Comparing within‐subject classification and regularization methods in fMRI for large and small sample sizes

Abstract

In recent years, a variety of multivariate classifier models have been applied to fMRI, with different modeling assumptions. When classifying high-dimensional fMRI data, we must also regularize to improve model stability, and the interactions between classifier and regularization techniques are still being investigated. Classifiers are usually compared on large, multisubject fMRI datasets. However, it is unclear how classifier/regularizer models perform for within-subject analyses, as a function of signal strength and sample size. We compare four standard classifiers: Linear and Quadratic Discriminants, Logistic Regression and Support Vector Machines. Classification was performed on data in the linear kernel (covariance) feature space, and classifiers are tuned with four commonly-used regularizers: Principal Component and Independent Component Analysis, and penalization of kernel features using L₁ and L₂ norms. We evaluated prediction accuracy (P) and spatial reproducibility (R) of all classifier/regularizer combinations on single-subject analyses, over a range of three different block task contrasts and sample sizes for a BOLD fMRI experiment. We show that the classifier model has a small impact on signal detection, compared to the choice of regularizer. PCA maximizes reproducibility and global SNR, whereas Lp -norms tend to maximize prediction. ICA produces low reproducibility, and prediction accuracy is classifier-dependent. However, trade-offs in (P,R) depend partly on the optimization criterion, and PCA-based models are able to explore the widest range of (P,R) values. These trends are consistent across task contrasts and data sizes (training samples range from 6 to 96 scans). In addition, the trends in classifier performance are consistent for ROI-based classifier analyses.

An empirical evaluation of multivariate lesion behaviour mapping using support vector regression

“…On the other hand, improvements beyond ~100 subjects were very small and appeared to approach a limiting value. This closely resembles findings on multivariate modelling of fMRI data (Churchill et al, 2014). The authors also observed that increases in sample size led to a plateau in regards to prediction accuracy already with small samples, while reproducibility of model parameters improved if already large samples were increased.…”

Assessing and tuning brain decoders: Cross-validation, caveats, and guidelines

“…One option is to average the predictions of a set of suitable models [44, chap. 35], [8,23,29] -see [18, chap. 8] for a description outside of neuroimaging.…”

Decoding the encoding of functional brain networks: An fMRI classification comparison of non-negative matrix factorization (NMF), independent component analysis (ICA), and sparse coding algorithms

“…Although the biological interpretations of these algorithms are not mutually exclusive to ICA, the success of a specific dictionary learning method may prioritize theories of encoding. Moreover, given the finding that the choice of regularizer may be more important than the choice of classifier ([19]), judicious selection a priori of a regularization method may allow us to better understand the network dynamics of cognitive processes.…”