This thesis details the implementation and evaluation of seven machine learning classifiers for the detection of Autism Spectrum Disorder (ASD) using resting-state functional MRI (rs-fMRI) data from the ABIDE I dataset. Two feature representations were compared: traditional Pearson correlation features and graph-based features. Logistic Regression (LR), Support Vector Machine (SVM), and Multi-Layer Perceptron (MLP) achieved the highest performance on Pearson correlation features, satisfying all predefined non-functional requirements, with average balanced accuracies up to 64.4% (SVM) and standard deviations below 2.5%. Linear Discriminant Analysis (LDA) narrowly missed the standard deviation constraint with 0.5%.
In contrast to the Pearson correlation features, graph-based features yielded consistently lower balanced accuracies, typically ranging from 54% to 59% across classifiers, underscoring their limited informativeness in the current implementation. Feature importance analysis on Pearson correlation data revealed connections between brain regions involving the inferior occipital gyrus, middle temporal pole, precuneus, and cerebellum as
consistently influential for classification. To facilitate neuroscientific exploration, an interactive tool, NASDA (Neuroimaging Autism Spectrum Disorder Analyser), was developed and demonstrated to fulfil all functional and non-functional requirements for Pearson correlation based analysis using LR as the recommended classification model.
These results highlight the dependency of classifiers performance on the quality of input features and contribute to ongoing efforts to localise robust neurological biomarkers for ASD.

This research is part of a broader study on Autism Spectrum Disorder (ASD) detection, which consists of three main components: feature design, classification, and feature selection. The primary objective of this study is to investigate whether different graph inference methods and graph-based features can contribute to finding new biomarkers in the brains of individuals with ASD. To easily interpret and visualize the results, a graphical user interface (GUI) was developed. The feature design phase includes both classical statistical features and innovative graph-based features, derived through a variety of statistical and Graph Signal Processing (GSP) graph inference methods. The top three performing methods were: sample covariance with graphical lasso (59.3% accuracy, 64.2% AUC), rSpecT-adjacency (58.0% accuracy, 62.5% AUC), and rSpecT-laplacian (58.8% accuracy, 61.5% AUC). the connectivity patterns visualized using the developed tools revealed distinct differences between methods, offering promising insights for neuroscientific interpretation and potential biomarker discovery.

In the Netherlands, 3% of people above 4 years old are diagnosed with autism. Diagnosing is currently done with a psychological assessment, but classifying people with autism using resting state functional magnetic imaging, or rs-fMRI, has become promising. The goal of this project was to see if new features could be found, based on the graph of rs-fMRI data stored in the Autism Brain Imaging Data Exchange (ABIDE) dataset, that had a significant positive influence on the accuracy of a predictive model. To do this, several feature selection modules were researched and coded in Python. Subsequently, these were tested with the features and classification methods created by our partner subgroups. The best performing model, using all data, had an accuracy of 74.26% and a sensitivity of 65.35%. The best performing model using graph features, on all data, had an accuracy of 60.4% and a sensitivity of 46.6%. This indicates that there were no graph features developed that had a significant positive influence on classifying whether someone has autism.

In the last decade, the new functional neuroimaging technique functional ultrasound (fUS) has emerged as a potential new tool for clinical and neuroscientific applications. Unlike several conventional methods for functional brain imaging, fUS offers an unparalleled combination of submillimeter-subsecond spatiotemporal resolution and the ability to penetrate deep into brain tissue and capture large areas of interest. This makes fUS an exceptionally valuable tool for investigating brain function. The principal physiological mechanism utilized by fUS imaging is neurovascular coupling, which connects neuronal activity with local hemodynamic changes. Within the brain, increased neuronal activity results in the dilation of nearby blood vessels, subsequently leading to an increased local blood flow and volume.
In fUS, the increase in blood volume is measured by the intensity of the Doppler signal caused by the Doppler effect. To obtain relevant information from functional neuroimaging data, blind source separation (BSS) is often used. BSS is the separation of a set of source signals from a set of mixed signals, without the aid of information about the source signals or the mixing process. A commonly used BSS method for identifying brain networks and artifacts in functional neuroimaging techniques like fMRI is independent component analysis, which is a low-rank matrix decomposition. This work explores the use of canonical polyadic decomposition (CPD), a low-rank tensor decomposition, for BSS of fUS data. The CPD approximates a 3rd-order tensor, consisting of a space, frequency, and time dimension, by a sum of rank-1 tensors. The proposed method offers three key advantages that make it highly relevant to the field of fUS signal processing: 1) Tensor-based BSS by CPD allows for frequency as a third dimension to complement the spatial and temporal dimensions; 2) Tensor-based BSS by CPD allows for the use of all raw fUS data (compound images) rather than only power doppler images, exploiting more available information; and 3) Tensor-based BSS by CPD allows for a BSS method without strong constraints such as the case with matrix-based methods like independent component analysis. The full processing pipeline developed in this work consists of four distinct stages. In stage one, pre-processing is carried out through SVD filtering, temporal demeaning, and pixel-based normalization. During the SVD filtering, normalization of the singular values is performed, which is novel. In stage two, compression is performed using truncated multilinear singular value decomposition. In stage three, BSS is carried out using CPD on the compressed data. In stage four, stable components are extracted by running the CPD 100 times and clustering the resulting components. These clusters were then averaged to create mean components. The CPD rank was estimated based on two quantifiable aspects: 1) Correlation between mean components and 2) Frequency of occurrence of mean components. The method is entirely data-driven, can be applied to entire raw fUS datasets, and can run on a laptop with standard RAM. Application to task experiment data of a mouse verified that activity that is temporally correlated with the stimulus can be extracted in expected regions. The components also indicate expected functional connectivity, which is often not the case for matrix-based methods. Moreover, frequency spectra showed different characteristics for different types of components, which displays the relevance of this dimension for BSS, especially for nuisance components. In conclusion, the proposed tensor-based blind source separation pipeline for raw fUS data was able to identify artifacts and meaningful neurological components based on distinctive characteristics in the temporal, spatial, and spectral domains. This work serves as a starting point for more advanced denoising, the identification of novel brain networks, and the comparison of brain networks across healthy and pathological conditions. Nevertheless, further research is necessary to ensure the utility of the method.