fMRI Analysis via One-class Machine Learning Techniques

2010

In this paper, we demonstrate how one-class recognition of cognitive brain functions across multiple subjects can be performed at the 90% level of accuracy via an appropriate choices of features which can be chosen automatically. The importance of this work is that while one-class is often the appropriate classification setting for identifying cognitive brain functions, most work in the literature has focused on two-class methods. Our work extends one-class work by [1], where such classification was first shown to be possible in principle albeit with an accuracy of about 60%. The results are also comparable to work of various groups around the world e.g.[2], [3] and [4] which have concentrated on two-class classification. The strengthening in the feature selection was accomplished by the use of a genetic algorithm run inside the context of a wrapper approach around a compression neural network for the basic one-class identification. In addition, versions of one-class SVM due to [5] and [6] were investigated.

2007 IEEE Conference on Computer Vision and Pattern Recognition, 2007

Machine learning methods, such as support vector machine (SVM), have been applied to fMRI data analysis, where most studies focus on supervised detection and classification of cognitive states. In this work, we study the general fMRI activation detection using SVM in an unsupervised way instead of the classification of cognitive states. Specifically, activation detection is formulated as an outlier (activated voxels) detection problem of the one-class support vector machine (OCSVM). An OCSVM implementation, ν-SVM, is used where parameter ν controls the outlier ratio, and is usually unknown. We propose a detection method that is not sensitive to ν randomly set within a range known a priori. In cases that this range is also unknown, we consider ν estimation using geometry and texture features. Results from both synthetic and experimental data demonstrate the effectiveness of the proposed methods.

Proceedings of the 9th International Joint Conference on Biomedical Engineering Systems and Technologies

Support Vector Machine (SVM) classifiers are widely used to analyse features extracted from brain MRI data to identify useful biomarkers of pathology in several disease conditions. They are trained to distinguish patients from healthy control subjects by making a binary classification of image features extracted by image processing algorithms. This task is particularly challenging when dealing with psychiatric disorders, as the reported neuroanatomical alterations are often very small and quite un-replicated within different studies. Subtle signs of pathology are difficult to catch especially in extremely heterogeneous conditions such as Autism Spectrum Disorders (ASD). We propose the use of the One-Class Classification (OCC) or Data Description method that, in contrast with two-class classification, is based on a description of one class of objects only. Then, new examples are tested for their similarity to the examples of this target class, end eventually considered as outliers. The application of the OCC to features extracted from brain MRI of children affected by ASD and control subjects demonstrated that a common pattern of features characterize the ASD population.

International Journal of Machine Learning and Cybernetics, 2011

It is generally assumed that one-class machine learning techniques can not reach the performance level of two-class techniques. The importance of this work is that while one-class is often the appropriate classification setting for identifying cognitive brain functions, most work in the literature has focused on two-class methods. In this paper, we demonstrate how one-class recognition of cognitive brain functions across multiple subjects can be performed at the 90% level of accuracy via an appropriate choice of features which can be chosen automatically. Our work extends one-class work by Hardoon and Manevitz (fMRI analysis via one-class machine learning techniques. In: Proceedings of the Nineteenth IJCAI, pp [1604][1605] 2005), where such classification was first shown to be possible in principle albeit with an accuracy of about 60%. The results of this paper are also comparable to work of various groups around the world e.g. Cox and Savoy (NeuroImage 19:261-270, 2003), Mourao-Miranda et al. (NeuroImage, 2006) and Mitchell et al., (Mach Learn 57:145-175, 2004) which have concentrated on two-class classification. The strengthening in the feature selection was accomplished by the use of a genetic algorithm run inside the context of a wrapper approach around a compression neural network for the basic one-class identification. In addition, versions of one-class SVM due to Scholkopf et al. (Estimating the support of a high-dimensional distribution. Technical Report MSR-TR-99-87, Microsoft Research, 1999) and Manevitz and Yousef (J Mach Learn Res 2:139-154, 2001) were investigated.

Journal of Physics: Conference Series

The human mind recognition system is a complex state of mind which can be connected to trigger an external reaction based on internal stimuli. This paper investigates the possibility of recognizing brain signal using signal processing and intelligent technique. The system is differenciate four difference mind thinking, namely moving left hand, moving right hand, moving foot and moving tangue, respectively. The Linear Prediction Coding (LPC) is used to extract the signal, in which the extracted signal is later used as input to the Support Vector Machine (SVM) based identifier. The brain signal parameters are compared and classified to identify the human actions that are intended to be perform. The results of the computer simulation show that this technique produces better accuracy to that of the existing technique based on artificial neural network (ANN), and it requires less training time of one second.

2016

Support Vector Machine (SVM) classifiers are widely used to analyse features extracted from brain MRI data to identify useful biomarkers of pathology in several disease conditions. They are trained to distinguish patients from healthy control subjects by making a binary classification of image features extracted by image processing algorithms. This task is particularly challenging when dealing with psychiatric disorders, as the reported neuroanatomical alterations are often very small and quite un-replicated within different studies. Subtle signs of pathology are difficult to catch especially in extremely heterogeneous conditions such as Autism Spectrum Disorders (ASD). We propose the use of the One-Class Classification (OCC) or Data Description method that, in contrast with two-class classification, is based on a description of one class of objects only. Then, new examples are tested for their similarity to the examples of this target class, end eventually considered as outliers. T...

We describe three experiments combining neuroimaging and machine learning. The first experiment compares the performance of maximum likelihood and neural net classifiers for "brain reading" of fMRI data in the visual cortex. The second experiment applies the optimal classifier to measure the development of the face region in children and adolescents. While the previous experiments used block designs, the third experiment describes an event-related experiment where the classification algorithm learned something real, but not what was planned. The corroboration and validation of the classification results with brain images will be demonstrated.

2009

Pattern recognition methods, which recently have shown promising potential in the analysis of neurophysiological data, are typically model-free and can thus be applied in the analysis of any type of signal. This study demonstrates the feasibility of, after suitable pre-processing steps, applying identical state-of-the-art pattern recognition method to single-trial classification of brain state data acquired with the fundamentally different techniques EEG and fMRI. We investigated linear and non-linear support vector machines (SVM) and artificial neural networks (ANNs), and it was found that the SVM is highly suitable for the classification of both fMRI and EEG single patterns. However, the non-linear classifiers performed better than the linear ones on the EEG data (linear ANN: 66.2%, SVM: 78.9% vs. non-linear ANN: 71.8%, SVM: 83.2%), whereas the opposite was true for the fMRI dataset (linear ANN: 74.4%, SVM: 77.2% vs. non-linear ANN: 70.5%, SVM: 74.2%). The exciting possibility of concurrent EEG and fMRI registration warrants a need for a unified analysis method for both modalities, and we propose pattern recognition for this purpose. The ability to identify cortical patterns on a single-trial basis allows for brain computer interfaces, lie detection, bio-feedback, the tracking of mental states over time, and in the design of interactive, dynamic fMRI and EEG studies.

NeuroImage, 2011

This paper describes a general kernel regression approach to predict experimental conditions from activity patterns acquired with functional magnetic resonance image (fMRI). The standard approach is to use classifiers that predict conditions from activity patterns. Our approach involves training different regression machines for each experimental condition, so that a predicted temporal profile is computed for each condition. A decision function is then used to classify the responses from the testing volumes into the corresponding category, by comparing the predicted temporal profile elicited by each event, against a canonical haemodynamic response function. This approach utilizes the temporal information in the fMRI signal and maintains more training samples in order to improve the classification accuracy over an existing strategy. This paper also introduces efficient techniques of temporal compaction, which operate directly on kernel matrices for kernel classification algorithms such as the support vector machine (SVM). Temporal compacting can convert the kernel computed from each fMRI volume directly into the kernel computed from beta-maps, average of volumes or spatial-temporal kernel. The proposed method was applied to three different datasets. The first one is a block-design experiment with three conditions of image stimuli. The method outperformed the SVM classifiers of three different types of temporal compaction in single-subject leave-one-block-out cross-validation. Our method achieved 100% classification accuracy for six of the subjects and an average of 94% accuracy across all 16 subjects, exceeding the best SVM classification result, which was 83% accuracy (p=0.008). The second dataset is also a block-design experiment with two conditions of visual attention (left or right). Our method yielded 96% accuracy and SVM yielded 92% (p=0.005). The third dataset is from a fast event-related experiment with two categories of visual objects. Our method achieved 77% accuracy, compared with 72% using SVM (p=0.0006).

2008

In this paper we describe a method based on Support Vector machines for Regression (SVR) to decode cognitive states from functional Magnetic Resonance Imaging (fMRI) data. In the context of the Pittsburgh Brain Activity Interpretation Competition (PBAIC, 2007), three participants were scanned during three runs of 20-minute immersion in a Virtual Reality Environment (VRE) where they played a game that engaged them in various search tasks. A set of objective feature ratings was automatically extracted from the VRE during the scanning session, whereas a set of subjective features was then derived from each individual experience. The aim of the present study was to explore the feasibility of the SVR approach in the case of an extremely complex regression problem, in which subjective experience of participants immersed in a VRE had to be predicted from their fMRI data. The proposed methodology was modeled as a multiphase process: a pre-processing phase, based on a filter approach, for fMRI image voxel selection, and a prediction phase, implemented by nonlinear SVR, for decoding subjective cognitive states from the selected voxel time series. Results highlight the generalization ability of nonlinear SVR, making this approach particularly interesting for real world application of Brain Computer Interface (BCI).