Automatic Bayesian Classification of Healthy Controls, Bipolar Disorder, and Schizophrenia Using Intrinsic Connectivity Maps From fMRI Data

Abstract

We present a method for supervised, automatic and reliable classification of healthy controls, patients with bipolar disorder and patients with schizophrenia using brain imaging data. The method uses four supervised classification learning machines trained with a stochastic gradient learning rule based on the minimization of Kullback-Leibler divergence and an optimal model complexity search through posterior probability estimation. Prior to classification, given the high dimensionality of functional magnetic resonance imaging data, a dimension reduction stage comprising of two steps is performed: first, a one sample univariate t-test mean difference T score approach is used to reduce the number of significant discriminative functional activated voxels, and then singular value decomposition (SVD) is performed to further reduce the dimension of the input patterns to a number comparable to the limited number of subjects available for each of the three classes. Experimental results using functional brain imaging (fMRI) data include receiver operation characteristic (ROC) curves for the 3-way classifier with area under curve (AUC) values around 0.82, 0.89, and 0.90 for healthy control versus non-healthy, bipolar disorder versus non-bipolar and schizophrenia patients versus nonschizophrenia, binary problems respectively. The average 3-way correct classification rate is in the range of 70 − 72%, for the test set, remaining close to the estimated Bayesian optimal correct classification rate theoretical upper bound of about 80%, estimated from the performance of the 1-nearest neighbor classifier over the same data.

Multivariate Classification of Blood Oxygen Level-Dependent fMRI Data with Diagnostic Intention: A Clinical Perspective

“…Only 5 recent studies have tried to overcome this issue by introducing individual confidence measures. 39,48,54,56,57 As seen in Fig 2 and On-line larger body of independent work. The diversity of scientific backgrounds of recent studies is reflected by a striking heterogeneity of reported methodologic details, sample characteristics, validation strategies, and performance measures.…”

Amygdala Functional Connectivity Predicts Pharmacotherapy Outcome in Pediatric Bipolar Disorder

“…Functional connectivity fMRI techniques (Friston, 2002) have emerged as useful and informative methods to test hypotheses about the integration of brain regions implicated as dysfunctional in psychiatric disorders. Despite promising functional connectivity results in adult bipolar disorder (e.g., Anand et al, 2009;Arribas et al, 2010;Calhoun et al, 2008b;Chepenik et al, 2010;Foland et al, 2008;Pompei et al, 2011;Wang et al, 2009), to the authors' knowledge, only two studies of functional connectivity in children with PBD have been published, and both utilized a seed voxel technique. In one study, amygdala connectivity with emotional face processing regions was attenuated during a facial emotion-rating task (Rich et al, 2008).…”

Pixel Classification Using General Adaptive Neighborhood-Based Features

“…In this case, a ROC analysis of each "binarized" problem (i.e. one class vs. the others) has been made, as it is done in [16]. The ROC space for each problem is presented in Fig.…”

“…All these descriptors have been classified using a feedforward Artificial Neural Network [15] with one hidden layer, and softmax activation function in the hidden and output layers (to guarantee that outputs satisfy the probability constraints) [16].…”