Parkinson's disease (PD) is the second most common and chronic neurodegenerative disorder after Alzheimer's Disease. At the same time, stroke is one of the leading causes of disability and death. Both PD and stroke are neurological diseases and are almost always coupled with EEG alterations. However, the studies that correlate the EEG features with the clinical scales and treatment outcomes are rare. As a prerequisite for the efficient diagnosis, disease progression monitoring, neurorehabilitation, and outcome prediction, it is required to study the alterations of stroke and PD subjects' cerebral rhythms. Series of novel therapeutic protocols for motor performance improvement, including those based on BCI, can benefit from these findings. BCIs have shown a promising result for motor and cognitive neurorehabilitation for PD and stroke patients. In conjunction with Motor Imagery, BCI can guide a patient to a functional recovery by real-time acquisition, processing, and feeding back information on his task engagement. The MI-BCI application creates a more controlled rehabilitation environment since the MI-induced oscillatory activity can be monitored to assess whether the patient performs the task correctly. For MI-BCI to reach the point where it is consistently successful as a neurorehabilitation tool, the new EEG signal processing techniques have to be studied and developed. Moreover, the alteration of EEG rhythms in PDs and stroke has to be considered to design a personalised and robust BCI system. During the PhD course, a series of studies have been conducted to study EEG alterations and neurophysiological deficits of stroke and PD's, signal processing, machine learning, and classification techniques needed for BCI modelling. Our results on EEG spectral features and clinical data show that the EEG confirmed to be a sensitive measure for brain functions in the earliest phase of cerebral ischemia and that EEG can be used as complementary in the evaluation of stroke severity and as a potentially useful tool in monitoring and mapping longitudinal changes in acute stroke patients. Furthermore, the significant correlation between EEG spectral features and symptom-specific motor decline indicates that EEG assessment may be a useful biomarker for objective monitoring of disease progression and as evaluation measure of the effect of PD's rehabilitation approaches. In the second part of the thesis, our study demonstrates the effect of cortical modulation induced by MI on the EEG resting-state and provides support for further development of Motor Imagery based BCI (MI-BCI). The third part shows that PD's and stroke patients could control MI-BCI, with high accuracy and that FBCSP may be used as the MI-BCI approach for complementary neurorehabilitation. The thesis's additional novelty is the proposal of the transfer learning-based multi-session extension FBCSP approach. The new approach has been tested, and the study shows that significantly improves BCI model calibration accuracy in PD patients. Finally, our last study results showed that the signal nonstationarities and power covariance shifts significantly reduce BCI models' accuracy. However, only after introducing the Stationary Subspace Analysis (SSA) preprocessing the classifier's performance is significantly increased. The abovementioned main findings of this thesis may improve the present BCI systems in terms of accuracy and usability and enhance the diffusion and beneficial aspects of MI-BCI neurorehabilitation to PD and stroke patients.

La malattia di Parkinson (PD) è la seconda malattia neurodegenerativa più comune dopo la malattia di Alzheimer. Allo stesso tempo, l'ictus cerebrale è una delle principali cause di disabilità e morte nel mondo. La mattina di Parkinson e l'ictus cerebrale ischemico sono patologie neurologiche che provocano alterazioni del segnale elettroencefalografico (EEG). Tuttavia, ci sono pochi studi che mettono in relazione le alterazioni EEG con il deficit neurologico per un migliore monitoraggio della progressione della malattia, per la neuroriabilitazione personalizzata e per la previsione dell’outcome clinico. I protocolli terapeutici avanzati per il miglioramento delle prestazioni motorie, compresi quelli basati sulla Brain Computer Interface (BCI), possono beneficiare dell’identificazione delle alterazioni EEG e del legame tra queste e il deficit specifico. Le tecniche BCI si sono dimostrate promettenti nella neuroriabilitazione motoria e cognitiva nei pazienti parkinsoniani e post-ictus specialmente attraverso l’utilizzo dell’Immaginazione Motoria (Motor Imagery, MI) supportata dalla BCI (MI-BCI) in grado di creare un ambiente di riabilitazione più controllato venendo fornito al paziente un feedback sulla corretta esecuzione del task riabilitativo. Per migliorare l’applicabilità, la prestazione e l’efficacia di questa strategia riabilitativa, le nuove tecniche di elaborazione del segnale EEG devono essere studiate e sviluppate. Inoltre, le alterazioni dei parametri EEG nei pazienti PD e ictus devono essere considerate nella progettazione di un sistema BCI personalizzato e robusto. Durante il corso di dottorato, sono stati condotti una serie di studi per identificare le correlazioni tra alterazioni EEG e il deficit neurofisiologico relativo all’ ictus e alla malattia di Parkinson. Inoltre, sono stati condotti degli studi per identificare le tecniche di elaborazione del segnale, di apprendimento automatico e di classificazione più appropriate per la modellazione BCI in queste popolazioni di pazienti. I risultati ottenuti ed in particolare la correlazione significativa tra i parametri spettrali dell’EEG e le scale cliniche di interesse hanno confermato l’ipotesi che i parametri EEG sono sensibili ai cambiamenti delle funzioni cerebrali nella prima fase dell'ischemia, che possono quindi essere utilizzati sia nella valutazione della gravità dell'ictus sia come strumento di monitoraggio e mappatura dei cambiamenti longitudinali nel paziente con ictus. Inoltre, le correlazioni identificate tra il parametri spettrali dell’EEG e il deficit motorio nei parkinsoniani indicano che la valutazione EEG può essere un biomarcatore utile per il monitoraggio obiettivo della progressione della patologia e dell'efficacia delle diverse strategie riabilitative. Nella seconda parte della tesi, viene descritto uno studio che ha evidenziato la modulazione corticale indotta dalla MI sull’EEG durante il resting-state, supportando l’ipotesi dell’efficacia della MI-BCI come strategia neuroriabilitativa. Nella terza parte sono riportati i risultati degli studi condotti su pazienti PD e su quelli con ictus che hanno dimostrato che entrambe le popolazioni, caratterizzate di deficit motorio, erano in grado di controllare la MI-BCI con elevata precisione. Inoltre, viene dimostrato che la migliore performance in termini di accuratezza di classificazione è stata ottenuta con la tecnica di preprocessing Filter Bank Common Spatial Patterns (FBCSP). Nel lavoro di tesi si propone anche una estensione dell'approccio FBCSP basato sul multi-session transfer learning. I risultati di questa tesi possono contribuire al miglioramento in termini di accuratezza di classificazione e di usabilità degli attuali sistemi BCI, aumentando la diffusione e gli aspetti benefici della neuroriabilitazione MI-BCI nei pazienti con PD e ictus.

Advanced MI-BCI procedures for neurorehabilitation of PD's and post-stroke patients / Miladinović, Aleksandar. - (2021 Apr 19).

Advanced MI-BCI procedures for neurorehabilitation of PD's and post-stroke patients

MILADINOVIĆ, ALEKSANDAR
2021-04-19

Abstract

Parkinson's disease (PD) is the second most common and chronic neurodegenerative disorder after Alzheimer's Disease. At the same time, stroke is one of the leading causes of disability and death. Both PD and stroke are neurological diseases and are almost always coupled with EEG alterations. However, the studies that correlate the EEG features with the clinical scales and treatment outcomes are rare. As a prerequisite for the efficient diagnosis, disease progression monitoring, neurorehabilitation, and outcome prediction, it is required to study the alterations of stroke and PD subjects' cerebral rhythms. Series of novel therapeutic protocols for motor performance improvement, including those based on BCI, can benefit from these findings. BCIs have shown a promising result for motor and cognitive neurorehabilitation for PD and stroke patients. In conjunction with Motor Imagery, BCI can guide a patient to a functional recovery by real-time acquisition, processing, and feeding back information on his task engagement. The MI-BCI application creates a more controlled rehabilitation environment since the MI-induced oscillatory activity can be monitored to assess whether the patient performs the task correctly. For MI-BCI to reach the point where it is consistently successful as a neurorehabilitation tool, the new EEG signal processing techniques have to be studied and developed. Moreover, the alteration of EEG rhythms in PDs and stroke has to be considered to design a personalised and robust BCI system. During the PhD course, a series of studies have been conducted to study EEG alterations and neurophysiological deficits of stroke and PD's, signal processing, machine learning, and classification techniques needed for BCI modelling. Our results on EEG spectral features and clinical data show that the EEG confirmed to be a sensitive measure for brain functions in the earliest phase of cerebral ischemia and that EEG can be used as complementary in the evaluation of stroke severity and as a potentially useful tool in monitoring and mapping longitudinal changes in acute stroke patients. Furthermore, the significant correlation between EEG spectral features and symptom-specific motor decline indicates that EEG assessment may be a useful biomarker for objective monitoring of disease progression and as evaluation measure of the effect of PD's rehabilitation approaches. In the second part of the thesis, our study demonstrates the effect of cortical modulation induced by MI on the EEG resting-state and provides support for further development of Motor Imagery based BCI (MI-BCI). The third part shows that PD's and stroke patients could control MI-BCI, with high accuracy and that FBCSP may be used as the MI-BCI approach for complementary neurorehabilitation. The thesis's additional novelty is the proposal of the transfer learning-based multi-session extension FBCSP approach. The new approach has been tested, and the study shows that significantly improves BCI model calibration accuracy in PD patients. Finally, our last study results showed that the signal nonstationarities and power covariance shifts significantly reduce BCI models' accuracy. However, only after introducing the Stationary Subspace Analysis (SSA) preprocessing the classifier's performance is significantly increased. The abovementioned main findings of this thesis may improve the present BCI systems in terms of accuracy and usability and enhance the diffusion and beneficial aspects of MI-BCI neurorehabilitation to PD and stroke patients.
19-apr-2021
BABICH, FULVIO
33
2019/2020
Settore ING-INF/06 - Bioingegneria Elettronica e Informatica
Università degli Studi di Trieste
File in questo prodotto:
File Dimensione Formato  
Aleksandar_Miladinovic_BCI_PhD_Thesis.pdf

accesso aperto

Descrizione: Aleksandar_Miladinovic_BCI_PhD_Thesis
Tipologia: Tesi di dottorato
Dimensione 3.56 MB
Formato Adobe PDF
3.56 MB Adobe PDF Visualizza/Apri
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11368/2988317
 Avviso

Registrazione in corso di verifica.
La registrazione di questo prodotto non è ancora stata validata in ArTS.

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact