scispace - formally typeset
Search or ask a question
Proceedings ArticleDOI

Recent machine learning advancements in sensor-based mobility analysis: Deep learning for Parkinson's disease assessment

TL;DR: This paper compared standard machine learning pipelines with deep learning based on convolutional neural networks and showed that deep learning outperformed other state-of-the-art machine learning algorithms in terms of classification rate.
Abstract: The development of wearable sensors has opened the door for long-term assessment of movement disorders. However, there is still a need for developing methods suitable to monitor motor symptoms in and outside the clinic. The purpose of this paper was to investigate deep learning as a method for this monitoring. Deep learning recently broke records in speech and image classification, but it has not been fully investigated as a potential approach to analyze wearable sensor data. We collected data from ten patients with idiopathic Parkinson's disease using inertial measurement units. Several motor tasks were expert-labeled and used for classification. We specifically focused on the detection of bradykinesia. For this, we compared standard machine learning pipelines with deep learning based on convolutional neural networks. Our results showed that deep learning outperformed other state-of-the-art machine learning algorithms by at least 4.6 % in terms of classification rate. We contribute a discussion of the advantages and disadvantages of deep learning for sensor-based movement assessment and conclude that deep learning is a promising method for this field.
Citations
More filters
Journal ArticleDOI
TL;DR: In this paper, the authors present a conciliatory explanation for the present publication, in which, it is acknowledged, that mere conjecture takes the place of experiment; and, that analogy is the substitute for anatomical examination, the only sure foundation for pathological knowledge.
Abstract: PREFACE The advantages which have been derived from the caution with which hypothetical statements are admitted, are in no instance more obvious than in those sciences which more particularly belong to the healing art. It therefore is necessary, that some conciliatory explanation should be offered for the present publication: in which, it is acknowledged, that mere conjecture takes the place of experiment; and, that analogy is the substitute for anatomical examination, the only sure foundation for pathological knowledge. When, however, the nature of the subject, and the circumstances under which it has been here taken up, are considered, it is hoped that the offering of the following pages to the attention of the medical public, will not be severely censured. The disease, respecting which the present inquiry is made, is of a nature highly afflictive. Notwithstanding which, it has not yet obtained a place in the classification of nosologists; some have regarded its characteristic symptoms as distinct and different diseases, and others have given its name to diseases differing essentially from it; whilst the unhappy sufferer has considered it as an evil, from the domination of which he had no prospect of escape. The disease is of long duration: to connect, therefore, the symptoms which occur in its later stages with those which mark its commencement, requires a continuance of observation of the same case, or at least a correct history of its symptoms, even for several years. Of both these advantages the writer has had the opportunities of availing himself, and has hence been led particularly to observe several other cases in which the disease existed in different stages of its progress. By these repeated observations, he hoped that he had been led to a probable conjecture as to the nature of the malady, and that analogy had suggested such means as might be productive of relief, and perhaps even of cure, if employed before the disease had been too long established. He therefore considered it to be a duty to submit his opinions to the examination of others, even in their present state of immaturity and imperfection. To delay their publication did not, indeed, appear to be warrantable. The disease had escaped particular notice; and the task of ascertaining its nature and cause by anatomical investigation, did not seem likely to be taken up by those who, from their abilities and opportunities, were most likely to accomplish it. That these friends to humanity and medical science, who have already unveiled to us many of the morbid processes by which health and life is abridged, might be excited to extend their researches to this malady, was much desired; and it was hoped, that this might be procured by the publication of these remarks. Should the necessary information be thus obtained, the writer will repine at no censurewhich the precipitate publication of mere conjectural suggestions may incur: but shall think himself fully rewarded by having excited the attention of those, who may point out the most appropriate means of relieving a tedious and most distressing malady.

869 citations

Journal ArticleDOI
TL;DR: The focus of this review is to provide in-depth summaries of deep learning methods for mobile and wearable sensor-based human activity recognition, and categorise the studies into generative, discriminative and hybrid methods.
Abstract: Human activity recognition systems are developed as part of a framework to enable continuous monitoring of human behaviours in the area of ambient assisted living, sports injury detection, elderly care, rehabilitation, and entertainment and surveillance in smart home environments. The extraction of relevant features is the most challenging part of the mobile and wearable sensor-based human activity recognition pipeline. Feature extraction influences the algorithm performance and reduces computation time and complexity. However, current human activity recognition relies on handcrafted features that are incapable of handling complex activities especially with the current influx of multimodal and high dimensional sensor data. With the emergence of deep learning and increased computation powers, deep learning and artificial intelligence methods are being adopted for automatic feature learning in diverse areas like health, image classification, and recently, for feature extraction and classification of simple and complex human activity recognition in mobile and wearable sensors. Furthermore, the fusion of mobile or wearable sensors and deep learning methods for feature learning provide diversity, offers higher generalisation, and tackles challenging issues in human activity recognition. The focus of this review is to provide in-depth summaries of deep learning methods for mobile and wearable sensor-based human activity recognition. The review presents the methods, uniqueness, advantages and their limitations. We not only categorise the studies into generative, discriminative and hybrid methods but also highlight their important advantages. Furthermore, the review presents classification and evaluation procedures and discusses publicly available datasets for mobile sensor human activity recognition. Finally, we outline and explain some challenges to open research problems that require further research and improvements.

601 citations

Proceedings ArticleDOI
03 Nov 2017
TL;DR: The proposed methods and CNNs are applied to the classification of the motor state of Parkinson’s Disease patients, which is challenging due to small dataset size, noisy labels, and large intra-class variability.
Abstract: While convolutional neural networks (CNNs) have been successfully applied to many challenging classification applications, they typically require large datasets for training. When the availability of labeled data is limited, data augmentation is a critical preprocessing step for CNNs. However, data augmentation for wearable sensor data has not been deeply investigated yet. In this paper, various data augmentation methods for wearable sensor data are proposed. The proposed methods and CNNs are applied to the classification of the motor state of Parkinson’s Disease patients, which is challenging due to small dataset size, noisy labels, and large intra-class variability. Appropriate augmentation improves the classification performance from 77.54% to 86.88%.

348 citations


Cites methods from "Recent machine learning advancement..."

  • ...Also, [7] applies a CNN to the classi€cation of bradykinesia present and absent states based on the wearable sensor data collected during several motor tasks....

    [...]

Journal ArticleDOI
TL;DR: This review focuses on wearable devices for PD applications and identifies five main fields: early diagnosis, tremor, body motion analysis, motor fluctuations (ON–OFF phases), and home and long-term monitoring.
Abstract: Background: Parkinson’s disease is a common and disabling pathology that is characterized by both motor and non-motor symptoms and affects millions of people worldwide. The disease significantly affects quality of life of those affected. Many works in literature discuss the effects of the disease. The most promising trends involve sensor devices, which are low cost, low power, unobtrusive, and accurate in the measurements, for monitoring and managing the pathology. Objectives: This review focuses on wearable devices for Parkinson’s disease applications and identifies five main fields: early diagnosis, tremor, body motion analysis, motor fluctuations (ON-OFF phases), and home and long-term monitoring. The concept is to obtain an overview of the pathology at each stage of development, from the beginning of the disease to consider early symptoms, during disease progression with analysis of the most common disorders, and including management of the most complicated situations (i.e. motor fluctuations and long-term remote monitoring). Data sources: The research was conducted within three databases: IEEE Xplore®, Science Direct®, and PubMed Central®, between January 2006 and December 2016. Study eligibility criteria: Since 1429 articles were found, accurate definition of the exclusion criteria and selection strategy allowed identification of the most relevant papers. Results: Finally, 136 papers were fully evaluated and included in this review, allowing a wide overview of wearable devices for the management of Parkinson's disease.

250 citations


Cites background or methods from "Recent machine learning advancement..."

  • ...…tapping) for upper limb motion analysis (Fukawa et al., 2007; Okuno et al., 2007; Patel et al., 2009; Yokoe et al., 2009; Hoffman andMcNames, 2011; Cavallo et al., 2013; Robles-García et al., 2013; Jia et al., 2014; Delrobaei et al., 2016; Djurić-Jovičić et al., 2016; Eskofier et al., 2016)....

    [...]

  • ...Alternatively, a recent work proposed the use of deep learning as a promising method to analyse wearable sensor data in place of machine learning approaches (Eskofier et al., 2016)....

    [...]

  • ...Also, Eskofier et al. (2016) performed an assessment of bradykinesia, but they introduced the use of deep learning instead of machine learning techniques as a promising method to analyse wearable sensor data....

    [...]

Proceedings ArticleDOI
TL;DR: In this article, various data augmentation methods for wearable sensor data are proposed and applied to the classification of the motor state of Parkinson's disease patients, which is challenging due to small dataset size, noisy labels, and large intra-class variability Appropriate augmentation improves the classification performance from 7754% to 8688%
Abstract: While convolutional neural networks (CNNs) have been successfully applied to many challenging classification applications, they typically require large datasets for training When the availability of labeled data is limited, data augmentation is a critical preprocessing step for CNNs However, data augmentation for wearable sensor data has not been deeply investigated yet In this paper, various data augmentation methods for wearable sensor data are proposed The proposed methods and CNNs are applied to the classification of the motor state of Parkinson's Disease patients, which is challenging due to small dataset size, noisy labels, and large intra-class variability Appropriate augmentation improves the classification performance from 7754\% to 8688\%

201 citations

References
More filters
Proceedings Article
01 Jan 2015
TL;DR: This work introduces Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments, and provides a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework.
Abstract: We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and/or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and/or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm.

111,197 citations


"Recent machine learning advancement..." refers methods in this paper

  • ...For the minimization, we employed Adam [25], a gradient-based optimizer for stochastic objective functions. an optimization algorithm for stochastic loss functions....

    [...]

  • ...For the minimization, we employed Adam [25], a gradient-based optimizer for stochastic objective functions....

    [...]

  • ...The Adam algorithm was used in the configuration recommended by Kingma & Ba (learning rate 10-3, β1 = 0.9, β2 = 0.999, ε = 10-8) [25]....

    [...]

Journal ArticleDOI
28 May 2015-Nature
TL;DR: Deep learning is making major advances in solving problems that have resisted the best attempts of the artificial intelligence community for many years, and will have many more successes in the near future because it requires very little engineering by hand and can easily take advantage of increases in the amount of available computation and data.
Abstract: Deep learning allows computational models that are composed of multiple processing layers to learn representations of data with multiple levels of abstraction. These methods have dramatically improved the state-of-the-art in speech recognition, visual object recognition, object detection and many other domains such as drug discovery and genomics. Deep learning discovers intricate structure in large data sets by using the backpropagation algorithm to indicate how a machine should change its internal parameters that are used to compute the representation in each layer from the representation in the previous layer. Deep convolutional nets have brought about breakthroughs in processing images, video, speech and audio, whereas recurrent nets have shone light on sequential data such as text and speech.

46,982 citations

Journal Article
TL;DR: It is shown that dropout improves the performance of neural networks on supervised learning tasks in vision, speech recognition, document classification and computational biology, obtaining state-of-the-art results on many benchmark data sets.
Abstract: Deep neural nets with a large number of parameters are very powerful machine learning systems. However, overfitting is a serious problem in such networks. Large networks are also slow to use, making it difficult to deal with overfitting by combining the predictions of many different large neural nets at test time. Dropout is a technique for addressing this problem. The key idea is to randomly drop units (along with their connections) from the neural network during training. This prevents units from co-adapting too much. During training, dropout samples from an exponential number of different "thinned" networks. At test time, it is easy to approximate the effect of averaging the predictions of all these thinned networks by simply using a single unthinned network that has smaller weights. This significantly reduces overfitting and gives major improvements over other regularization methods. We show that dropout improves the performance of neural networks on supervised learning tasks in vision, speech recognition, document classification and computational biology, obtaining state-of-the-art results on many benchmark data sets.

33,597 citations

Posted Content
TL;DR: In this article, the adaptive estimates of lower-order moments are used for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimate of lowerorder moments.
Abstract: We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and/or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and/or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm.

23,486 citations

Journal ArticleDOI
TL;DR: This paper provides an introduction to the WEKA workbench, reviews the history of the project, and, in light of the recent 3.6 stable release, briefly discusses what has been added since the last stable version (Weka 3.4) released in 2003.
Abstract: More than twelve years have elapsed since the first public release of WEKA. In that time, the software has been rewritten entirely from scratch, evolved substantially and now accompanies a text on data mining [35]. These days, WEKA enjoys widespread acceptance in both academia and business, has an active community, and has been downloaded more than 1.4 million times since being placed on Source-Forge in April 2000. This paper provides an introduction to the WEKA workbench, reviews the history of the project, and, in light of the recent 3.6 stable release, briefly discusses what has been added since the last stable version (Weka 3.4) released in 2003.

19,603 citations


"Recent machine learning advancement..." refers background in this paper

  • ...6), which is a collection of ML algorithms for data mining tasks [20]....

    [...]

Related Papers (5)
Christopher G. Goetz, Barbara C. Tilley, Stephanie R. Shaftman, Glenn T. Stebbins, Stanley Fahn, Pablo Martinez-Martin, Werner Poewe, Cristina Sampaio, Matthew B. Stern, Richard Dodel, Bruno Dubois, Robert G. Holloway, Joseph Jankovic, Jaime Kulisevsky, Anthony E. Lang, Andrew J. Lees, Sue Leurgans, Peter A. LeWitt, David L. Nyenhuis, C. Warren Olanow, Olivier Rascol, Anette Schrag, Jeanne A. Teresi, Jacobus J. van Hilten, Nancy R. LaPelle, Pinky Agarwal, Saima Athar, Yvette Bordelan, Helen Bronte-Stewart, Richard Camicioli, Kelvin L. Chou, Wendy Cole, Arif Dalvi, Holly Delgado, Alan Diamond, Jeremy P.R. Dick, John E. Duda, Rodger J. Elble, Carol Evans, V. G. H. Evidente, Hubert H. Fernandez, Susan H. Fox, Joseph H. Friedman, Robin D. Fross, David A. Gallagher, Deborah A. Hall, Neal Hermanowicz, Vanessa K. Hinson, Stacy Horn, Howard I. Hurtig, Un Jung Kang, Galit Kleiner-Fisman, Olga Klepitskaya, Katie Kompoliti, Eugene C. Lai, Maureen L. Leehey, Iracema Leroi, Kelly E. Lyons, Terry McClain, Steven W. Metzer, Janis M. Miyasaki, John C. Morgan, Martha Nance, Joanne Nemeth, Rajesh Pahwa, Sotirios A. Parashos, Jay S. Schneider, Kapil D. Sethi, Lisa M. Shulman, Andrew Siderowf, Monty Silverdale, Tanya Simuni, Mark Stacy, Robert Malcolm Stewart, Kelly L. Sullivan, David M. Swope, Pettaruse M. Wadia, Richard Walker, Ruth H. Walker, William J. Weiner, Jill Wiener, Jayne R. Wilkinson, Joanna M. Wojcieszek, Summer C. Wolfrath, Frederick Wooten, Allen Wu, Theresa A. Zesiewicz, Richard M. Zweig 
Trending Questions (1)
How Companies Are Using machine learning and deep learning to improve human experience?

Our results showed that deep learning outperformed other state-of-the-art machine learning algorithms by at least 4.6 % in terms of classification rate.