and improves the chances for successful adoption. This review provides a broad taxonomy of machine learning algorithms, followed by a more detailed description of each algorithm class, their purpose and capabilities, and examples of their applications, particularly in geriatric medicine. Additional focus is given on the clinical implications and challenges involved in relying on devices with reduced interpretability and the progress made in counteracting the latter via the development of explainable machine learning.

Day-to-day regularity and diurnal switching of physical activity reduce depression-related behaviors: a time-series analysis of wearable device data

BACKGROUND

Wearable devices have been widely used in research to understand the relationship between habitual physical activity and mental health in the real world. However, little attention has been paid to the temporal variability in continuous physical activity patterns measured by these devices. Therefore, we analyzed time-series patterns of physical activity intensity measured by a wearable device and investigated the relationship between its model parameters and depression-related behaviors.

METHODS

Sixty-six individuals used the wearable device for one week and then answered a questionnaire on depression-related behaviors. A seasonal autoregressive integral moving average (SARIMA) model was fitted to the individual-level device data and the best individual model parameters were estimated via a grid search.

RESULTS

Out of 64 hyper-parameter combinations, 21 models were selected as optimal, and the models with a larger number of affiliations were found to have no seasonal autoregressive parameter. Conversely, about half of the optimal models indicated that physical activity on any given day fluctuated due to the previous day's activity. In addition, both irregular rhythms in day-to-day activity and low-level of diurnal variability could lead to avoidant behavior patterns.

CONCLUSION

Automatic and objective physical activity data from wearable devices showed that diurnal switching of physical activity, as well as day-to-day regularity rhythms, reduced depression-related behaviors. These time-series parameters may be useful for detecting behavioral issues that lie outside individuals' subjective awareness.