Recent research has demonstrated that functional connectivity profiles act as a unique fingerprint that can accurately identify subjects from a large group. As a result, brain connectomics has emerged as a rapidly growing research field that focuses on identifying individuals based on inter-individual variability in brain connectivity during resting-state and task-evoked responses. The main objective of this study, which is part of the ERC HANDmade project (SH4, ERC-2017-STG) led by Prof. Betti Viviana, is to investigate and predict inter-individual differences in motor tasks based on resting-state functional connectivity measured with magnetoencephalography (MEG). Several regression-based models are discussed and compared to improve the identification rate on 51 subjects from the Human Connectome Project. A residualisation approach is also presented to enhance model performance by increasing variability across subjects. Results indicate that interindividual differences in brain connectomes during spontaneous and task-evoked activity can be accurately predicted. However, it is observed that no improvement can be obtained in the identification rate considering the muscle activity, compared to Vettoruzzo’s MSc thesis. Regardless of the model complexity and the number of the EMG features considered, the results show the inadequacy of predicting muscular activity from individual brain connectivity. Therefore, the investigation is addressed to understand the relationship between the brain activity and muscular activity, and non-linear approaches employing neural networks are considered the best choices for this type of motor tasks. The study’s findings can have implications in clinical research and rehabilitation fields for comprehending the neural mechanism underlying various neurological and muscular disorders and for developing personalized treatment approaches.

Recent research has demonstrated that functional connectivity profiles act as a unique fingerprint that can accurately identify subjects from a large group. As a result, brain connectomics has emerged as a rapidly growing research field that focuses on identifying individuals based on inter-individual variability in brain connectivity during resting-state and task-evoked responses. The main objective of this study, which is part of the ERC HANDmade project (SH4, ERC-2017-STG) led by Prof. Betti Viviana, is to investigate and predict inter-individual differences in motor tasks based on resting-state functional connectivity measured with magnetoencephalography (MEG). Several regression-based models are discussed and compared to improve the identification rate on 51 subjects from the Human Connectome Project. A residualisation approach is also presented to enhance model performance by increasing variability across subjects. Results indicate that interindividual differences in brain connectomes during spontaneous and task-evoked activity can be accurately predicted. However, it is observed that no improvement can be obtained in the identification rate considering the muscle activity, compared to Vettoruzzo’s MSc thesis. Regardless of the model complexity and the number of the EMG features considered, the results show the inadequacy of predicting muscular activity from individual brain connectivity. Therefore, the investigation is addressed to understand the relationship between the brain activity and muscular activity, and non-linear approaches employing neural networks are considered the best choices for this type of motor tasks. The study’s findings can have implications in clinical research and rehabilitation fields for comprehending the neural mechanism underlying various neurological and muscular disorders and for developing personalized treatment approaches.

Development of machine learning-based regression models to predict inter-individual differences from MEG-based resting-state functional connectivity

FONGARO, ENRICO
2022/2023

Abstract

Recent research has demonstrated that functional connectivity profiles act as a unique fingerprint that can accurately identify subjects from a large group. As a result, brain connectomics has emerged as a rapidly growing research field that focuses on identifying individuals based on inter-individual variability in brain connectivity during resting-state and task-evoked responses. The main objective of this study, which is part of the ERC HANDmade project (SH4, ERC-2017-STG) led by Prof. Betti Viviana, is to investigate and predict inter-individual differences in motor tasks based on resting-state functional connectivity measured with magnetoencephalography (MEG). Several regression-based models are discussed and compared to improve the identification rate on 51 subjects from the Human Connectome Project. A residualisation approach is also presented to enhance model performance by increasing variability across subjects. Results indicate that interindividual differences in brain connectomes during spontaneous and task-evoked activity can be accurately predicted. However, it is observed that no improvement can be obtained in the identification rate considering the muscle activity, compared to Vettoruzzo’s MSc thesis. Regardless of the model complexity and the number of the EMG features considered, the results show the inadequacy of predicting muscular activity from individual brain connectivity. Therefore, the investigation is addressed to understand the relationship between the brain activity and muscular activity, and non-linear approaches employing neural networks are considered the best choices for this type of motor tasks. The study’s findings can have implications in clinical research and rehabilitation fields for comprehending the neural mechanism underlying various neurological and muscular disorders and for developing personalized treatment approaches.
2022
Development of machine learning-based regression models to predict inter-individual differences from MEG-based resting-state functional connectivity
Recent research has demonstrated that functional connectivity profiles act as a unique fingerprint that can accurately identify subjects from a large group. As a result, brain connectomics has emerged as a rapidly growing research field that focuses on identifying individuals based on inter-individual variability in brain connectivity during resting-state and task-evoked responses. The main objective of this study, which is part of the ERC HANDmade project (SH4, ERC-2017-STG) led by Prof. Betti Viviana, is to investigate and predict inter-individual differences in motor tasks based on resting-state functional connectivity measured with magnetoencephalography (MEG). Several regression-based models are discussed and compared to improve the identification rate on 51 subjects from the Human Connectome Project. A residualisation approach is also presented to enhance model performance by increasing variability across subjects. Results indicate that interindividual differences in brain connectomes during spontaneous and task-evoked activity can be accurately predicted. However, it is observed that no improvement can be obtained in the identification rate considering the muscle activity, compared to Vettoruzzo’s MSc thesis. Regardless of the model complexity and the number of the EMG features considered, the results show the inadequacy of predicting muscular activity from individual brain connectivity. Therefore, the investigation is addressed to understand the relationship between the brain activity and muscular activity, and non-linear approaches employing neural networks are considered the best choices for this type of motor tasks. The study’s findings can have implications in clinical research and rehabilitation fields for comprehending the neural mechanism underlying various neurological and muscular disorders and for developing personalized treatment approaches.
Machine learning
AI
connectivity
EMG
MEG
File in questo prodotto:
File Dimensione Formato  
Fongaro_Enrico.pdf

accesso aperto

Dimensione 20.51 MB
Formato Adobe PDF
20.51 MB Adobe PDF Visualizza/Apri

The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.12608/45156