The goal of the 2016 PhysioNet/CinC Challenge is the development of an algorithm to classify normal/abnormal heart sounds. A total of 124 time-frequency features were extracted from the phonocardiogram (PCG) and input to a variant of the AdaBoost classifier. A second classifier using convolutional neural network (CNN) was trained using PCGs cardiac cycles decomposed into four frequency bands. The final decision rule to classify normal/abnormal heart sounds was based on an ensemble of classifiers combining the outputs of AdaBoost and the CNN. The algorithm was trained on a training dataset (normal= 2575, abnormal= 665) and evaluated on a blind test dataset. Our classifier ensemble approach obtained the highest score of the competition with a sensitivity, specificity, and overall score of 0.9424, 0.7781, and 0.8602, respectively. IntroductionHeart auscultation is the primary tool for screening and diagnosis in primary health care [1]. Availability of digital stethoscopes and mobile devices provides clinicians an opportunity to record and analyze heart sounds (PCG) for diagnostic purposes. The goal of the 2016 PhysioNet/CinC Challenge is the development of algorithms to classify normal/abnormal heart sound recordings [2]. We proposed an ensemble of a feature-based classifier and a deep learningbased classifier to boost the classification performance of heart sounds. Method and MaterialA block diagram of the proposed approach to classify normal/abnormal PCG is shown in Fig. 1. Challenge DatabaseThe challenge database provided PCG recordings of healthy subjects and pathological patients collected at either a clinical or non-clinical environment. Details about the challenge dataset can be found in [2]. For algorithm development, in-house training and test sets were generated by randomly taking 80% and 20% of the records from each database, while keeping the same prevalence of abnormal classes. In-house training set was used for training and cross-validation of different models, and in-house test set was used for evaluation of the classification performance independently from the blind test dataset. Pre-processingEach PCG was resampled to 1000 Hz, band-pass filtered between 25 Hz and 400 Hz, and then pre-processed to remove any spikes in the PCG [3]. Furthermore, preprocessed PCGs were segmented into four heart sound states using a segmentation method proposed by Springer et al. [4]. Each PCG is comprised of more than one cardiac cycle (beat), and each beat is comprised of four heart sound states (i.e. S1, systole, S2, and diastole). Feature-based ApproachIn this approach, a variant of AdaBoost classifier [5] was trained for classification of normal/abnormal PCGs using time and frequency-domain features. Time-domain FeaturesMean and standard deviation (SD) of the following parameters were used as time-domain features (36 features): 1. PCG intervals: RR intervals, S1 intervals, S2 intervals, systolic intervals, diastolic intervals, ratio of systolic interval to RR interval of each heart beat, ratio of diastolic...

Background: Individuals with diabetic peripheral neuropathy (DPN) have deficits in sensory and motor skills leading to inadequate proprioceptive feedback, impaired postural balance and higher fall risk. Objective: This study investigated the effect of sensor-based interactive balance training on postural stability and daily physical activity in older adults with diabetes. Methods: Thirty-nine older adults with DPN were enrolled (age 63.7 ± 8.2 years, BMI 30.6 ± 6, 54% females) and randomized to either an intervention (IG) or a control (CG) group. The IG received sensor-based interactive exercise training tailored for people with diabetes (twice a week for 4 weeks). The exercises focused on shifting weight and crossing virtual obstacles. Body-worn sensors were implemented to acquire kinematic data and provide real-time joint visual feedback during the training. Outcome measurements included changes in center of mass (CoM) sway, ankle and hip joint sway measured during a balance test while the eyes were open and closed at baseline and after the intervention. Daily physical activities were also measured during a 48-hour period at baseline and at follow-up. Analysis of covariance was performed for the post-training outcome comparison. Results: Compared with the CG, the patients in the IG showed a significantly reduced CoM sway (58.31%; p = 0.009), ankle sway (62.7%; p = 0.008) and hip joint sway (72.4%; p = 0.017) during the balance test with open eyes. The ankle sway was also significantly reduced in the IG group (58.8%; p = 0.037) during measurements while the eyes were closed. The number of steps walked showed a substantial but nonsignificant increase (+27.68%; p = 0.064) in the IG following training. Conclusion: The results of this randomized controlled trial demonstrate that people with DPN can significantly improve their postural balance with diabetes-specific, tailored, sensor-based exercise training. The results promote the use of wearable technology in exercise training; however, future studies comparing this technology with commercially available systems are required to evaluate the benefit of interactive visual joint movement feedback.

Advances in wearable technology allow for the objective assessment of motor performance in both in-home and in-clinic environments and were used to explore motor impairments in Parkinson’s disease (PD). The aims of this study were to: 1) assess differences between in-clinic and in-home gait speed, and sit-to-stand and stand-to-sit duration in PD patients (in comparison with healthy controls); and 2) determine the objective physical activity measures, including gait, postural balance, instrumented Timed-up-and-go (iTUG), and in-home spontaneous physical activity (SPA), with the highest correlation with subjective/semi-objective measures, including health survey, fall history (fallers vs. non-fallers), fear of falling, pain, Unified Parkinson's Disease Rating Scale, and PD stage (Hoehn and Yahr). Objective assessments of motor performance were made by measuring physical activities in the same sample of PD patients (n = 15, Age: 71.2±6.3 years) and age-matched healthy controls (n = 35, Age: 71.9±3.8 years). The association between in-clinic and in-home parameters, and between objective parameters and subjective/semi-objective evaluations in the PD group was assessed using linear regression-analysis of variance models and reported as Pearson correlations (R). Both in-home SPA and in-clinic assessments demonstrated strong discriminatory power in detecting impaired motor function in PD. However, mean effect size (0.94±0.37) for in-home measures was smaller compared to in-clinic assessments (1.30±0.34) for parameters that were significantly different between PD and healthy groups. No significant correlation was observed between identical in-clinic and in-home parameters in the PD group (R = 0.10–0.25; p>0.40), while the healthy showed stronger correlation in gait speed, sit-to-stand duration, and stand-to-sit duration (R = 0.36–0.56; p<0.03). This suggests a better correlation between supervised and unsupervised motor function assessments in healthy controls compared to PD group. In the PD group, parameters related to velocity and range-of-motion of lower extremity within gait assessment (R = 0.58–0.84), and turning duration and velocity within iTUG test (R = 0.62–0.77) demonstrated strong correlations with PD stage (p<0.01).

Background: Impairment of physical function is a major indicator of frailty. Functional performance tests have been shown to be useful for identification of frailty in older adults. However, these tests are often not translatable into unsupervised and remote monitoring of frailty status at home and/or community settings. Objective: In this study, we explored daily postural transition quantified using a chest-worn wearable technology to identify frailty in community-dwelling older adults. Methods: Spontaneous daily physical activity was monitored over 24 h in 120 community-dwelling elderly (age: 78 ± 8 years) using an unobtrusive wearable sensor (PAMSys™, BioSensics LLC, Watertown, MA, USA). Participants were classified as non-frail and pre-frail/frail using Fried's criteria. A validated software package was used to identify body postures and postural transition between each independent postural activity such as sit-to-stand, stand-to-sit, stand-to-walk, and walk-to-stand. The transition from walking to sitting was further classified as quick sitting and cautious sitting based on presence/absence of a standing posture pause between sitting and walking. A general linear model univariate test was used for between-group comparison. Pearson's correlation was used to determine the association between sensor-derived parameters and age. Logistic regression model was used to identify independent predictors of frailty. Results: According to Fried's criteria, 63% of participants were pre-frail/frail. The total number of postural transitions, stand-to-walk, and walk-to-stand were, respectively, 25.2, 30.2, and 30.6% lower in the pre-frail/frail group when compared to the non-frail group (p < 0.05, Cohen's d = 0.73-0.79). Furthermore, the ratio of cautious sitting was significantly higher by 6.2% in pre-frail/frail compared to non-frail (p = 0.025, Cohen's d = 0.22). Total number of postural transitions and the ratio of cautious sitting also showed significant negative and positive correlations with age, respectively (r = -0.51 and 0.29, p < 0.05). After applying a logistic regression model, among tested parameters, walk-to-stand (odds ratio [OR] = 0.997 p = 0.013), quick sitting (OR = 1.036, p = 0.05), and age (OR = 1.073, p = 0.016) were recognized as independent variables to identify frailty status. Conclusions: This study demonstrated that daily number of specific postural transitions such as walk-to-stand and quick sitting could be used for monitoring frailty status by unsupervised monitoring of daily physical activity. Further study is warranted to explore whether tracking the daily number of specific postural transitions is also sensitive to track change in the status of frailty over time.

Background: Frailty is a geriatric syndrome that leads to impairment in interrelated physiological systems and progressive homeostatic dysregulation in physiological systems. Objective: The focus of the present systematic review was to study the association between the activity of the cardiac autonomic nervous system (ANS) and frailty. Methods: A systematic literature search was conducted in multiple databases: PubMed/MEDLINE, Embase, Cochrane Library, Web of Science, CINAHL, and ClinicalTrials.gov; the last search was performed in March 2015. Inclusion criteria were: (1) that the studied population was classified for frailty according to a standard definition, such as Fried's criteria; (2) that the study had a nonfrail control group, and (3) that heart rate (HR) and/or heart rate variability (HRV) were parameters of interest in the study. Results: Of the 1,544 articles screened, 54 were selected for full-text review and 6 studies met the inclusion criteria. Assessment of HRV using different standard time domain, frequency domain, and nonlinear domain approaches confirmed the presence of an impaired cardiac ANS function in frail compared to nonfrail participants. Furthermore, HR changes while performing a clinical test (e.g., the seated step test or the lying-to-standing orthostatic test) were decreased in the frail group compared to the nonfrail group. Conclusions: The current systematic review provides evidence that the cardiac ANS is impaired in frail compared to nonfrail older adults, as indicated by a reduction in the complexity of HR dynamics, reduced HRV, and reduced HR changes in response to daily activities. Four out of 6 included articles recruited only female participants, and in the other 2 articles the effect of gender on impairment of cardiac ANS was insufficiently investigated. Therefore, further studies are required to study the association between cardiac ANS impairments and frailty in males. Furthermore, HRV was studied only during static postures such as sitting, or without considering the level of activity as a potential confounder. Accordingly, simultaneous measurement of both physiological (i.e., HRV) and kinematic (e.g., using wearable sensor technology) information may provide a better understanding of cardiac ANS impairments with frailty while controlling for activity.

No abstract

No abstract

No abstract