Currently, he is working as a Ph.D. in the Laboratory of Embedded Systems in EPFL with Prof. David Atienza. His research includes design and optimization of heterogeneous multi-cores systems for processing multiple bio-signals in real time.
Entries sorted reverse-chronologically
Event-Triggered Sensing for High-Quality and Low-Power Cardiovascular Monitoring Systems
IEEE Design & Test. 2020-01-01.
DOI : 10.1109/MDAT.2019.2951126.
In the context of wearable medical systems, resources are scarce while performance requirements are high. Traditional sampling strategies create large amounts of data, which hinders the device's battery lifetime. However, energy savings are possible when relying on an event-triggered strategy, following the brain example. In this paper, we explore the use of non-Nyquist sampling for cardiovascular monitoring systems, with an in-depth analysis of the performance of a knowledge-based adaptive sampling strategy. By reducing the average sampling rate from 360 Hz down to 13.6 Hz, we can increase the battery lifetime by 4× with a marginal impact on the accuracy of heart-rate analysis.
Low Power Sensing and Processing in Wearable Biomedical Devices for Personalized Health Monitoring
Lausanne, EPFL, 2019.
DOI : 10.5075/epfl-thesis-9721.
Whether it is for personal use or for medical application, wearable sensors are becoming more and more widespread. This is the industry answer to two parallel trends. First, the public show a wish to collect data about their own lifestyle. This rather new effect appeared with the rise of smartphones, smartwatches and other heart-rate belts for athletes, with the promise of understanding and improving their health and performance. From a medical point-of-view, these commercial devices are not usable because their results come from an unproven algorithm, far from clinical trials. However, the health-care professionals see a strong benefit of having trustworthy wearable devices, which can be used by their patients for extended periods of time. Having in situ medical-grade data collection is able to give insights about the patient's health status and its development. There are however limitations to what can be performed. Ideally, such a device should record as many data as possible, without requiring any set-up nor maintenance. Because of technical limitations, there are choices to be made. For example, which signals are captured, or what should be the minimal battery life? Any additional burden put on the patient hinders using the device, especially in the case of a non immediate life-threatening situation such as a daily screening. State-of-the-art biomedical wearable devices deploy multiple strategies to match the specification's stringent requirements. First it is engetically expensive to sense signals. The more data is collected, the smaller the battery lifetime. Therefore, extensive research is pursued beforehand to minimize the number of signals required. If a pathology can be reliably identified with fewer data, it is a meaningful saving. Additional energy savings are possible by minimizing the amount of data to transmit. Pushing this idea to the extreme, if the diagnostics algorithm can be efficiently run on the device itself, sending only the results is an excellent approach adopted by a number of sensor nodes. In this thesis, I first propose event-driven approaches for sensing bio-signals, designed to take into account the signal's behavior. Changing the sampling strategy is based on the fact that oversampling is frequent, because this approach provides strong guarantees about the data quality. However, it is not taking into account the actual signal's temporal characteristics, where a constant value is digitized with the same rate as a high-frequency pattern. Taking into account the signal's evolution, several kinds of events are envisioned for triggering the measure, from a simple level-crossing strategy to a more refined knowledge-based one. This event-driven sampling paradigm minimizes the amount of data to process, leading to an increased battery lifetime. Secondly, I consider a widespread but under-diagnosed disease: obstructive sleep apnea. This disease is connected to increased risks of cardiovascular diseases, motivating the need for diagnostic and treatment. The target device I consider is an existing wearable biomedical node. Following the optimization process described previously, I rely on a single biological signal processed directly on the sensor. Using machine learning, a limited number of features are extracted and filtered before being used for the obstructive sleep apnea assessment. Afterwards, the device only requires to send the results to a base station, being a non-instrusive OSA screening system.
Online Obstructive Sleep Apnea Detection on Medical Wearable Sensors
IEEE Transactions on Biomedical Circuits and Systems. 2018-08.
DOI : 10.1109/TBCAS.2018.2824659.
Obstructive Sleep Apnea (OSA) is one of the main under-diagnosed sleep disorder. It is an aggravating factor for several serious cardiovascular diseases, including stroke. There is, however, a lack of medical devices for long-term ambulatory monitoring of OSA since current systems are rather bulky, expensive, intrusive, and cannot be used for long-term monitoring in ambulatory settings. In this paper, we propose a wearable, accurate, and energy efficient system for monitoring obstructive sleep apnea on a long-term basis. As an embedded system for Internet of Things (IoT), it reduces the gap between home health-care and professional supervision. Our approach is based on monitoring the patient using a single-channel electrocardiogram (ECG) signal. We develop an efficient time-domain analysis to meet the stringent resources constraints of embedded systems to compute the sleep apnea score. Our system, for a publicly available database (PhysioNet Apnea-ECG), has a classification accuracy of up to 88.2% for our new online and patient-specific analysis, which takes the distinct profile of each patient into account. While accurate, our approach is also energy efficient and can achieve a battery lifetime of 46 days for continuous screening of OSA.
Wireless Monitoring of Endogenous and Exogenous Biomolecules on an Android Interface
IEEE Sensors Journal. 2016.
DOI : 10.1109/Jsen.2016.2524631.
Monitoring patients in intensive care units is generally expensive and time consuming. A prompter medical intervention for those critical patients is a key factor for their safety. Therefore, a system that offers immediate visualization of the monitored data represents a great advance in the field. In this paper, the design, the development, and the validation of an android interface for the continuous and wireless monitoring of up to five compounds are described. Continuous monitoring of the biomolecules is addressed by using a fully integrated hardware platform consisting of biosensors connected to a read-out circuit on a printed circuit board. The electrochemical platform uses Roving's Bluetooth module RN-42 to send the measured data to the mobile device. For the validation of the system, some biomolecules are taken as reference: glucose and lactate for endogenous metabolites and paracetamol for exogenous biomolecules. Chronoamperometries are performed at +650 mV for glucose and lactate and at +450 mV for paracetamol. Multi-walled carbon nanotubes are deposited on working electrodes for glucose and lactate for enhanced signal. Instead, for paracetamol, bare working electrodes are used. The measured data are continuously displayed on the screen of the mobile device because of the android interface. Current step for every variation of glucose, lactate, or paracetamol is clearly visible by the trend of the graphs.
Low-Power Wearable System for Real-Time Screening of Obstructive Sleep Apnea
2016. IEEE Computer Society Annual Symposium on VLSI , Pittsburgh, Pennsylvania, U.S.A. , July 11-13, 2016. p. 230-235.
DOI : 10.1109/Isvlsi.2016.51.
Obstructive Sleep Apnea (OSA) is one of the main sleep disorders, but only 10% of the cases are diagnosed. Moreover, there is a lack of tools for long-term monitoring of OSA, since current systems are too bulky and intrusive to be used continuously. In this context, recent studies have shown that it is possible to detect it automatically based on single- lead ECG recordings. This approach can be used in non-invasive smart wearable sensors which measure and process bio-signals online. This work focuses on the implementation, optimization and integration of an algorithm for OSA detection for preventive health-care. It relies on a frequency-domain analysis while tar- geting an ultra-low power embedded wearable device. As it must share its resources usage with other computations, it must be as lightweight as possible. Our current results based on publicly available signals show a classification accuracy of up to 83.2% for both the offline analysis and the embedded online one. This system gives an even better classification accuracy than the best offline algorithm when using the same features for classification
Design of Ultra-Low-Power Smart Wearable Systems
2015. IEEE 16th Latin-American Test Symposium 2015 (LATS) , Puerto Vallarta, Mexico , March 25-27, 2015. p. 1-2.
DOI : 10.1109/LATW.2015.7102527.
Latest progress in microelectronics have enabled a new generation of low cost, low power, miniaturized, yet, smart sensor nodes. This new generation of wearable sensor nodes promise to deploy automated complex bio-signals analysis. In this paper, we present INYU, a wearable sensor device for physical and emotional health monitoring. The device obtains key vital signs of the user, namely Electrocardiogram (ECG), respiration and skin conductance continuously. Using this information, INYU can deliver a novel real-time algorithm for on-line heart-beat classification and correction that relies on a probabilistic model to determine whether a heartbeat is likely to happen under certain timing conditions. Thus, using this algorithm INYU can quickly decide if a beat is occurring at an expected time or if there is a problem in the series (e.g., a skipped, an extra or a misplaced beat). This new algorithm has been integrated in the processing pipeline of automated Heart-Rate Variability (HRV) analysis, both for time-domain (RMSSD, SDNN) and frequency-domain (LF/HF) algorithms.
Real-Time Probabilistic Heart Beat Classification and Correction for Embedded Systems
2015. Computing in Cardiology 2015 , Nice, France , September 06-09, 2015.
With the emergence of wearable and non-intrusive med- ical devices, one major challenge is the real-time analysis of the acquired signals in real-life and ambulatory con- ditions. This paper presents a lightweight algorithm for on-line heart beat classification and correction that relies on a probabilistic model to determine whether a heart beat is likely to happen under certain timing conditions or not. It can quickly decide if a beat is occurring at an expected time or if there is a problem in the series (e.g., a skipped, an extra or a misplaced beat). If an error is detected, the series is repaired accordingly. The algorithm has been carefully optimized to minimize the required processing power and memory usage in order to enable its real-time embedded implementation on a wearable sensing device. Our experimental results, based on the PhysioNet Fanta- sia database, show that the proposed algorithm achieves 99.5% sensitivity in the detection and correction of erro- neous beats. In addition, it features a fast response time when the activity level of the user changes, thus enabling its use in situations where the heart rate quickly changes.
Full System for Translational Studies of Personalized Medicine with Free-Moving Mice (invited)
2015. International Symposium on Circuits and Systems (ISCAS) , Lisbon, Portugal , May 24-27, 2015. p. 1774-1777.
DOI : 10.1109/ISCAS.2015.7168998.
A full remotely powered system for metabolism monitoring of free-moving mice is presented here. The fully implantable sensing platform hosts two ASICs, one off-the-shelf micro-controller, four biosensors, two other sensors, a coil to receive power, and an antenna to transmit data. Proper enzymes ensure specificity for animal metabolites while Multi-Walled Carbon Nanotubes ensure the due sensitivity. The remote powering is indeed provided by inductive coils located under the floor of the mouse' cage. Two different approaches were investigated to ensure freedom of movement to the animal. The application to studies for personalized medicine is demonstrated by showing continuous monitoring of both glucose and paracetamol.