Abstract
Purpose :
Functional Magnetic Resonance Imaging (fMRI) measures neural activity indirectly by monitoring local changes in blood oxygenation and blood flow. Altered patterns of cerebral blood flow that occur, for instance, in stroke patients, may affect the analysis and interpretation of fMRI data. Here we investigate relations between experimental design and altered hemodynamic responses using computational approaches modeled on the early visual system.
Methods :
Population receptive field mapping is used to localize cortical representation of different regions of the visual field. We used computational modeling to determine how the accuracy and precision of population receptive field mapping is affected by altered hemodynamic responses in two conditions: 1) stimuli presented sweeping across the visual field and 2) stimuli presented randomly across the visual field. Hemodynamic response functions for both analyses were modeled using a 2-gamma function with a time-to-peak varying between 0 and 12 repetition times (TRs) from stimulus onset.
Results :
We found that the analysis was more robust to an altered hemodynamic response when the stimuli were presented in a random sequence. Sequential stimulus presentation led to systematic errors, that were not present with the random presentation sequence, leading to an increase in Type I errors (false positive). For both stimulus presentation methods, the sensitivity and specificity of the analysis greatly deteriorated if the time-to-peak was delayed by greater than +/- 1 TR.
Conclusions :
These results highlight the importance of tailoring the experimental design to altered hemodynamic responses and point to the need for subject- and voxel-specific hemodynamic response functions when analyzing fMRI data in patient populations with neurovascular abnormalities. Without these modifications, researchers run the risk of making incorrect conclusions about retinotopic reorganization.
This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.