Abstract
While achieving higher spatial resolution is an important goal in fMRI studies, it also results in reduced signal-to-noise ratio (SNR). NORDIC is a noise reduction method based on patchwise principal component analysis (PCA) and increases SNR by removing thermal noise components. Critically, the NORDIC approach should reduce spatial smoothing effects that were frequently described with other noise-reduction methods. Here we investigate NORDIC's effects on population receptive field (pRF) mapping, particularly on pRF size estimation. We acquired fMRI data from three healthy participants using a SIEMENS PrismaFit 3T scanner. A bar aperture moving in eight directions was presented, revealing reversing checkerboards over a 9° radius field of view. NORDIC denoising was applied after standard scanner reconstruction. Both original and denoised data underwent minimal preprocessing using fMRIPrep, followed by pRF mapping analysis using containerized solutions prfprepare and prfanalyze-vista. Our findings show that NORDIC increases variance-explained values without inflicting differences in pRF position (eccentricity and polar angle), while pRF size estimations increase considerably (median 22% increase in pRF size with NORDIC). Cohen’s d effect sizes show a small effect on pRF size and a large effect on variance explained. It might be concluded that this increase in pRF sizes is caused by an increase in SNR due to NORDIC. Our simulations, however, clearly show that increased SNR yields reduced pRF sizes. We also studied the effects of spatial smoothing in the pRF sizes estimated and found that spatial smoothing leads to increases in pRF sizes. Taken together, these increases in pRF size seem not to be directly linked to spatial image smoothness, but may arise from the retinotopic organization of neighboring voxels. Although NORDIC only marginally increases image smoothness, its impact on pRF size estimations necessitates careful interpretation. Our results underscore the importance of considering NORDIC's influence on pRF size in fMRI preprocessing.