Home Research Publications People Related Sites
   Multimodality Brain Mapping Group/Digital Image Processing Laboratory
Research Description

Motion correction in fMRI using the map-slice-to-volume (MSV), 2D-to-3D, approach - Our fMRI processing is based on a capability of mapping individual EPI slices in time series data to an anatomically correct MRI volume data, map-slice-to-volume (MSV), to correctly register voxel intensities acquired from a human subject exhibiting head motion during the fMRI data collection (Kim 1999, Kim 2008). For language studies, non-linear field artifacts are particularly evident in EPI data that include temporal lobes and cause difficulty in motion correction of language activation data. The confounding effect from the head motion requires robust solution for the local field variation and spin saturation (i.e., spin history) effect.

Dynamically changing B0 term in motion estimation - For the effect of changes in B0 inhomogeneity due to head motion, we construct the transform matrix of the data acquisition (forward) model that includes the field inhomogeneity term updated with motion parameters. Our distortion model incorporates the field map estimation that changes with head motion (Yeo 2008, Takeda 2012, Takeda 2013).

Modeling of head motion by joint mapping of slice to volume - By the joint MSV (JMSV) method, motion parameters for slices can be jointly estimated by maximizing the objective function, in which smoothness through a roughness penalty term is implemented (Park 2004).

Language activation - overt speech fMRI - A cortical surface activation map of an overt speech processed using our MSV motion correction. The EPI time series was acquired without physical head-restrain. It is particularly important to correct individual slice in fMRI time series of verbalized speech since the head motion associated with articulation is unavoidable and the magnitude is much larger than the range of motion that can be handled by volume registration algorithms or simple smoothing or interpolation.

Evaluation of MSV using mathematical phantom time series against known truths - The availability of truth to compare the accuracy of the detection algorithms is important for evaluation of activation accuracy. A comprehensive synthetic time series data was constructed with (1) known motion parameters for each slice, (2) geometric distortion exerted on each slice computed with a dynamic field inhomogeneity map and (3) image intensity changes from the spin saturation effect as a result of out-of-plane head motion. The simulated fMRI time series are used for analyses with motion correction, for the evaluation of motion recovery and activation localization (Kim 2008).

Susceptibility and field map estimations - Susceptibility map was estimated using several 3D image restoration algorithms for a simulated noisy 3D field map modeling, thresholded inverse SVC filter estimate (NRMSE=26.83%.), Wiener filter estimate and quadratic penalized least squares estimate (Yeo 2008).

Intensity correction of spin saturation (i.e., spin history) effect in fMRI time series - Spin saturation effect causes intensity attenuation in the potion of the incorrectly selected regions at the previous slice selection as exhibited as dark bands in an EPI image shown. The accurate account and correction of the spin saturation effect require knowledge of slice positions at the time of the slice excitation, and time (t) of RF excitation pulse to identify voxels with dt < TR. After the initial development using a single T1 presented earlier (Bhagalia & Kim, 2007, 2008), effective-T1 correction with more complete account of the partial volume effect, which involves multiple T1s for different tissue types in the affected EPI voxels, was computed.

Multi-modality, multi-scale and functional brain maps - Electrocortical stimulation (ECS) mapping, a clinical open cranium procedure, is performed to identify the functionally important sites in pre-surgical evaluation of epilepsy patients. Multiscale measurements of neural processes from fMRI, DTI, optical imaging, PET, can be combined through multimodality image registration on anatomical MRI and CT scans for detailed maps of language, motor and sensory sites and compared in the same coordinate system (Yeo et al., 2008).

Brain Functional Mapping

Header image