Unobtrusive Accelerometer-based Longitudinal ADL Monitoring among Different Populations

Abstract

Approximately 15 million people worldwide suffer from stroke each year and more than 10 million people are living with Parkinson’s disease[1]. As the third leading cause of death in the United States, stroke is a medical emergency that leads to the death of more than 140,000 individuals each year with a cost of $5,392 each year for medication and $11,689 yearly for rehabilitation[2]. Using accelerometers, one can longitudinally monitor these individuals, thus can help maintain individuals’ quality of life through Activities of daily living (ADL) targeted interventions. Although wearable technology can monitor and collect data needed to make the predictions about one’s health status, new methods for interpreting longitudinal datasets are lacking. Therefore, long duration data is unable to be translated into usable results that state a person’s health status. This study utilizes accelerometer data to investigate ADL and its variation throughout the day are correlated with the health status of individuals[3]. ADL is an important factor in these populations since it is a result of health impact due to illnesses on the individuals[4]. The amount of movement performed, number of transitions occurring during sleep and when awake, variations in transitions of ADL, how much time in between each movement transition, and how often these movements transitions occur will all be considered aspects of daily living activities that could be measured by the accelerometer sensors. Transitions are proven as important indicator in assessment of the health status among young and older[5] and obese adults[6].

Fig. 1a. Participant wearing sensor at low back, 1b. 3-day longitudinal accelerometer data

Methods: In this study, participants wore a wearable accelerometer device on their lower back and performed unhindered movements in their natural settings. The wearable sensor Dynaport (Motion Monitor+, McRoberts BV, The Hague, The Netherlands) was worn like a belt on the participants lower back, specifically above the posterior iliac spine at the L5/ S1 vertebrae. The sampling rate was set to 100 Hz and sensor recorded and stored data on a built-in SD chip. Prior to participation, written informed consent was given by all the participants which was approved by Chapman University IRB. The 24-hours of day were divided into 4 phases (each of 6 hours). Multivariate Analysis of Variance (MANOVA) was performed with subjects as random effects. The independent effects being groups (Healthy Young, healthy Old, PD and Stroke) and Phases (Phase 1, Phase 2, Phase 3 and Phase 4).

Results:

Fig. 2. The number of transitions in sleep phase 1 were significantly higher (p<0.01) than sleep phase 4 among all 4 groups (Healthy, young, old , PD and Stroke).

Fig. 3. Healthy young and older adults had similar trends, however PD group had lower transition duration compared to other groups. Stroke individuals produced significantly higher transition durations during phase 4 of sleep(p=0.017).

Discussion: We found that stroke patients had a lower activity magnitude than healthy old, healthy young, and Parkinson's patients. Individuals with stroke lose their motor function and are not able to walk and perform other daily activities [7]. Without proper brain function, it is difficult for them to move as much as normal healthy adults. Individuals with Parkinson's disease suffer from a disorder in the central nervous system that leads to stiff movement, tremor, difficulty with balance and coordination and as the disease progresses, they have a more difficult time in walking and speaking[8]. Healthy older individuals and individuals with Parkinson’s disease were found to have similar activity magnitudes due to old age as Parkinson’s disease is often diagnosed in people over 60 years old [9]. It was found that healthy young and older adults had similar trends when sleeping. However, Parkinson's disease individuals had a lower transition duration compared to the other groups, this may be attributed to restless leg syndrome and REM sleep behavior disorder[10]. Stroke individuals produced significantly higher transition durations (longer duration) during phase 4 of sleep, which could be caused by sleep disordered breathing[11] often diagnosed in stroke. Conclusion:  This study revealed important information on ADL and sleep behavior in the four different populations. The new signal processing methods utilizing sensor data is helpful for assessing ADL characteristics unobtrusively and continuously in natural settings. These novel signal processing algorithms will aid in improving health care by understanding of how the progression of diseases or the recovery from a disease will influence the movement of individuals.