Self-Aware Machine Learning for Chronic Pathology Monitoring on Wearable Devices

Remote health monitoring has attracted a lot of attention over the past decades to provide the opportunity for early detection of pathological health conditions. This early detection improves the quality of life for the patients and significantly reduces the burden on their family members. Moreover, this improvement reduces the socioeconomic consequences that are caused due to the disability of patients to work despite their health conditions. Wearable technologies offer a promising solution in pervasive health monitoring by relaxing the constraints concerning time and location. There are two major challenges in developing these technologies. The first challenge is real-time monitoring of the patients to fulfill early detection of life-threatening conditions using advanced machine learning techniques. As a solution for this issue, in this thesis, I considered a paradigm shift toward self-aware approaches and frameworks, which resulted in less energy consumption and longer battery lifetime, hence enabling ambulatory patient monitoring. I provided the possibility of switching between energy-efficient and high-performance modes in these systems. The second challenge is improving performance while guaranteeing the accuracy of the system. In this work, I have addressed this second challenge by combining multi-parametric bio-signals analysis and personalizing the detection for each patient. Throughout this thesis, I focused on epileptic seizure detection systems as a principal case study to analyze and evaluate the proposed techniques.

Self-Aware Wearable Systems in Epileptic Seizure Detection

Amir Aminifar, David Atienza Alonso, Farnaz Forooghifar

Today, wearable systems are facing fundamental barriers in terms of battery lifetime and quality of their results. The main challenge in wearable systems is to increase the battery lifetime, while maintaining the machine-learning performance of the system. A recently proposed concept for overcoming this challenge is self-awareness, which increases system's knowledge of itself and the surrounding environment. This is precisely what health monitoring wearable systems require to adapt to different situations. To demonstrate the impact of introducing self-awareness in wearable technologies, we consider the epileptic seizure detection problem, as a case study. Epilepsy affects around 1% of the world's population, which can dramatically degrade the quality of life and represents a major public health issue. As a result, detection of epileptic seizures has become more important over the past decades. In this paper, we aim to introduce a new generation of self-aware wearable systems to decrease energy consumption and improve their seizures detection capabilities by introducing the notion of self-awareness in such systems. These techniques include switching to low-power mode to reduce the energy consumption and machine-learning model enhancement to improve detection quality. We incorporated our proposed techniques in the machine learning module, which detects epileptic seizures by monitoring the cardiac and respiratory systems. We evaluated the performance of our approach based on an epilepsy database of more than 141 hours, provided by the Lausanne University Hospital (CHUV). Our self-aware wearable system achieves 36% reduction in computational complexity and 10.51% improvement in detection performance.

2018

Assessment of Foot Signature Using Wearable Sensors for Clinical Gait Analysis and Real-Time Activity Recognition

Benoît Mariani

Locomotion is one of the most important abilities of humans. Actually, gait locomotion provides mobility, and symbolizes freedom and independence. However, gait can be affected by several pathologies, due to aging, neurodegenerative disease, or trauma. The evaluation and treatment of mobility diseases thus requires clinical gait assessment, which is commonly done by using either qualitative analysis based on subjective observations and questionnaires, or expensive analysis established in complex motion laboratories settings. This thesis presents a new wearable system and algorithmic methods for gait assessment in natural conditions, addressing the limitations of existing methods. The proposed system provides quantitative assessment of gait performance through simple and precise outcome measures. The system includes wireless inertial sensors worn on the foot, that record data unobtrusively over long periods of time without interfering with subject's walking. Signal processing algorithms are presented for the automatic calibration and online virtual alignment of sensor signals, the detection of temporal parameters and gait phases, and the estimation of 3D foot kinematics during gait based on fusion methods and biomechanical assumptions. The resulting 3D foot trajectory during one gait cycle is defined as Foot Signature, by analogy with hand-written signature. Spatio-temporal parameters of interest in clinical assessment are derived from foot signature, including commonly parameters, such as stride velocity and gait cycle time, as well as original parameters describing inner-stance phases of gait, foot clearance, and turning. Algorithms based on expert and machine learning methods have been also adapted and implemented in real-time to provide input features to recognize locomotion activities including level walking, stairs, and ramp locomotion. Technical validation of the presented methods against gold standard systems was carried out using experimental protocols on subjects with normal and abnormal gait. Temporal aspects and quantitative estimation of foot-flat were evaluated against pressure insoles in subjects with ankle treatments during long-term gait. Furthermore, spatial parameters and foot clearance were compared in young and elderly persons to data obtained from an optical motion capture system during forward gait trials at various speeds. Finally, turning was evaluated in children with cerebral palsy and people with Parkinson's disease against optical motion capture data captured during timed up and go and figure-of-8 tests. Overall, the results demonstrated that the presently proposed system and methods were precise and accurate, and showed agreement with reference systems as well as with clinical evaluations of subjects' mobility disease using classical scores. Currently, no other methods based on wearable sensors have been validated with such precision to measure foot signature and subsequent parameters during unconstrained walking. Finally, we have used the proposed system in a large-scale clinical application involving more than 1800 subjects from age 7 to 77. This analysis provides reference data of common and original gait parameters, as well as their relationship with walking speed, and allows comparisons between different groups of subjects with normal and abnormal gait. Since the presented methods can be used with any foot-worn inertial sensors, or even combined with other systems, we believe our work to open the door to objective and quantitative routine gait evaluations in clinical settings for supporting diagnosis. Furthermore, the present studies have high potential for further research related to rehabilitation based on real-time devices, the investigation of new parameters' significance and their association with various mobility diseases, as well as for the evaluation of clinical interventions.

EPFL2012

Distributed Time Series Analytics

Tian Guo

In recent years time series data has become ubiquitous thanks to affordable sensors and advances in embedded technology. Large amount of time-series data are continuously produced in a wide spectrum of applications, such as sensor networks, medical monitoring and so on. Availability of such large scale time series data highlights the importance of of scalable data management, efï¬cient querying and analysis. Meanwhile, in the online setting time series carries invaluable information and knowledge about the real-time status of involved entities or monitored phenomena, which calls for online time series data mining for serving timely decision making or event detection. In this thesis we aim to address these important issues pertaining to scalable and distributed analytics techniques for massive time series data. Concretely, this thesis is centered around the following three topics: As the number of sensors that pervade our lives signiï¬cantly increases (e.g., environmental sensors, mobile phone sensors, IoT applications, etc.), the efï¬cient management of massive amount of time series from such sensors is becoming increasingly important. The inï¬nite nature of sensor data poses a serious challenge for query processing even in a cloud infrastructure. Traditional raw sensor data management systems based on relational databases lack scalability to accommodate large scale sensor data efï¬ciently. Thus, distributed key-value stores in the cloud are becoming a prime tool to manage sensor data. However, currently there are no techniques for indexing and/or query optimization of the model-view sensor time series data in the cloud. In Chapter 2, we propose an innovative index for modeled segments in key-value stores, namely KVI-index. KVI-index consists of two interval indices on the time and sensor value dimensions respectively, each of which has an in-memory search tree and a secondary list materialized in the key-value store. The dramatic increase in the availability of data streams fuels the development of many distributed real-time computation engines (e.g., Storm, Samza, Spark Streaming, S4 etc.). In Chapter 3, we focus on a fundamental time series mining task in such a new computation paradigm, namely continuously mining dynamic (lagged) correlations in time series via a distributed real-time computation engine. Correlations reveal the hidden and temporal interactions across time series and are widely used in scientiï¬c data analysis, data-driven event detection, ï¬nance markets and so on. We propose the P2H framework consisting of a parallelism-partitioning based data shufï¬ing and a hypercube structure based computation pruning method, so as to enhance both the communication and computation efï¬ciency for mining correlations in the distributed context. In numerous real-world applications large datasets collected from observations and measurements of physical entities are inevitably noisy and contain outliers. The outliers in such large and noisy datasets can dramatically degrade the performance of standard distributed machine learning approaches such as s regression trees. In Chapter 4 we present a novel distributed regression tree approach that utilizes robust regression statistics, statistics that are more robust to outliers, for handling large and noisy datasets. Then we present an adaptive gradient learning method for recurrent neural networks (RNN) to forecast streaming time series in the presence of both outliers and change points.