Elisabetta De Giovanni
Elisabetta De Giovanni received the M.Sc. degree in Biomedical Engineering, with specialization in Technology for Healthcare, at University of Pavia, Italy, in April 2016. The final thesis, done in the Embedded System Laboratory (ESL) of EPFL, involved applying a smart and low power algorithm to compute vital parameters from physiological signals implemented in an embedded system.
She is currently working in the Embedded Systems Laboratory (ESL), in EPFL, with Prof. David Atienza. The work involves system level co-design of smart multi-parametric Wireless Body Sensor Networks (WBSN), detecting and processing physiological signals.
Entries sorted reverse-chronologicallyLes donées en-ligne ne sont pas accessibles
Hierarchical Cardiac-Rhythm Classification Based on Electrocardiogram Morphology
2017. Computing in Cardiology (CinC) , Rennes, France , September 24-27, 2017.
Atrial Fibrillation (AF) is a type of cardiac arrhythmia that significantly increases the risk of stroke and heart failure. In general, in the case of patients affected by AF, their electrocardiogram (ECG) shows a typical pattern of irregular RR intervals and abnormal P waves. However, discriminating AF from a normal sinus rhythm or from other types of rhythms remains a challenging problem today. Methods: We analyze the database of PhysioNet/Computing in Cardiology Challenge 2017 to validate our heart rhythm classification technique. The database contains short-term ECG recordings, labelled as normal sinus rhythm, AF, other types of rhythm, and noise. We extract different morphology-based features of ECG signals, and we design a multiclass classifier based on error-correcting output codes, along with a random forest classifier for binary decision making. Results: We test the performance of our classifiers based on the F1 score of each class and the average F1 score of all the classes. The final F1 score obtained on the hidden test set of challenge is 80%. Conclusions: Our results show that our classifier is robust and that it is able to discriminate AF from normal sinus, other rhythms, and noise, based on the morphology of the ECG signal.
A Patient-Specific Methodology for Prediction of Paroxysmal Atrial Fibrillation Onset
2017. Computing in Cardiology , Rennes, France , September 24-27, 2017.
DOI : 10.22489/CinC.2017.285-191.
In spite of the progress in management of Atrial Fibrillation (AF), this arrhythmia is one of the major causes of stroke and heart failure. The progression of this pathology from a silent paroxysmal form (PAF) into a sustained AF can be prevented by predicting the onset of PAF episodes. Moreover, since AF is caused by heterogeneous mechanisms in different patients, as we demonstrate in this paper, a patient-specific approach offers a promising solution. In this work, we consider two ECG recordings, one close to PAF onset and one far away from any PAF episode. For each patient, we extract two 5-minute ECG segments approximately 20 minutes apart. Next, we train a linear Support Vector Machine (SVM) classifier using patient-specific sets of time- and amplitude-domain features. In particular, we consider the P-waves and the QRS complexes in short windows of 5 consecutive heart beats. Finally, we validate the method on the PAF Prediction Challenge (2001) PhysioNet database predicting the onset with an F1 score of 97.1\%, sensitivity of 96.2\% and specificity of 98.1\%.
Guidelines for the use and interpretation of assays for monitoring autophagy (3rd edition)
DOI : 10.1080/15548627.2015.1100356.
Ultra-Low Power Estimation of Heart Rate Under Physical Activity Using a Wearable Photoplethysmographic System
2016. 19th IEEE/Euromicro Conference On Digital System Design (DSD 2016) , Limassol, Cyprus , August 31 - September 2, 2016. p. 1-10.
DOI : 10.1109/Dsd.2016.101.
In the last years, the need for enhancing health and preventing problems with remote monitoring is increasing. A non-invasive low-cost technique for processing bio-signals and monitoring vital parameters, at rest and during physical activity, is the use of wearable PhotoPlethysmoGraphic (PPG) systems. However, in order to detect a relevant vital parameter, such as the heart rate during demanding exercises, motion artifacts must be removed from the signals retrieved. In this paper, we present a fast and easy to implement algorithm to estimate the heart rate value which does not need to reconstruct the noise-free signal nor does it apply adaptive filtering as existing algorithms, thus gaining computational time and stored memory space. The method consists of applying the Fast Fourier Transform on short windows of data and removing motion artifacts relying on single-sided amplitude spectrum analysis of PPG and 3-axis accelerometer signals. The results show that our algorithm manages to remove a wide range of motion artifacts achieving an average absolute error of only 1.27 BPM between the heart rate estimated by the algorithm every second and the ground-truth value. The method was successfully implemented on a wearable PPG device achieving an execution time of 226 ms per second, hence obtaining a battery lifetime of 9.37 days.
A neural approach to drugs monitoring for personalized medicine
2015. International Joint Conference on Neural Networks 2015 (IJCNN) , Killarney, Ireland , July 12-17, 2015.
DOI : 10.1109/IJCNN.2015.7280611.
The development of fast and mobile drug detection is an important aspect of personalized medicine. It enables the quick assessment of inter-individual differences in drug metabolism and corresponding adjustments of the dose. Recent developments of amperometric biosensors using cytochrome P450 (CYP) show great promise, by lowering the detection limit to physiological range for several drugs via the usage of Multi Walled Carbon Nanotubes (MWCNT). The next challenge is to develop algorithms for processing the resulting sensor data compatible with low-power hardware, which would allow the development of portable battery-powered devices. In this work we pursue a novel approach to this problem. Here we provide a proof of principle by demonstrating how sensor data could be analyzed using a conventional multi-layer perceptron network with error-backpropagation.