Generalized q-Sampling Imaging (GQI)
GQI is a model-free approach that estimates water diffusion orientation distribution functions (dODFs) using an analytic transform of the diffusion signal (Yeh et al., 2010). Like the other methods, GQI produces a number of parametric microstructure maps such as generalized fractional anisotropy (GFA), quantitative anisotropy (QA), and isotropic component (ISO) (Yeh et al., 2013).
GQI in QSIRecon
GQI reconstruction is supported in QSIRecon using the DSI Studio package: see init_dsi_studio_recon_wf().
This is accessible in a reconstruction specification by using a node with action: reconstruction, software: DSI Studio, and parameters: method: gqi.
Also see qsirecon.interfaces.dsi_studio.DSIStudioGQIReconstruction.
GQI Foundational Papers
GQI Concept Introduction: Generalized q-Sampling Imaging (GQI) was introduced as an analytical, model-free diffusion MRI reconstruction method that bridges Q-ball imaging (single-shell HARDI) and diffusion spectrum imaging (grid sampling) (Yeh et al., 2010). It estimates the spin distribution function (SDF), an ODF-like representation of diffusion directions, from a variety of sampling schemes.
Anisotropy quantification (GFA): Generalized Fractional Anisotropy (GFA), originally defined in QBI (Tuch, 2004), is an SDF-derived analog of the DTI-based Fractional Anisotropy (FA). It quantifies the angular variance of diffusion and is output by GQI for each voxel. GFA provides a unitless 0-1 measure of anisotropy but, like FA, can be confounded by complex fiber architecture (e.g., crossing fibers can lower GFA).
Quantitative Anisotropy (QA): QA was a key new metric introduced with GQI (Yeh et al., 2010). Unlike FA/GFA which are normalized measures per voxel, QA measures the absolute density of diffusional spins along a specific SDF lobe. SDF lobes are interpreted as fiber orientations (fixels in MRtrix terms). By subtracting the isotropic background signal, QA directly reflects the relative volume fraction or “strength” of each fixel in the SDF.
Isotropic diffusion (ISO): ISO refers to the isotropic component separated out by GQI. In practice, GQI estimates the background isotropic diffusion in a voxel as the minimum SDF value (Yeh et al., 2010).
Validation and impact: Foundational studies validated GQI and its metrics: Yeh et al. (2010) showed GQI’s ODF reconstruction accuracy comparable to DSI/QBI and introduced QA. Subsequent work (Yeh et al., 2013) demonstrated QA’s practical advantages: it remains stable despite crossing fibers or edema, leading to improved tractography with fewer false positives.
GQI Studies Across The Lifespan
GQI has predominantly been used as the basis for tractography. There are, however some studies looking at GQI-based metrics across the lifespan.
Development (animal model): GQI-derived metrics generally increase during brain maturation, reflecting strengthening and organization of white matter. For example, Lim et al. (2015) showed that rabbit brains from infancy to adulthood had rising GFA in major tracts, mirroring the well-known FA increases in development. Importantly, GQI metrics unveiled maturation in regions like the hippocampus (gray matter) that DTI FA did not capture, indicating GQI’s potential to detect subtler microstructural ordering.
Pathological aging and injury (animal model): GQI metrics have been applied to models of aging-related brain injury (e.g., radiation exposure in older rabbits). Shen et al. (2015) found QA and ISO could track complex temporal changes (inflammation, edema, recovery) post-radiation that weren’t fully captured by FA. Notably, ISO (isotropic diffusion) tends to increase in situations of edema or tissue loss which is often more pronounced in aging brains (e.g. ventricular expansion, lesions). GQI explicitly quantifies this isotropic water component, which can help disentangle true axonal changes (QA) from mere fluid-related changes (ISO) in elderly or disease populations.
GQI Methodological Warnings and Caveats
Data requirements: GQI’s accuracy and outputs depend on adequate diffusion sampling. In practice, users should ensure high angular resolution or multi-shell data for GQI.
Reproducibility and normalization: Because QA is an absolute signal metric, it can vary with scanner hardware and protocol. Between scanners or sessions, differences in coil gains, b-value, or SNR could lead to QA changes unrelated to biology. This is a concern especially in multi-center studies or longitudinal studies with system upgrades.
Model limitations vs. DTI: GQI expands capability at the cost of complexity. Unlike the simple tensor, GQI has a parameter (the length ratio ~1.2-1.3) to tune the sensitivity to diffusion distance. Using a wrong value could under- or over-estimate QA and fiber (fixel) counts.
References
Fang-Cheng Yeh, Van J Wedeen, and Wen-Yih Isaac Tseng. Generalized q-sampling imaging. IEEE Transactions on Medical Imaging, 29(9):1626–1635, 2010.
Fang-Cheng Yeh, Timothy D Verstynen, Yu Wang, Juan C Fernández-Miranda, and Wen-Yih Isaac Tseng. Deterministic diffusion fiber tracking improved by quantitative anisotropy. PLOS ONE, 8(11):e80713, 2013.
David S Tuch. Q-ball imaging. Magnetic Resonance in Medicine: An Official Journal of the International Society for Magnetic Resonance in Medicine, 52(6):1358–1372, 2004.
S. Y. Lim, Y. S. Tyan, Y. P. Chao, F. Y. Nien, and J. C. Weng. New insights into the developing rabbit brain using diffusion tensor tractography and generalized q-sampling mri. PLOS ONE, 10(3):e0119932, 2015. doi:10.1371/journal.pone.0119932.
C. Y. Shen, Y. S. Tyan, L. W. Kuo, C. W. Wu, and J. C. Weng. Quantitative evaluation of rabbit brain injury after cerebral hemisphere radiation exposure using generalized q-sampling imaging. PLOS ONE, 10(7):e0133001, 2015. doi:10.1371/journal.pone.0133001.