Repo of papers.
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
This is a repository with the written papers over the years.
Given the importance eHealth has assumed in the recent years, in this paper I present
how IoT can be used in the healthcare field in particular using LoRa.
A paper in Italian about serendipity and Oersted’s electromagnetism discovery.
The following papers have been written by some collegues and me.
In this page I’m reporting the papers titles and their abstract, you can check them out by clicking on their respective links.
Multi-class classification from single lead ECG recordings
The automatic classification of heart rhythms using
short time single lead ECG recordings is a challenging task that
has been widely studied recently.
In this paper we present our work that aims at classifying these
kind of ECG signals as Atrial Fibrillation (Afib), Normal, Other
rhythms or too noisy to be classified (Noisy). We developed three
different binary classifiers as Recurrent Neural Networks (RNNs)
both with a binary cross-entropy loss function and a weighted
version of it. We used these three RNNs to develop a cascade
classifier for the samples of the given dataset, considering the
problem as a multiple binary classification problem.
We obtained similar results, with a slightly better result using
the unweighted loss function, with an accuracy of 81.18% vs
80.01% and a F1 score of 0.77 vs 0.76.
A Motor Imagery based Brain Computer Interface to restore upper limb movements
Spinal Cord Injury (SCI) is a condition that
causes, for patients suffering from it, a huge lack of autonomy.
This is very expensive, both for families and society, as people
are often totally dependent on others also for the most basic
and everyday situations. In the recent years lot of investments
have been made for improving their lifestyle and autonomy.
Although several different approaches have been developed for
many BCI systems, we decided to implement our own setup
for SCI patients based on MI literature, and in particular on
MI training before the actual use of the BCI. Studies revealed
that in SCI there are several departures from healthy subjects
brain patterns, along with other preserved motor functions.
We analysed these brain activation patterns for upper limb
movements and we developed both a non-invasive and an
invasive BCI system. The former is based on FES, electrical
stimulation of arm and hand muscles, and the latter on an
implanted device called bridge, which aims to restore the
damages in the spinal cord bypassing them. Supported by the
literature, our results seem promising and we now expect to
implement the actual system and start the clinical trial.