A novel framework for groupwise registration of fMRI images based on common functional networks

Abstract

Accurate registration plays a critical role in group-wise functional Magnetic Resonance Imaging (fMRI) image analysis, as spatial correspondence among different brain images is a prerequisite for inferring meaningful patterns. However, the problem is challenging and remains open, and more effort should be made to advance the state-of-the-art image registration methods for fMRI images. Inspired by the observation that common functional networks can be reconstructed from fMRI image across individuals, we propose a novel computational framework for simultaneous groupwise fMRI image registration by utilizing those common functional networks as references for spatial alignments. In this framework, firstly, individualized functional networks in each subject are inferred using Independent Component Analysis (ICA); secondly, congealing groupwise registration that takes entropy of stacked independent components (ICs) from all the subjects as objective function is applied to register individual functional maps for maximal matching. The proposed framework is evaluated by and applied to an Alzheimer’s Disease (AD) fMRI dataset and shows reasonably good results.

Multimodal Cross Enhanced Fusion Network for Diagnosis of Alzheimer’s Disease and Subjective Memory Complaints

“…Many studies (3; 4; 5) have found that a range of neuroimaging biomarkers can be used to diagnosis both AD and SMC. Structural magnetic resonance imaging (sMRI) is a noninvasive method that can detect high-resolution structural changes in the brain caused by atrophy, such as changes in thickness, volume, shape, and texture (8). Similarly, 18-Fluoro-Deoxy-Glucose Positron Emission computed Tomography (FDG-PET) uses radioactive tracers to track cerebral metabolic rate of glucose reflecting hemodynamic and detect changes in brain function (7).…”

“…Recently, Deep learning methods have demonstrated excellent performance on high-dimensional complex data, and have been widely applied to the diagnosis of AD (8). These methods typically involve three basic steps: 1) identifying regions of interest (ROI), 2) extracting features, and 3) constructing classification models.…”

“…These methods typically involve three basic steps: 1) identifying regions of interest (ROI), 2) extracting features, and 3) constructing classification models. In the ROI identifying step , brain imaging data is often automatically divided into multiple brain regions using the generic template AAL (Automated Anatomical Labeling) (8), or determined manually based on relevant brain regions in anatomy (9). For feature extraction, existing networks can be classified into 2D-based, 3D-based, and transformer-based networks according to their architecture.…”

A Coarse-to-Fine Deformable Transformation Framework for Unsupervised Multi-Contrast MR Image Registration with Dual Consistency Constraint

“…ulti-modal medical imaging plays an important role in many clinical applications [1][2][3][4][5][6][7][8][9][10][11], such as image-guided intervention, disease diagnosis, and treatment planning. Among them, multi-contrast magnetic resonance (MR) imaging is one of the most prevalent techniques utilized in brain imaging as different MR imaging sequences can highlight different regions of interest.…”

rfDemons: Resting fMRI-Based Cortical Surface Registration Using the BrainSync Transform

“…Here we explore an alternative to this approach in which we use a topologically-constrained nonrigid deformation of the cortical surfaces to perform inter-subject registration. An alternative recent method also used rfMRI data for this purpose [7]. In that case, z-score maps derived from ICA analysis of the functional activations were used to perform group-wise cortical surface registration.…”