A family of locally constrained CCA models for detecting activation patterns in fMRI

Abstract

Canonical correlation analysis (CCA) has been used in functional Magnetic Resonance Imaging (fMRI) for improved detection of activation by incorporating time series from multiple voxels in a local neighborhood. To improve the specificity of local CCA methods, spatial constraints were previously proposed. In this study, constraints are generalized by introducing a family model of spatial constraints for CCA to further increase both sensitivity and specificity in fMRI activation detection. The proposed locally-constrained CCA (cCCA) model is formulated in terms of a multivariate constrained optimization problem and solved efficiently with numerical optimization techniques. To evaluate the performance of this cCCA model, simulated data are generated with a Signal-To-Noise Ratio of 0.25, which is realistic to the noise level contained in episodic memory fMRI data. Receiver operating characteristic (ROC) methods are used to compare the performance of different models. The cCCA model with optimum parameters (called optimum-cCCA) obtains the largest area under the ROC curve. Furthermore, a novel validation method is proposed to validate the selected optimum-cCCA parameters based on ROC from simulated data and real fMRI data. Results for optimum-cCCA are then compared with conventional fMRI analysis methods using data from an episodic memory task. Wavelet-resampled resting-state data are used to obtain the null distribution of activation. For simulated data, accuracy in detecting activation increases for the optimum-cCCA model by about 43% as compared to the single voxel analysis with comparable Gaussian smoothing. Results from the real fMRI data set indicate a significant increase in activation detection, particularly in hippocampus, para-hippocampal area and nearby medial temporal lobe regions with the proposed method.

The relation between statistical power and inference in fMRI

“…[ 80 , 81 ]) or structural equation modeling (SEM; [ 82 ]) Such approaches typically increase power by reducing the number of variables tested (e.g., 40 ICA components rather than 200,000 voxels), affording the use of more liberal statistical thresholds. Canonical Correlation Analyses (CCA) is another powerful multivariate tool to address the correspondence of two datasets, see for instance refs [ 83 , 84 ]. Graph or network theory forms another informative and elegant alternative to mass univariate analyses [ 85 – 88 ].…”

A robust deep neural network for denoising task-based fMRI data: An application to working memory and episodic memory

“…Since fMRI activation maps are usually thresholded with a strict p value to eliminate false positive voxels, the AUC value in the low FPR range 0-0.1 is more meaningful than the AUC value in the full FPR range 0-1 for controlling false positives. Consistent with our previous studies (Yang et al, 2018;Zhuang et al, 2017) having AUC value for fMRI data in the range of 0.05-0.07, the signal fraction f was adjusted if the simulated data with initial signal fraction value had an AUC value not in this range. rnd(3,1) is a 3 x 1 vector generated from standard normal distribution to introduce the variability between voxels within the same active region.…”

A Projection CCA Method for Effective fMRI Data Analysis

“…DCT coefficients are calculated for several voxels within a neighborhood of the ROI. The average spectrum obtained from the DCT coefficients in the ROI is used to determine high energy frequency components K. The desired temporal structure is enforced onto the subspace by selecting those DCT basis functions, b k , from (17), which represent 90% of the fMRI time course energies. The matrix B consisting of K−DCT basis vectors selected using indices of coefficients obtained using (18) is embedded in the CCA objective function for estimating canonical vectors.…”

“…Maximizing the variance leads to principal component analysis (PCA) [4], the independence of components leads to the spatial and temporal independent component analysis (sICA and tICA) [5], the sparsity of components leads to dictionary learning [6]. Maximum correlation constraints, which leads to CCA, has also been successfully employed to analyse fMRI data [7]- [17]. Generally, CCA analyses fMRI data from a temporal or spatial perspective.…”