RT info:eu-repo/semantics/doctoralThesis
T1 From local activation to global synchronisation: Methodological challenges of M/EEG analysis for biomedical applications
A1 Rodríguez González, Víctor
A2 Universidad de Valladolid. Escuela de Doctorado
K1 Electroencefalografía
K1 Electroencephalography
K1 Electroencefalografía
K1 Magnetoencephalography
K1 Magnetoencefalografía
K1 Neural signal analysis
K1 Análisis de señales neuronales
K1 2490 Neurociencias
AB Neuroscience has dedicated centuries to the study of the human brain and its functions, leading to significant strides in comprehending the neural underpinnings of complex cognitive processes. Since their introduction in the 20th century, neurophysiological recordings have emerged as a pivotal tool that has greatly contributed to this expansive comprehension of the intricacies of the brain. Despite the increasing knowledge gained in the last decades, the human brain remains the most profound mystery in our anatomy, with numerous challenges and opportunities for further exploration and understanding.There are a wide variety of metrics that have been used to characterise electroencephalographic and magnetoencephalographic (M/EEG) signals and the alterations that the neurological and psychiatric disorders elicit on them. Nevertheless, most of them can be grouped into one of these two categories: (i) local activation analyses, which assess M/EEG signals from the sensors or brain regions individually; and (ii) global synchronisation analyses, that consider the brain operation as an integrated network. This Doctoral Thesis is focused on delving into the existing literature about the analysis of M/EEG signals, identifying potential methodological challenges and, subsequently, proposing effective solutions to address them. Furthermore, these methodological advancements opened new avenues for innovative clinical applications of M/EEG signals in the context of mild cognitive impairment (MCI) and dementia associated with Alzheimer's disease (AD).Throughout the studies included in this Doctoral Thesis, we have employed five different databases of M/EEG recordings, accumulating more than 550 resting-state neural recordings encompassing healthy controls, patients with MCI, and patients with AD. Of note, this comprehensive dataset also includes longitudinal recordings from patients with AD that underwent a non-pharmacological therapy (NPT) against dementia.The main contributions of this Thesis were organised based on the two aforementioned levels of analysis, with a methodological advance followed by a clinical application within each level. First, (i) it was observed a substantial consistency in the local activation parameters between sensor- and source-level M/EEG signals, which suggests that, for these metrics, the source inversion process becomes necessary only when the spatial dimension significantly influences the results. Also, (ii) it was demonstrated the validity of local activation metrics to predict the outcome of a NPT against AD, with the spatial dimension emerging as a crucial factor in this predictive capability. Moreover, (iii) it was developed a subject-specific data-driven algorithm for frequency band segmentation that overcomes the limitations of the current approaches: the Connectivity-based Meta-Bands (CMB) algorithm. Finally, (iv) the potential of CMB to detect alterations that MCI and AD provoke in the frequency structure of the neural network was assessed.The contributions of the Thesis have successfully addressed the long-standing question of whether to conduct local-activation analyses at the sensor or source level. The findings revealed a high consistency in metrics at both levels, albeit with some spatial dimension information missing at the sensor level.  This broadened knowledge about this type of metrics led to consider them as potential predictors of the outcome of a NPT against AD, demonstrating their predictive potential. Additionally, the introduction of the CMB algorithm marked a significant advancement by providing a data-driven, subject-specific frequency band segmentation. This algorithm unveiled the underlying frequency-dependent structure of functional neural networks, presenting an intriguing alternative to canonical frequency bands. Lastly, the application of CMB facilitated the characterisation of pathological alterations induced by MCI and AD in the frequency-dependent structure of the brain network. In conclusion, the present Doctoral Thesis has addressed two methodological open challenges, proposed relevant solutions to them, and assessed potential clinical applications that opened up new possibilities for future research in the field.
YR 2024
FD 2024
LK https://uvadoc.uva.es/handle/10324/69490
UL https://uvadoc.uva.es/handle/10324/69490
LA eng
NO Escuela de Doctorado
DS UVaDOC
RD 21-nov-2024