In diffusion magnetic resonance imaging, or dMRI, an incomplete field of view (FOV) means that part of the brain is missing from the scan. This can seriously affect analyses of white matter connectivity, including tractography, which maps the pathways of nerve fiber bundles across the brain. Although previous studies have used deep generative models to estimate or “impute” the missing regions, it is still unclear how to best use additional information from paired multi-modality data, such as combining dMRI with structural T1-weighted (T1w) MRI, to improve the quality of reconstruction and support downstream analyses.

To address this, the researchers developed a new framework that imputes missing dMRI regions by integrating diffusion features from the acquired portion of the scan with information about the complete brain anatomical structure derived from paired imaging data. The idea is that using anatomical guidance from other modalities can improve how the missing diffusion signals are reconstructed. They tested the framework on two cohorts from different sites, including a total of 96 participants, and compared it with a baseline method that treated T1w and dMRI information equally without specifically leveraging their complementary roles.

The proposed framework significantly improved imputation quality, as measured by the angular correlation coefficient, and improved the accuracy of downstream tractography, as measured by the Dice score. These results suggest that carefully integrating paired multi-modality data leads to more accurate reconstruction of incomplete dMRI scans. By improving whole-brain tractography, this approach may reduce uncertainty in analyses of white matter bundles, particularly those relevant to neurodegenerative diseases.

Fig. 1.

Visualization (left) and histogram (right) of 103 real cases of dMRI scans with incomplete FOV that failed quality assurance. In the left figure, horizontal regions indicate the distribution of the incomplete part of FOV with an estimated position of a brain mask. The total cutoff distance from the incomplete FOV to the top of the brain is estimated using a corresponding and registered T1w image. Its histogram is presented in the right figure.

Brain network models can be enhanced by combining structural data from white matter (via diffusion MRI) with functional data from gray matter (via resting-state BOLD signals). Diffusion MRI shows how white matter tracts connect gray matter regions, and we developed a method to assess how each white matter voxel contributes to brain function by integrating these connections. This creates detailed maps of white matter engagement, identifying areas critical for brain communication. These maps are highly consistent across individuals and highlight how gray matter activity depends on white matter organization. We also observed that white matter engagement varies over time and differs between genders, indicating its potential as a tool for understanding brain function and identifying neurological disorders.��

Fig. 1. Reconstruction and Analysis of Dynamic White Matter Engagement.��

1a: Workflow of WM engagement reconstruction. The brain is parcellated into 90 GM regions using the AAL atlas. Following the left path, fiber tracking is performed considering each WM voxel x as a seed point. This tracking process retrieves all fibers that transverse x and connects to different pairs of GM regions. This created a 90×90 SC matrix, in which each element S(i,j) indicates the number of fibers connecting regions iand j and pass through x. Following the right path, FC is calculated based on the fMRI time series of the 90 regions, producing an FC matrix based on Pearson’s correlation coefficients. Using graph theory, a 90×90 EBC matrix is calculated where each element E(i,j) represents the functional importance of the connection between i and j. Voxel engagement (engmt) of x is then computed as a weighted average of the EBC elements, with the SC values serving as weights.��

1b: Dynamic Engagement Analysis. To capture the temporal dynamics of voxel engagement, the time course is divided into overlapping windows of 36 seconds. The standard deviation (STD) of voxel engagement across these windows quantifies temporal variability. The clustering of concatenated engagement maps from all windows and subjects via k-means results in 31 mode maps, each representing a cluster center. The silhouette coefficient method determines the optimal number of clusters. Mode occurrence is measured by the frequency of each mode map across the time course for each subject, reflecting the dynamic engagement of voxels in the network.��

This study investigates the variability in diffusion tensor imaging (DTI) data, particularly when data are merged from multiple sites, which is crucial for large-scale analyses. DTI measures can be affected by spatially varying and correlated noise, making it important to understand how different factors—like physiology, subject behavior, and scanner interaction—impact the reliability of the results. The researchers focused on characterizing the sources of variance in DTI metrics in different brain regions to improve the accuracy of future analyses.

Using data from 1,035 subjects, aged 22 to 103, in the Baltimore Longitudinal Study of Aging, the study analyzed how DTI variance changes over time and across multiple factors. Each subject had up to 12 longitudinal DTI scans, and the authors examined how factors such as age, scan interval, motion, sex, and session order affected DTI variance in different regions of the brain. They found that the impact of these factors was complex and varied across regions. For example, the time between scans was associated with increased variance in some areas (like the cuneus and occipital gyrus) but decreased variance in others (such as the caudate nucleus). Additionally, males showed higher variability in specific regions, and head motion had a mixed impact on DTI variance across different regions.

The findings highlight the need for researchers to consider the variability in DTI metrics when analyzing data and planning studies. By accounting for these regional variations in variability, researchers can improve the accuracy and reliability of DTI-based analyses, especially in large, multi-site studies. This work also emphasizes the importance of including variance estimates in data sharing to enhance the quality of future research.

This study focuses on improving brain diffusion magnetic resonance imaging (dMRI) when the field of view (FOV) is incomplete, a common issue that can hinder analysis of brain tissue microstructure and connectivity. Rather than discarding corrupted dMRI data due to missing slices, the authors propose a deep generative model to impute, or estimate, the absent brain regions directly from existing scans. This allows for more complete whole-brain tractography, which is essential for mapping brain connections.

The framework leverages both the diffusion characteristics from diffusion-weighted images (DWIs) and anatomical details from structural images to accurately reconstruct missing slices in the dMRI data. Testing on the Wisconsin Registry for Alzheimer’s Prevention (WRAP) and National Alzheimer’s Coordinating Center (NACC) datasets showed that the imputed slices significantly improved the accuracy of whole-brain tractography. Specifically, the model enhanced the precision of tractography for 72 brain tracts, resulting in improved analysis of brain connectivity, which is crucial for studying diseases like Alzheimer’s.

Overall, this approach provides an effective alternative to discarding incomplete dMRI data, enabling more reliable analyses by reconstructing missing information and extending the FOV. The method helps reduce uncertainty in tractography, offering valuable insights into brain structure, especially in clinical studies related to neurodegenerative conditions.

Diffusion tensor imaging (DTI) is a magnetic resonance imaging technique that provides unique insights into white matter microstructure in the brain. However, it is susceptible to confounding effects introduced by scanner or acquisition differences. ComBat is a leading approach for addressing these site biases. Despite its frequent use for harmonization, ComBat’s robustness towards site dissimilarities and overall cohort size has not yet been evaluated in the context of DTI.

To address this, we matched 358 participants from two sites to create a “silver standard” cohort for multi-site harmonization. We harmonized mean fractional anisotropy (FA) and mean diffusivity (MD) calculated from participant DTI data for regions of interest defined by the JHU EVE-Type III atlas. To quantify the reliability of ComBat, we performed bootstrapping over 10 iterations at 19 levels of total sample size, 10 levels of sample size imbalance between sites, and 6 levels of mean age difference between sites. We measured three key metrics: (i) β_AGE, the linear regression coefficient of the relationship between FA and age; (ii) γ_sf, the ComBat-estimated site-shift; and (iii) δ_sf, the ComBat-estimated site-scaling. We evaluated the reliability of ComBat by calculating the root mean squared error (RMSE) in these metrics and examined the correlation between the reliability of ComBat and the violation of model assumptions.

Our results indicate that ComBat performs reliably for β_AGE when the total sample size is greater than 162 and the mean age difference between sites is less than 4 years. The assumptions of the ComBat model regarding the normality of residual distributions are not violated as the model becomes unstable.

In conclusion, before harmonizing DTI data with ComBat, it is crucial to examine the input cohort for size and covariate distributions at each site. Direct assessment of residual distributions is less informative on stability than bootstrap analysis. We advise caution when using ComBat in situations that do not conform to the identified thresholds.

Connectivity matrices derived from diffusion MRI (dMRI) offer an interpretable and generalizable way to understand the human brain connectome. However, dMRI is subject to inter-site and between-scanner variations, which can hinder cross-dataset analysis and affect the robustness and reproducibility of results. To evaluate different harmonization approaches on connectivity matrices, we compared graph measures derived from these matrices before and after applying three harmonization techniques: mean shift, ComBat, and CycleGAN.

The sample consisted of 168 age-matched, sex-matched normal subjects from two studies: the ������ý���� Memory and Aging Project (VMAP) and the Biomarkers of Cognitive Decline Among Normal Individuals (BIOCARD). First, we plotted the graph measures and used the coefficient of variation (CoV) and the Mann-Whitney U test to assess the effectiveness of each method in removing site effects on the matrices and the derived graph measures.

ComBat effectively eliminated site effects for global efficiency and modularity, outperforming the other two methods. However, all methods exhibited poor performance when harmonizing average betweenness centrality. Second, we examined whether the harmonization methods preserved correlations between age and graph measures. All methods, except for CycleGAN in one direction, improved the correlations between age and global efficiency and between age and modularity, changing them from insignificant to significant with p-values less than 0.05.

These findings suggest that while ComBat is particularly effective for certain graph measures, challenges remain in harmonizing other measures like betweenness centrality. Nonetheless, the ability of these methods to enhance the significance of age-related correlations highlights their potential in improving the robustness of dMRI connectivity analyses across different datasets.

Data harmonization is essential for eliminating confounding effects in multi-site diffusion image analysis. One such harmonization method, LinearRISH, scales rotationally invariant spherical harmonic (RISH) features from a “target” site to match those from a “reference” site, aiming to reduce scanner-related confounding effects. However, the designation of reference and target sites is not arbitrary, and this choice can bias the resulting diffusion metrics such as fractional anisotropy and mean diffusivity. This study introduces MidRISH, a method that projects both sites to a mid-space, thereby avoiding the bias introduced by reference site selection. The MidRISH method was validated through two experiments: harmonizing scanner differences in 37 matched patients free of cognitive impairment, and harmonizing acquisition and study differences in 117 matched patients free of cognitive impairment. The results demonstrate that MidRISH reduces the bias associated with reference site selection while maintaining the harmonization efficacy of LinearRISH. Users should be cautious when using LinearRISH harmonization, as choosing a reference site impacts the effect size of diffusion metrics. The proposed MidRISH method eliminates the bias-inducing site selection step, offering a more robust approach to harmonization.

Diffusion magnetic resonance imaging (dMRI) enables the assessment of subvoxel brain microstructure by extracting biomarkers like fractional anisotropy and reconstructing white matter fiber trajectories to reveal brain connectivity. However, accurate analysis is challenging at the interface between cerebrospinal fluid (CSF) and white matter, where the MRI signal is a mix of contributions from both CSF and white matter partial volumes. This mixture introduces significant bias in estimating diffusion properties, thus limiting the clinical utility of diffusion-weighted imaging (DWI).

Furthermore, existing mathematical models often struggle with single-shell acquisitions, which are common in clinical settings. Without proper regularization, direct model fitting is impractical.

To address these challenges, we propose a novel voxel-based deep learning method for mapping and correcting free-water partial volume contamination in DWI. This approach leverages data-driven techniques to reliably estimate plausible free-water volumes across various diffusion MRI acquisition schemes, including single-shell acquisitions. Our evaluation shows that this methodology consistently produces more accurate and reliable results than previous methods.

By effectively mitigating the impact of free water partial volume effects, our approach enhances the accuracy and reliability of DWI analysis for single-shell dMRI. This advancement expands the potential applications of dMRI in assessing brain microstructure and connectivity, making it more useful in clinical settings.

Subject head motion during the acquisition of diffusion-weighted imaging (DWI) of the brain can induce artifacts and negatively affect image quality. Understanding the frequency and extent of motion can highlight the most critical aspects of motion correction needed. This study investigates the extent of translational and rotational movements among participants and examines how these motions change during scan acquisition.

The analysis includes 5,380 DWI scans from 1,034 participants. The study measures rotations and translations in the sagittal, coronal, and transverse planes required to align each volume to the first and previous volumes, as well as overall displacement. The different types of motion are compared with each other and over time.

Results show that the largest rotation occurs around the right-left axis (median 0.378°/min, range 0.000 – 11.466°) and the largest translation occurs along the anterior-posterior axis (median 1.867 mm/min, range 0.000 – 10.944 mm). Additionally, spikes in movement are frequently observed at the beginning of the scan, particularly in anterior-posterior translation.

These findings indicate that all scans are affected by subtle head motion, which may impact subsequent image analysis. Understanding these motion patterns can help in developing better motion correction techniques to improve the quality of DWI scans.

Gradient nonlinearities in magnetic resonance imaging (MRI) not only cause spatial distortions but also create discrepancies between the intended and acquired diffusion sensitization in diffusion-weighted (DW) MRI. With advances in scanner performance, correcting these gradient nonlinearities has become increasingly important. Common methods for estimating gradient nonlinear fields rely on phantom calibration field maps, which are often impractical, especially for retrospective data.

This study presents a new approach to estimate the complete gradient nonlinear field, denoted as L(r), by formulating a quadratic minimization problem. This method begins with the corrupted diffusion signal and estimates L(r) under two scenarios: (1) when the true diffusion tensor is known, and (2) when the true diffusion tensor is unknown and must be estimated. The validity of this mathematical approach is demonstrated both theoretically and through tensor simulation.

The estimated field is evaluated using diffusion tensor metrics: mean diffusivity (MD), fractional anisotropy (FA), and principal eigenvector (V1). Simulations with 300 diffusion tensors indicate that the formulation is stable and not ill-posed. When the true diffusion tensor is known, the change in the determinant of the estimated L(r) field relative to the true field is near zero, and the median difference in corrected diffusion metrics compared to true values is also near zero. The results show that the accuracy of L(r) estimation depends on the level of corruption in L(r).

This work introduces a novel mathematical method to estimate the gradient field without requiring additional calibration scans, offering a significant advancement for correcting gradient nonlinearities in DW MRI.