IntroductionCurrent PD assessment methods have inherent limitations. There is need for an objective method to assist clinical decisions and to facilitate evaluation of treatments. Accelerometers, and analysis using artificial neural networks (ANN), have shown potential as a method of motor symptom evaluation. This work describes the development of a novel PD disease state detection system informed by algorithms based on data collected in an unsupervised, home environment. We evaluated whether this approach can reproduce patient-completed symptom diaries and clinical assessment of disease state. Methods34 participants with PD wore bilateral wrist-worn accelerometers for 4 h in a research facility (phase 1) and for 7 days at home whilst completing symptom diaries (phase 2). An ANN to predict disease state was developed based on home-derived accelerometer data. Using a leave-one-out approach, ANN performance was evaluated against patient-completed symptom diaries and against clinician rating of disease state. ResultsIn the clinical setting, specificity for dyskinesia detection was extremely high (0.99); high specificity was also demonstrated for home-derived data (0.93), but with low sensitivity (0.38). In both settings, sensitivity for on/off detection was sub-optimal. ANN-derived values of the proportions of time in each disease state showed strong, significant correlations with patient-completed symptom diaries.
Mobile phone data are an interesting new data source for official statistics. However, multiple problems and uncertainties need to be solved before these data can inform, support or even become an integral part of statistical production processes. In this paper, we focus on arguably the most important problem hindering the application of mobile phone data in official statistics: detecting home locations. We argue that current efforts to detect home locations suffer from a blind deployment of criteria to define a place of residence and from limited validation possibilities. We support our argument by analysing the performance of five home detection algorithms (HDAs) that have been applied to a large, French, Call Detailed Record (CDR) dataset (~18 million users, 5 months). Our results show that criteria choice in HDAs influences the detection of home locations for up to about 40% of users, that HDAs perform poorly when compared with a validation dataset (the 35°-gap), and that their performance is sensitive to the time period and the duration of observation. Based on our findings and experiences, we offer several recommendations for official statistics. If adopted, our recommendations would help in ensuring a more reliable use of mobile phone data vis-à-vis official statistics.
Human activity recognition (HAR) in ubiquitous computing is beginning to adopt deep learning to substitute for well-established analysis techniques that rely on hand-crafted feature extraction and classification techniques. From these isolated applications of custom deep architectures it is, however, difficult to gain an overview of their suitability for problems ranging from the recognition of manipulative gestures to the segmentation and identification of physical activities like running or ascending stairs. In this paper we rigorously explore deep, convolutional, and recurrent approaches across three representative datasets that contain movement data captured with wearable sensors. We describe how to train recurrent approaches in this setting, introduce a novel regularisation approach, and illustrate how they outperform the state-of-the-art on a large benchmark dataset. Across thousands of recognition experiments with randomly sampled model configurations we investigate the suitability of each model for different tasks in HAR, explore the impact of hyperparameters using the fANOVA framework, and provide guidelines for the practitioner who wants to apply deep learning in their problem setting.
The majority of activity recognition systems in wearable computing rely on a set of statistical measures, such as means and moments, extracted from short frames of continuous sensor measurements to perform recognition. These features implicitly quantify the distribution of data observed in each frame. However, feature selection remains challenging and labour intensive, rendering a more generic method to quantify distributions in accelerometer data much desired. In this paper we present the ECDF representation, a novel approach to preserve characteristics of arbitrary distributions for feature extraction, which is particularly suitable for embedded applications. In extensive experiments on six publicly available datasets we demonstrate that it outperforms common approaches to feature extraction across a wide variety of tasks.
Our evaluations show that frequency features perform better than motion texture features, which in-turn perform better than symbol-/word-based features. Put succinctly, skill classification accuracy is positively correlated with motion granularity as demonstrated by our results on two challenging video datasets.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.