Abstract
Mapping the organization of the visual system has enabled considerable progress in understanding different stages of visual processing. This mapping is typically achieved with fMRI-based retinotopy, which requires large amounts of data and central fixation. This can be problematic in populations such as infants, who cannot be instructed to fixate and who tend to produce only ~5 minutes of usable data per fMRI session. We were recently able to perform retinotopy in infants 5–23 months by modifying standard paradigms to reduce the impact of eye movements. However, this was still quite time-consuming, making it difficult to collect enough data in every participant or to perform additional experiments testing the response properties of the regions of interest identified through retinotopy. Here, we evaluate the feasibility of using movie fMRI data (collected for other purposes in a subset of sessions) to predict the retinotopic organization of infant visual cortex, eliminating the need for a separate retinotopy paradigm. We utilized two approaches with different strengths. First, independent components analysis (ICA) was used to extract components of the movie data that reflect visual evoked signal. With reference to the ground truth from these infants (obtained using retinotopy), ICA discovered eccentricity (foveal-peripheral) maps but rarely found phase (areal boundary) maps. Second, shared response modeling (SRM) was used to transform retinotopic maps between participants using functional alignment. SRM accurately predicted phase maps, but performed worse on eccentricity maps. By combining ICA and SRM, it may be possible to predict the retinotopic organization of infant visual cortex using as little as three minutes of movie data. Beyond the practical uses of obtaining retinotopic maps without specialized tasks, this work also shows the potential of functional alignment with infants and reveals that infant brain activity during passive movie viewing recapitulates the organization of the visual system.