Search

IntroductionResting state functional MRI (RS-fMRI) is currently used in numerous clinical and research settings. The localization of resting state networks (RSNs) has been utilized in applications ranging from group analysis of neurodegenerative diseases to individual network mapping for pre-surgical planning of tumor resections. Reproducibility of these results has been shown to require a substantial amount of high-quality data, which is not often available in clinical or research settings.MethodsIn this work, we report voxelwise mapping of a standard set of RSNs using a novel deep 3D convolutional neural network (3DCNN). The 3DCNN was trained on publicly available functional MRI data acquired in n = 2010 healthy participants. After training, maps that represent the probability of a voxel belonging to a particular RSN were generated for each participant, and then used to calculate mean and standard deviation (STD) probability maps, which are made publicly available. Further, we compared our results to previously published resting state and task-based functional mappings.ResultsOur results indicate this method can be applied in individual subjects and is highly resistant to both noisy data and fewer RS-fMRI time points than are typically acquired. Further, our results show core regions within each network that exhibit high average probability and low STD.DiscussionThe 3DCNN algorithm can generate individual RSN localization maps, which are necessary for clinical applications. The similarity between 3DCNN mapping results and task-based fMRI responses supports the association of specific functional tasks with RSNs.

Background and objectiveBlood oxygenation level-dependent (BOLD) functional MRI (fMRI) has been extensively used as a marker of brain dysfunction and subsequent recovery following stroke. However, growing evidence suggests that straightforward interpretation of BOLD fMRI changes with aging and disease is challenging. In this study, we investigated the effect of calibrating task fMRI data by applying a hemodynamic calibration method using the resting-state fluctuation amplitude (RSFA). Task fMRI responses were obtained during a covert verbal fluency task in a group of early stage stroke patients and matched healthy normal controls.MethodsFifteen acute left hemisphere stroke patients (less than 7 days from stroke; aged 44–84 years, average ~64 years) and 21 healthy controls (aged 55–77 years, average ~61 years) were prospectively studied. All subjects completed a 3-min covert verbal fluency task, and a 10-min eyes-closed resting-state fMRI scan, from which the calibration factor (RSFA) was computed. A behavioral measure on the verbal fluency task was also collected outside the scanner. Whole brain activation volumes and region-of-interest (ROI)-wise percent signal change and activation volumes before and after calibration were computed.ResultsBetween-group differences in whole brain activation volumes, although statistically significant before calibration failed to be significant after calibration. There were significant within-group differences before and after calibration with RSFA. Statistically significant between-group differences on ROI-wise measures before calibration also significantly reduced after calibration. Exploratory brain-behavior correlations revealed a similar pattern: significant correlations before calibration failed to survive after calibration.Discussion and conclusionBOLD fMRI changes with aging and disease is confounded by changes in neurofunctional coupling leading to challenges in the straightforward interpretation of task fMRI results. Application of the hemodynamic calibration using the RSFA technique in the current study appeared to mitigate any differences between stroke and age-matched healthy controls. Our study indicates that estimating neural activity after applying hemodynamic scaling is important for studies of aging and for studies tracking post-stroke changes. We recommend that further investigation of hemodynamic calibration with RSFA in healthy subjects and in stroke in larger samples is necessary.