Reconstruction#
Below, an overview of all reconstruction models available on DIPY.
Note
Some reconstruction models do not have a tutorial yet
Method |
Single Shell |
Multi Shell |
Cartesian |
Paper Data Descriptions |
References |
---|---|---|---|---|---|
Yes |
Yes |
Yes |
Typical b-value = 1000 s/mm^2, maximum b-value 1200 s/mm^2 (some success up to 1500 s/mm^2) |
Basser et al. [BMLeBihan94b] |
|
Yes |
Yes |
Yes |
Typical b-value = 1000 s/mm^2, maximum b-value 1200 s/mm^2 (some success up to 1500 s/mm^2) |
Chang et al. [CJP05], Chung et al. [CLH06], Yendiki et al. [YKK+14] |
|
No |
Yes |
No |
DTI-style acquisition, multiple b=0, all shells should be within maximum b-value of 1000 s/mm^2 (or 32 directions evenly distributed 500 s/mm^2 and 1500 s/mm^2 per Neto Henriques et al. [NetoHenriquesRG+17]) |
Pasternak et al. [PSG+09], Neto Henriques et al. [NetoHenriquesRG+17] |
|
No |
Yes |
No |
Dual spin echo diffusion-weighted 2D EPI images were acquired with b values of 0, 500, 1000, 1500, 2000, and 2500 s/mm^2 (max b value of 2000 suggested as sufficient in brain tissue); at least 15 directions |
Jensen et al. [JHR+05] |
|
No |
Yes |
No |
None |
Dela Haije et al. [DelaHaijeOzarslanF20] |
|
No |
Yes |
No |
DKI-style acquisition: at least two non-zero b shells (max b value 2000), minimum of 15 directions; typically b-values in increments of 500 from 0 to 2000, 30 directions |
||
No |
Yes |
No |
b-values in increments of 500 from 0 to 2000 s/mm^2, 30 directions |
Neto Henriques [NetoHenriques18] |
|
Yes |
No |
No |
HARDI data (preferably 7T) with at least 200 diffusion images at b=3000 s/mm^2, or multi-shell data with high angular resolution |
Aganj et al. [ALS+10] |
|
Westins CSA |
Yes |
No |
No |
||
No |
Yes |
No |
low b-values are needed |
Le Bihan et al. [LeBihanBL+88] |
|
No |
Yes |
No |
Fadnavis et al. [FRF+19] |
||
SDT |
Yes |
No |
No |
QBI-style acquisition (60-64 directions, b-value 1000 s/mm^2) |
Descoteaux et al. [DDKnoscheA09] |
No |
No |
Yes |
515 diffusion encodings, b-values from 12,000 to 18,000 s/mm^2. Acceleration in subsequent studies with ~100 diffusion encoding directions in half sphere of the q-space with b-values = 1000, 2000, 3000 s/mm^2) |
||
No |
No |
Yes |
203 diffusion encodings (isotropic 3D grid points in the q-space contained within a sphere with radius 3.6), maximum b-value = 4000 s/mm^2 |
Canales-Rodríguez et al. [CRodriguezIMAlemanGomezMGarcia10] |
|
No |
Yes |
Yes |
Fits any sampling scheme with at least one non-zero b-shell, benefits from more directions. Recommended 23 b-shells ranging from 0 to 4000 in a 258 direction grid-sampling scheme |
Yeh et al. [YWT10] |
|
Yes |
Yes |
No |
At least 40 directions, b-value above 1000 s/mm^2 |
Rokem et al. [RYP+15] |
|
Yes |
No |
No |
At least 64 directions, maximum b-values 3000-4000 s/mm^2, multi-shell, isotropic voxel size |
Tuch [Tuc04], Descoteaux et al. [DAFD07], Tristán-Vega et al. [TristanVAFernandezW09] |
|
No |
Yes |
No |
Multi-shell HARDI data (500, 1000, and 2000 s/mm^2; minimum 2 non-zero b-shells) or DSI (514 images in a cube of five lattice-units, one b=0) |
Merlet and Deriche [MD13], Özarslan et al. [OzarslanKB08], Özarslan et al. [OzarslanKS+09] |
|
No |
Yes |
No |
Six unit sphere shells with b = 1000, 2000, 3000, 4000, 5000, 6000 s/mm^2 along 19, 32, 56, 87, 125, and 170 directions (see Olson et al. [OAM19] for candidate sub-sampling schemes) |
Özarslan et al. [OzarslanKS+13], Olson et al. [OAM19] |
|
No |
Yes |
No |
Dela Haije et al. [DelaHaijeOzarslanF20] |
||
MAPL |
No |
Yes |
No |
Multi-shell similar to WU-Minn HCP, with minimum of 60 samples from 2 shells b-value 1000 and 3000 s/mm^2 |
Fick et al. [FWCD16] |
Yes |
No |
No |
Minimum: 20 gradient directions and a b-value of 1000 s/mm^2; benefits additionally from 60 direction HARDI data with b-value = 3000 s/mm^2 or multi-shell |
Tournier et al. [TCGC04], Tournier et al. [TCC07], Descoteaux et al. [DAFD07] |
|
No |
Yes |
No |
5 b=0, 50 directions at 3 non-zero b-shells: b=1000, 2000, 3000 s/mm^2 |
Jeurissen et al. [JTD+14] |
|
No |
Yes |
No |
Multi-shell 64 direction b-values of 1000, 2000 s/mm^2 as in Alexander et al. [AEA+17]. Original model used 1480 s/mm^2 with 92 directions and 36 b=0 |
||
Yes |
Yes |
Yes |
HARDI data with 64 directions at b = 2500 s/mm^2, 3 b=0 images (full original acquisition: 256 directions on a sphere at b = 2500 s/mm^2, 36 b=0 volumes) |
Canales-Rodríguez et al. [CRodriguezDS+15] |
|
No |
Yes |
No |
Evenly distributed geometric sampling scheme of 216 measurements, 5 b-values (50, 250, 50, 1000, 200 s/mm^2), measurement tensors of four shapes: stick, prolate, sphere, and plane |
Westin et al. [WKP+16] |
|
No |
Yes |
No |
At least one b=0, minimum of 39 acquisitions with spherical and linear encoding; optimal 120 (see Morez et al. [MSdenDekker+23]), ideal 217 see Table 1 in Herberthson et al. [HBH+21] |
Herberthson et al. [HBH+21], Morez et al. [MSdenDekker+23] |
|
Ball & Stick |
Yes |
Yes |
No |
Three b=0, 60 evenly distributed directions per Jones et al. [JHS99] at b-value 1000 s/mm^2 |
Behrens et al. [BWMJ+03] |
No |
Yes |
No |
Minimum 200 volumes of multi-spherical dMRI (multi-shell, multi-diffusion time; varying gradient directions, gradient strengths, and diffusion times) |
Fick et al. [FPS+18] |
|
Power Map |
Yes |
Yes |
No |
HARDI data with 60 directions at b-value = 3000 s/mm^2, 7 b=0 (Minimum: HARDI data with at least 30 directions) |
Dell'Acqua et al. [DellAcquaLCS14] |
No |
Yes |
No |
72 directions at each of 5 evenly spaced b-values from 0.5 to 2.5 ms/μm^2, 5 b-values from 3 to 5 ms/μm^2, 5 b-values from 5.5 to 7.5 ms/μm^2, and 3 b-values from 8 to 9 ms/μm^2 / b=0 ms/μm^2, and along 33 directions at b-values from 0.2–3 ms/μm^2 in steps of 0.2 ms/μm^2 (24 point spherical design and 9 directions identified for rapid kurtosis estimation) |
Kaden et al. [KKC+16], Neto Henriques et al. [NetoHenriquesJS19] |
|
No |
Yes |
No |
Neto Henriques et al. [NetoHenriquesJS20], Neto Henriques et al. [NetoHenriquesJS21], Novello et al. [NNetoHenriquesI+22] |
Applying positivity constraints to Q-space Trajectory Imaging (QTI+)
Reconstruct with Diffusion Spectrum Imaging (DSI)
Reconstruction of the diffusion signal with the correlation tensor model (CTI)
DSI Deconvolution (DSID) vs DSI
Reconstruct with Generalized Q-Sampling Imaging
Continuous and analytical diffusion signal modelling with 3D-SHORE
Reconstruct with Constant Solid Angle (Q-Ball)
Reconstruction with the Sparse Fascicle Model (SFM)
Calculate DSI-based scalar maps
Reconstruction of the diffusion signal with the kurtosis tensor model (DKI)
Reconstruct with Q-space Trajectory Imaging (QTI)
Reconstruction of the diffusion signal with DTI (single tensor) model
K-fold cross-validation for model comparison
Crossing invariant fiber response function with FORECAST model
Local reconstruction using the Histological ResDNN
Reconstruction of the diffusion signal with the WMTI model (DKI-MICRO)
Using the RESTORE algorithm for robust tensor fitting
Using the free water elimination model to remove DTI free water contamination
Signal Reconstruction Using Spherical Harmonics
Reconstruction with Multi-Shell Multi-Tissue CSD
Continuous and analytical diffusion signal modelling with MAP-MRI
Reconstruction with Constrained Spherical Deconvolution model (CSD)
Intravoxel incoherent motion (IVIM)
Reconstruction of Bingham Functions from ODFs
Mean signal diffusion kurtosis imaging (MSDKI)
Reconstruction with Robust and Unbiased Model-BAsed Spherical Deconvolution (RUMBA)
Estimating diffusion time dependent q-space indices using qt-dMRI