A challenge in clinical neuroscience is to identify systematic differences between patients’ resting state networks (RSNs) that may relate to phenotypes and/or disease subtypes. Recently, Durieux...Show moreA challenge in clinical neuroscience is to identify systematic differences between patients’ resting state networks (RSNs) that may relate to phenotypes and/or disease subtypes. Recently, Durieux et al. (2022) proposed a novel unsupervised learning method for multi-subject rs-fMRI data, called Clusterwise Independent Component Analysis (C-ICA), which enables automatic clustering of subjects based on differences and similarities in subjects’ RSNs. A drawback of C-ICA, however, is that the method assumes that subject clusters show large difference in the RSNs that characterizes them. This, however, is a quite restrictive assumption as it can be expected that there exists important RSNs that are common to all subjects (across clusters). Not accounting for these common RSNs may negatively affect C-ICA’s ability to identify the subject clustering and the RSNs for each cluster. Therefore, in order to tackle this problem, two extensions of C-ICA (CCD1 and CCD2) are presented and compared to each other that allow to distinguish between (and extract from the data) both common and distinctive components. Both extensions differ in the procedure to identify the common components, with CCD2 selecting the common components out of a wider pool of possible components than CCD1. In a simulation study, the two C-ICA extensions are compared to C-ICA while manipulating the following factors: number of clusters, number of common and distinctive components, degree of overlap among clusters and amount of noise. Performance was determined by computing cluster recovery and component recovery, where component recovery was studied separately for the common components, the distinctive components and all components. Both extensions clearly outperformed C-ICA, with CCD2 performing the best. A mixed analysis of variance (ANOVA) showed that, the degree of overlap and the amount of noise in the data were the most influential factors, with large amounts of overlap and noise decreasing cluster and component recovery. Moreover, both extensions showed a slower deterioration of cluster and component recovery -with increasing overlap and noise- than C-ICA. Specifically for intermediate amounts of noise and overlap the differences between both extensions and C-ICA were the largest.Show less
Electroencephalography (EEG) has been used for over a century to measure electric potentials in the brain. The brain signals can be used in a Brain Computer Interface (BCI) to control machines,...Show moreElectroencephalography (EEG) has been used for over a century to measure electric potentials in the brain. The brain signals can be used in a Brain Computer Interface (BCI) to control machines, like prostheses or exoskeletons. In the last decades, consumer-grade EEG devices have become available. Their performance and reliability, however, does not match that of medical EEG equipment. In this project, the possibilities for classification with such a consumer-grade EEG device, the EMOTIV EPOC+, were explored. The main challenge was to overcome the large noise and nuisances in the EEG signal to be able to classify brain states induced by motoric tasks and sensory stimuli. The effect of several preprocessing techniques was studied: Independent Component Analysis, a denoising autoencoder, and adaptive filtering with an autoencoder. For the classification, various machine learning approaches were proposed to analyse the EEG signals, ranging from a naive Bayes classifier to Deep Convolutional Neural Networks. Furthermore, the effect of motivating the structure of a neural network for this classification task was investigated. It was found that using the spectral features of the EEG signal as a motivation for the structure of neural networks helps classifiers identify brain states that are not classified by other types of neural networks.Show less