Decoding the infant mind: Multivariate pattern analysis (MVPA) using fNIRS

Abstract

The MRI environment restricts the types of populations and tasks that can be studied by cognitive neuroscientists (e.g., young infants, face-to-face communication). FNIRS is a neuroimaging modality that records the same physiological signal as fMRI but without the constraints of MRI, and with better spatial localization than EEG. However, research in the fNIRS community largely lacks the analytic sophistication of analogous fMRI work, restricting the application of this imaging technology. The current paper presents a method of multivariate pattern analysis for fNIRS that allows the authors to decode the infant mind (a key fNIRS population). Specifically, multivariate pattern analysis (MVPA) employs a correlation-based decoding method where a group model is constructed for all infants except one; both average patterns (i.e., infant-level) and single trial patterns (i.e., trial-level) of activation are decoded. Between subjects decoding is a particularly difficult task, because each infant has their own somewhat idiosyncratic patterns of neural activation. The fact that our method succeeds at across-subject decoding demonstrates the presence of group-level multi-channel regularities across infants. The code for implementing these analyses has been made readily available online to facilitate the quick adoption of this method to advance the methodological tools available to the fNIRS researcher.

Fig 1. Functional near-infrared spectroscopy (fNIRS) records cortical hemodynamic responses in populations that cannot comfortably be inside the MR scanner such as young infants.

Pairs of detectors and emitters form an fNIRS channel (from Gervain et al., 2011 with permission) which covers a localizable region of the cortex.

Fig 2. Illustration of the multivariate methods applied to fNIRS in this paper.

Fig 3. Depiction of the two datasets and the decoding results (infant-level and trial-level) for each.

Error bars depict the bootstrapped confidence intervals of the mean across infants.

Fig 4. Decoding accuracy of infant-level activation patterns by subset size for Datasets #1 (purple boxes) and #2 (blue boxes).

Far right, decoding using three most informative channels (most informative channels determined using subset size 2, Fig 3). Note: For the subset size of 10 channels, there is only one subset and so there is no range to estimate.

Fig 5. Accuracy for each of the 10 NIRS channels for Dataset #1 (left) and Dataset #2 (right) in different subset sizes (from 2 to 10 channels with each line labeled at the right with the subset size).

Fig 6. Comparison of the relative informativeness across channels from multivariate analysis (from dark to light, least to most informative respectively) and channels which exhibit a significant difference between the same two conditions in a univariate analysis.

Across both datasets, only a single channel that exhibits a significant univariate response is one of the most informative channels in the multivariate analyses. In Dataset #2, not a single channel was significant for our univariate analysis but we achieve significant infant-level decoding in the multivariate analysis.