AGFS-Tractometry: A Novel Atlas-Guided Fine-Scale Tractometry Approach for Enhanced Along-Tract Group Statistical Comparison Using Diffusion MRI Tractography

TL;DR

AGFS-Tractometry addresses the challenge of detecting localized white-matter differences along tracts by combining atlas-guided fine-scale parcellation with a nonparametric, permutation-based statistical framework. It builds a population-wide tract profiling template from the ORG atlas to create fine-scale along-tract parcels and a spatial neighborhood among parcels. Subject-specific tract profiles are extracted by mapping data to the atlas and measuring parcel-wise diffusion metrics with distance-weighted averaging, followed by a CP-based suprathreshold community approach for group comparisons. Across synthetic and real datasets, AGFS-Tractometry shows enhanced sensitivity and specificity relative to AFQ and BUAN, identifying more localized, literature-consistent differences, with code available at GitHub.

Abstract

Diffusion MRI (dMRI) tractography is currently the only method for in vivo mapping of the brain's white matter (WM) connections. Tractometry is an advanced tractography analysis technique for along-tract profiling to investigate the morphology and microstructural properties along the fiber tracts. Tractometry has become an essential tool for studying local along-tract differences between different populations (e.g., health vs disease). In this study, we propose a novel atlas-guided fine-scale tractometry method, namely AGFS-Tractometry, that leverages tract spatial information and permutation testing to enhance the along-tract statistical analysis between populations. There are two major contributions in AGFS-Tractometry. First, we create a novel atlas-guided tract profiling template that enables consistent, fine-scale, along-tract parcellation of subject-specific fiber tracts. Second, we propose a novel nonparametric permutation testing group comparison method to enable simultaneous analysis across all along-tract parcels while correcting for multiple comparisons. We perform experimental evaluations on synthetic datasets with known group differences and in vivo real data. We compare AGFS-Tractometry with two state-of-the-art tractometry methods, including Automated Fiber-tract Quantification (AFQ) and BUndle ANalytics (BUAN). Our results show that the proposed AGFS-Tractometry obtains enhanced sensitivity and specificity in detecting local WM differences. In the real data analysis experiments, AGFS-Tractometry can identify more regions with significant differences, which are anatomically consistent with the existing literature. Overall, these demonstrate the ability of AGFS-Tractometry to detect subtle or spatially localized WM group-level differences. The created tract profiling template and related code are available at: https://github.com/ZhengRuixi/AGFS-Tractometry.git.

Figures (7)

• Figure 1: Different tract profiling strategies in tractometry, including AFQ, BUAN, and AGFS-Tractometry. (a) Illustration of the CST with color-coded clusters indicating anatomical subdivisions. (b) Tract profiling along the entire fiber tract using the traditional AFQ and BUAN methods, with results shown as color-coded parcels. (c) Our proposed AGFS-Tractometry method can achieve finer subdivisions, distinguishing different motor pathways for the leg, trunk, hand, and face.
• Figure 2: Method Overview. (a) Step 1: Tract Profiling Template Creation. Along-tract parcellation is guided by centerlines computed from the ORG-atlas, followed by neighborhood construction to generate a parcel neighborhood matrix; (b) Step 2: Subject-Specific Profile Extraction. Subject-specific tract parcellation is performed, and quantitative measures are computed for each parcel to generate individual tract profiles; (c) Step 3: Groupwise Statistical Comparison. Group-level comparisons are conducted at the parcel level. Statistically significant parcels are identified and grouped into communities based on the parcel neighborhood matrix. A permutation test determines the significance of each community according to size thresholds, resulting in the final detection of group differences.
• Figure 3: Illustration of the parcel neighborhood construction process: (a) Intra- and inter-cluster parcel neighbor relationship, (b) parcel neighborhood matrix.
• Figure 4: Method for Community Detection and Statistical Validation. The upper panel illustrates the community detection process. A parcel neighborhood matrix and a parcel-level statistical testing matrix are constructed and combined via dot multiplication to form a parcel graph. The clique percolation (CP) algorithm is then applied to detect parcel communities based on the graph structure. Communities are sorted by size in descending order. The lower panel depicts the suprathreshold community detection. A null distribution of maximum community sizes is generated by repeatedly permuting subject labels and applying community detection on the permuted datasets. Communities obtained from the true data are considered significant if their size exceeds the threshold derived from the null distribution (e.g., the 95th percentile), ensuring control of false positive rates.
• Figure 5: Visual comparison across the compared methods in the synthetic datasets. Three examples out of a total of 27 combinations of different ROIs and radii across the AF, CST, and CC2 tracts are shown. The first column shows the ground truth ROIs with artificially introduced differences, while the subsequent columns show the detected regions by AFQ, BUAN, and AFGS-tractometry. For each ROI, the ACC of the detected significant region is displayed.