rfick commited on
Commit
c6cfde0
·
verified ·
1 Parent(s): 6a1f85d

Upload folder using huggingface_hub

Browse files
.gitattributes CHANGED
@@ -58,3 +58,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
58
  # Video files - compressed
59
  *.mp4 filter=lfs diff=lfs merge=lfs -text
60
  *.webm filter=lfs diff=lfs merge=lfs -text
 
 
 
58
  # Video files - compressed
59
  *.mp4 filter=lfs diff=lfs merge=lfs -text
60
  *.webm filter=lfs diff=lfs merge=lfs -text
61
+ cactus-axons/single-bundle-g070-icvf061-3T.rpk filter=lfs diff=lfs merge=lfs -text
62
+ canonical-wm/g070-f055-3T.rpk filter=lfs diff=lfs merge=lfs -text
README.md CHANGED
@@ -1,3 +1,28 @@
1
  ---
2
  license: cc-by-4.0
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: cc-by-4.0
3
+ tags:
4
+ - diffusion-mri
5
+ - monte-carlo
6
+ - replay-pack
7
+ - microstructure
8
+ pretty_name: dmrai-lab substrate bank
9
  ---
10
+
11
+ # dmrai-lab substrate bank
12
+
13
+ Compressed Monte-Carlo **replay packs** (`.rpk`) for diffusion-MRI microstructure. Each pack is one master walk on a substrate plus the physics channels needed to **replay** any acquisition inside its certified envelope — *one walk, every acquisition* — in the open [Replay Pack Specification](https://github.com/dmrai-lab/replay-pack-spec) format.
14
+
15
+ ```python
16
+ from dmipy_sim import bank
17
+ pack = bank.pull("<id>", repo_id="dmrai-lab/substrate-bank")
18
+ S = pack.replay(waveform, relaxation=True)
19
+ ```
20
+
21
+ ## Substrates
22
+
23
+ | Substrate | Replay tiers | Size | License |
24
+ |---|---|---|---|
25
+ | [`cactus-axons/single-bundle-g070-icvf061-3T`](cactus-axons/single-bundle-g070-icvf061-3T.md) | C0 C1 C2 | 5.84 MB | CC-BY-4.0 |
26
+ | [`canonical-wm/g070-f055-3T`](canonical-wm/g070-f055-3T.md) | C0 | 18.23 MB | CC-BY-4.0 |
27
+
28
+ Each substrate links to its own **substrate card** with geometry, provenance, fidelity and load/replay instructions. Tier legend: C0 gradient · C1 bulk relaxation · C2 surface relaxivity · C3 field · C4 magnetization transfer.
SHA256SUMS ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ 9a7dd9015b0a5cf5a4e0715c406c483404e7a10234572a76f0504d1142ada5d3 cactus-axons/single-bundle-g070-icvf061-3T.rpk
2
+ bb57348ea115be2900cfb44eb0ad13b3ba307f1d2b967086136d9fbd23423106 canonical-wm/g070-f055-3T.rpk
cactus-axons/single-bundle-g070-icvf061-3T.croissant.jsonld ADDED
@@ -0,0 +1,120 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "@context": {
3
+ "@vocab": "https://schema.org/",
4
+ "cr": "http://mlcommons.org/croissant/"
5
+ },
6
+ "@type": "Dataset",
7
+ "name": "cactus-axons/single-bundle-g070-icvf061-3T",
8
+ "description": "dmipy-sim replay pack: compressed Monte-Carlo master walk (lowrank K=96) with replay envelope {'gradient': True, 'bulk_relaxation': True, 'surface_relaxivity': True, 'field': False, 'magnetization_transfer': False, 'diffusivity_fixed': True, 'acquisition': {'b_max': 3000000000.0, 'ogse_periods': [1, 2, 3, 5], 'B0_list': [3.0, 7.0, 11.7], 'note': 'temporal band set by max OGSE period / min delta'}}.",
9
+ "license": "CC-BY-4.0",
10
+ "citeAs": "Fick RHJ, dmrai-lab (2026). CACTUS single-bundle Monte-Carlo master walk. Geometry generator: Villarreal-Haro et al., CACTUS (https://github.com/Juanitovh/CACTUS).",
11
+ "cr:provenance": {
12
+ "substrate": "CACTUS single bundle",
13
+ "real_or_synthetic": "synthetic",
14
+ "generator": {
15
+ "name": "CACTUS",
16
+ "authors": "Villarreal-Haro et al.",
17
+ "url": "https://github.com/Juanitovh/CACTUS",
18
+ "patch": "workers/meta_grid.py: np.random.randn(n,3)->standard_normal((int(n),3)) for numba compat"
19
+ },
20
+ "g_ratio": 0.7,
21
+ "icvf": 0.6076,
22
+ "f_intra": 0.3486,
23
+ "f_myelin": 0.2589,
24
+ "f_extra": 0.3924,
25
+ "n_fibres": 366,
26
+ "box_um": [
27
+ 30.0,
28
+ 30.0,
29
+ 30.0
30
+ ],
31
+ "fibre_axis": "x",
32
+ "TE": 0.04,
33
+ "feature_radius_um": {
34
+ "intra": 0.25,
35
+ "extra": 0.4
36
+ },
37
+ "boundary": "reflecting voxel box (extra pool)",
38
+ "cactus_config": "#Step1: Substrate Initialization parameters\n\n#### Algorithm parameters\noutfile cactus_bundle\nn_cores 10\nverbose 1 \n\n\n\n#substrate constant parameters\nn_bundles 1\nlenght_side 30\nmin_rad .27\nmax_rad 2.2\nbias -.03\n\n\n### combinational parameters\nnumber_of_repetitions 1\nicvf 0.65\ndispersion 10 \n\ngamma_theta 2.40 \ngamma_kappa 0.25 \n\n\n\n\n#Step 2: joint fibre optimization parameters \n\noptimization_tolerance 50\nmax_iterations 1500\nalpha 1\nsubdivisions_number 10\n\n\n#Step 3: Fibre radial growth (FRG) and meshing parameteres\n\ngrid_size 0.35\ngrowth_iterations 4\nn_segments 4\ncolorless 1\nsim_vol simulations\ninn_out outer\ng_ratio 0.7\nn_erode 0\n\n\n",
39
+ "optimized_final_header": [
40
+ "30",
41
+ "366",
42
+ "18"
43
+ ],
44
+ "tiers_present": [
45
+ "C0-gradient",
46
+ "C1-bulk-relaxation",
47
+ "C2-surface-relaxivity"
48
+ ],
49
+ "tiers_deferred": {
50
+ "C3_field": "rpk susceptibility channel is 2D translationally-invariant; CACTUS needs 3D field maps",
51
+ "C4_mt": "MT binding needs R/25 sub-step -> ~3.5M mesh reflections/walker on sub-micron features"
52
+ }
53
+ },
54
+ "cr:replayEnvelope": {
55
+ "gradient": true,
56
+ "bulk_relaxation": true,
57
+ "surface_relaxivity": true,
58
+ "field": false,
59
+ "magnetization_transfer": false,
60
+ "diffusivity_fixed": true,
61
+ "acquisition": {
62
+ "b_max": 3000000000.0,
63
+ "ogse_periods": [
64
+ 1,
65
+ 2,
66
+ 3,
67
+ 5
68
+ ],
69
+ "B0_list": [
70
+ 3.0,
71
+ 7.0,
72
+ 11.7
73
+ ],
74
+ "note": "temporal band set by max OGSE period / min delta"
75
+ }
76
+ },
77
+ "cr:fidelity": {
78
+ "metric": "max_abs_complex_signal_error",
79
+ "err_max": 0.0005992180903185556,
80
+ "floor_max": 0.015136345027932882,
81
+ "noise_floor": 0.007071067811865475,
82
+ "within_2x_floor": true,
83
+ "per_family": {
84
+ "OGSE1": {
85
+ "err_max": 4.110324120514571e-05,
86
+ "floor_max": 0.01101313849143775
87
+ },
88
+ "OGSE2": {
89
+ "err_max": 3.0114713312470272e-05,
90
+ "floor_max": 0.007819370051004686
91
+ },
92
+ "OGSE3": {
93
+ "err_max": 0.00021818730268882545,
94
+ "floor_max": 0.008300622238353213
95
+ },
96
+ "OGSE5": {
97
+ "err_max": 0.00032166529279630717,
98
+ "floor_max": 0.008387572597074308
99
+ },
100
+ "PGSE": {
101
+ "err_max": 5.3429294722594866e-05,
102
+ "floor_max": 0.008324886028105324
103
+ },
104
+ "SHORTD": {
105
+ "err_max": 0.0005992180903185556,
106
+ "floor_max": 0.015136345027932882
107
+ }
108
+ },
109
+ "target_floor": 0.001,
110
+ "meets_target": false
111
+ },
112
+ "distribution": [
113
+ {
114
+ "@type": "cr:FileObject",
115
+ "name": "cactus-axons/single-bundle-g070-icvf061-3T.rpk",
116
+ "encodingFormat": "application/octet-stream",
117
+ "contentUrl": null
118
+ }
119
+ ]
120
+ }
cactus-axons/single-bundle-g070-icvf061-3T.md ADDED
@@ -0,0 +1,140 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-4.0
3
+ tags:
4
+ - diffusion-mri
5
+ - monte-carlo
6
+ - replay-pack
7
+ - microstructure
8
+ - dmipy
9
+ - cactus
10
+ - white-matter
11
+ pretty_name: single-bundle-g070-icvf061-3T
12
+ ---
13
+
14
+ # Substrate card — `cactus-axons/single-bundle-g070-icvf061-3T`
15
+
16
+ A **replay pack** (`.rpk`) — a compressed Monte-Carlo master walk on a *synthetic* diffusion-MRI substrate, plus every physics channel needed to **replay** any acquisition inside its certified envelope without re-simulating. One walk, every acquisition. Format: the open [Replay Pack Specification](https://github.com/dmrai-lab/replay-pack-spec).
17
+
18
+ ## Geometry & provenance
19
+
20
+ | Property | Value |
21
+ |---|---|
22
+ | Generator | [CACTUS](https://github.com/Juanitovh/CACTUS) |
23
+ | Generator authors | Villarreal-Haro et al. |
24
+ | Substrate | CACTUS single bundle |
25
+ | Fibres | 366 |
26
+ | g-ratio | 0.7 |
27
+ | ICVF (intra+myelin) | 0.6076 |
28
+ | f_intra | 0.3486 |
29
+ | f_myelin | 0.2589 |
30
+ | f_extra | 0.3924 |
31
+ | Voxel box (µm) | [30.0, 30.0, 30.0] |
32
+ | Fibre axis | x |
33
+ | Boundary condition | reflecting voxel box (extra pool) |
34
+
35
+ ## Replay capability
36
+
37
+ | Tier | Channel | Present | Meaning |
38
+ |---|---|:--:|---|
39
+ | C0 | Gradient | ✅ | positions — any q/b, direction, waveform (PGSE/PGSTE/OGSE/…) |
40
+ | C1 | Bulk relaxation | ✅ | per-compartment T2/T1 via the compartment map |
41
+ | C2 | Surface relaxivity | ✅ | boundary-local-time channel (any ρ) |
42
+ | C3 | Field / susceptibility | — | stored susceptibility field maps (any B0, χ, orientation) |
43
+ | C4 | Magnetization transfer | — | bound-fraction channel (MT/ihMT) |
44
+
45
+ **Deferred tiers (with reason):**
46
+
47
+ - **C3_field** — rpk susceptibility channel is 2D translationally-invariant; CACTUS needs 3D field maps
48
+ - **C4_mt** — MT binding needs R/25 sub-step -> ~3.5M mesh reflections/walker on sub-micron features
49
+
50
+ > Diffusivity is fixed at walk time (a geometry property, not a replay knob).
51
+
52
+ ## Fidelity
53
+
54
+ | Metric | Value |
55
+ |---|---|
56
+ | Max replay error | 0.0005992 |
57
+ | MC noise floor | 0.01514 |
58
+ | Within 2× floor | True |
59
+ | Target floor σ* | 0.001 |
60
+ | Meets σ* | False |
61
+ | Walkers | 20000 |
62
+ | Saved time points | 81 |
63
+ | T_max (s) | 0.04 |
64
+ | Compression | lowrank (K=96) |
65
+
66
+ ## Load & replay
67
+
68
+ ```python
69
+ import numpy as np
70
+ from dmipy_sim import bank
71
+
72
+ pack = bank.pull("cactus-axons/single-bundle-g070-icvf061-3T", repo_id="dmrai-lab/substrate-bank") # cached + hash-verified
73
+
74
+ # Build any waveform G on the pack's own time grid (shape (n_meas, pack.n_t, 3), T/m),
75
+ # e.g. a PGSE perpendicular to the fibre axis, then replay it:
76
+ S = pack.replay(waveform, relaxation=True) # C1 relaxation; add rho=... for C2 surface
77
+ # perpendicular vs parallel gradients recover the fibre anisotropy (S_perp > S_par).
78
+ ```
79
+
80
+ ## Generator configuration
81
+
82
+ The exact generator config that produced this geometry (also embedded in the `.rpk` provenance):
83
+
84
+ ```
85
+ #Step1: Substrate Initialization parameters
86
+
87
+ #### Algorithm parameters
88
+ outfile cactus_bundle
89
+ n_cores 10
90
+ verbose 1
91
+
92
+
93
+
94
+ #substrate constant parameters
95
+ n_bundles 1
96
+ lenght_side 30
97
+ min_rad .27
98
+ max_rad 2.2
99
+ bias -.03
100
+
101
+
102
+ ### combinational parameters
103
+ number_of_repetitions 1
104
+ icvf 0.65
105
+ dispersion 10
106
+
107
+ gamma_theta 2.40
108
+ gamma_kappa 0.25
109
+
110
+
111
+
112
+
113
+ #Step 2: joint fibre optimization parameters
114
+
115
+ optimization_tolerance 50
116
+ max_iterations 1500
117
+ alpha 1
118
+ subdivisions_number 10
119
+
120
+
121
+ #Step 3: Fibre radial growth (FRG) and meshing parameteres
122
+
123
+ grid_size 0.35
124
+ growth_iterations 4
125
+ n_segments 4
126
+ colorless 1
127
+ sim_vol simulations
128
+ inn_out outer
129
+ g_ratio 0.7
130
+ n_erode 0
131
+ ```
132
+
133
+ Geometry header (`optimized_final.txt`): voxel_side=30, n_fibres=366, n_control_points=18.
134
+
135
+ > Reproducibility note: workers/meta_grid.py: np.random.randn(n,3)->standard_normal((int(n),3)) for numba compat
136
+
137
+ ## Citation & license
138
+
139
+ **License:** CC-BY-4.0
140
+ **Cite:** Fick RHJ, dmrai-lab (2026). CACTUS single-bundle Monte-Carlo master walk. Geometry generator: Villarreal-Haro et al., CACTUS (https://github.com/Juanitovh/CACTUS).
cactus-axons/single-bundle-g070-icvf061-3T.rpk ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9a7dd9015b0a5cf5a4e0715c406c483404e7a10234572a76f0504d1142ada5d3
3
+ size 5838772
cactus-axons/single-bundle-g070-icvf061-3T__cactus_config.txt ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #Step1: Substrate Initialization parameters
2
+
3
+ #### Algorithm parameters
4
+ outfile cactus_bundle
5
+ n_cores 10
6
+ verbose 1
7
+
8
+
9
+
10
+ #substrate constant parameters
11
+ n_bundles 1
12
+ lenght_side 30
13
+ min_rad .27
14
+ max_rad 2.2
15
+ bias -.03
16
+
17
+
18
+ ### combinational parameters
19
+ number_of_repetitions 1
20
+ icvf 0.65
21
+ dispersion 10
22
+
23
+ gamma_theta 2.40
24
+ gamma_kappa 0.25
25
+
26
+
27
+
28
+
29
+ #Step 2: joint fibre optimization parameters
30
+
31
+ optimization_tolerance 50
32
+ max_iterations 1500
33
+ alpha 1
34
+ subdivisions_number 10
35
+
36
+
37
+ #Step 3: Fibre radial growth (FRG) and meshing parameteres
38
+
39
+ grid_size 0.35
40
+ growth_iterations 4
41
+ n_segments 4
42
+ colorless 1
43
+ sim_vol simulations
44
+ inn_out outer
45
+ g_ratio 0.7
46
+ n_erode 0
47
+
48
+
canonical-wm/g070-f055-3T.croissant.jsonld ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "@context": {
3
+ "@vocab": "https://schema.org/",
4
+ "cr": "http://mlcommons.org/croissant/"
5
+ },
6
+ "@type": "Dataset",
7
+ "name": "canonical-wm/g070-f055-3T",
8
+ "description": "dmipy-sim replay pack: compressed Monte-Carlo master walk (lowrank K=64) with replay envelope {'gradient': True, 'rf': True, 'B0_any': True, 'orientation_any': True, 'T1T2': True, 'rho': False, 'mt': False, 'permeability': False, 'diffusivity_fixed': True, 'acquisition': {'b_max': 3000000000.0, 'ogse_periods': [1, 2, 3, 5], 'B0_list': [3.0, 7.0, 11.7], 'note': 'temporal band set by max OGSE period / min delta'}}.",
9
+ "license": "CC-BY-4.0",
10
+ "citeAs": "Fick RHJ, dmrai-lab (2026), canonical WM master (synthetic).",
11
+ "cr:provenance": {
12
+ "g_ratio": 0.7,
13
+ "packing_density": 0.55,
14
+ "real_or_synthetic": "synthetic"
15
+ },
16
+ "cr:replayEnvelope": {
17
+ "gradient": true,
18
+ "rf": true,
19
+ "B0_any": true,
20
+ "orientation_any": true,
21
+ "T1T2": true,
22
+ "rho": false,
23
+ "mt": false,
24
+ "permeability": false,
25
+ "diffusivity_fixed": true,
26
+ "acquisition": {
27
+ "b_max": 3000000000.0,
28
+ "ogse_periods": [
29
+ 1,
30
+ 2,
31
+ 3,
32
+ 5
33
+ ],
34
+ "B0_list": [
35
+ 3.0,
36
+ 7.0,
37
+ 11.7
38
+ ],
39
+ "note": "temporal band set by max OGSE period / min delta"
40
+ }
41
+ },
42
+ "cr:fidelity": {
43
+ "metric": "max_abs_complex_signal_error",
44
+ "err_max": 0.0015449349215161125,
45
+ "floor_max": 0.007417712466507749,
46
+ "noise_floor": 0.002886751345948129,
47
+ "within_2x_floor": true,
48
+ "per_family": {
49
+ "OGSE1": {
50
+ "err_max": 2.6567877097349795e-05,
51
+ "floor_max": 0.007417712466507749
52
+ },
53
+ "OGSE2": {
54
+ "err_max": 5.227884402603549e-05,
55
+ "floor_max": 0.004460657597099843
56
+ },
57
+ "OGSE3": {
58
+ "err_max": 0.0002155028703048945,
59
+ "floor_max": 0.005062851187307158
60
+ },
61
+ "OGSE5": {
62
+ "err_max": 0.00032346565433251715,
63
+ "floor_max": 0.0037408410155284743
64
+ },
65
+ "PGSE": {
66
+ "err_max": 4.928880879743661e-05,
67
+ "floor_max": 0.004254416190880235
68
+ },
69
+ "SHORTD": {
70
+ "err_max": 0.0015449349215161125,
71
+ "floor_max": 0.003591027423547303
72
+ }
73
+ }
74
+ },
75
+ "distribution": [
76
+ {
77
+ "@type": "cr:FileObject",
78
+ "name": "canonical-wm/g070-f055-3T.rpk",
79
+ "encodingFormat": "application/octet-stream",
80
+ "contentUrl": null
81
+ }
82
+ ]
83
+ }
canonical-wm/g070-f055-3T.md ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-4.0
3
+ tags:
4
+ - diffusion-mri
5
+ - monte-carlo
6
+ - replay-pack
7
+ - microstructure
8
+ - dmipy
9
+ - white-matter
10
+ pretty_name: g070-f055-3T
11
+ ---
12
+
13
+ # Substrate card — `canonical-wm/g070-f055-3T`
14
+
15
+ A **replay pack** (`.rpk`) — a compressed Monte-Carlo master walk on a *synthetic* diffusion-MRI substrate, plus every physics channel needed to **replay** any acquisition inside its certified envelope without re-simulating. One walk, every acquisition. Format: the open [Replay Pack Specification](https://github.com/dmrai-lab/replay-pack-spec).
16
+
17
+ ## Geometry & provenance
18
+
19
+ | Property | Value |
20
+ |---|---|
21
+ | g-ratio | 0.7 |
22
+
23
+ ## Replay capability
24
+
25
+ | Tier | Channel | Present | Meaning |
26
+ |---|---|:--:|---|
27
+ | C0 | Gradient | ✅ | positions — any q/b, direction, waveform (PGSE/PGSTE/OGSE/…) |
28
+ | C1 | Bulk relaxation | — | per-compartment T2/T1 via the compartment map |
29
+ | C2 | Surface relaxivity | — | boundary-local-time channel (any ρ) |
30
+ | C3 | Field / susceptibility | — | stored susceptibility field maps (any B0, χ, orientation) |
31
+ | C4 | Magnetization transfer | — | bound-fraction channel (MT/ihMT) |
32
+
33
+ > Diffusivity is fixed at walk time (a geometry property, not a replay knob).
34
+
35
+ ## Fidelity
36
+
37
+ | Metric | Value |
38
+ |---|---|
39
+ | Max replay error | 0.001545 |
40
+ | MC noise floor | 0.007418 |
41
+ | Within 2× floor | True |
42
+ | Walkers | 120000 |
43
+ | Saved time points | 301 |
44
+ | T_max (s) | 0.06 |
45
+ | Compression | lowrank (K=64) |
46
+
47
+ ## Load & replay
48
+
49
+ ```python
50
+ import numpy as np
51
+ from dmipy_sim import bank
52
+
53
+ pack = bank.pull("canonical-wm/g070-f055-3T", repo_id="dmrai-lab/substrate-bank") # cached + hash-verified
54
+
55
+ # Build any waveform G on the pack's own time grid (shape (n_meas, pack.n_t, 3), T/m),
56
+ # e.g. a PGSE perpendicular to the fibre axis, then replay it:
57
+ S = pack.replay(waveform, relaxation=True) # C1 relaxation; add rho=... for C2 surface
58
+ # perpendicular vs parallel gradients recover the fibre anisotropy (S_perp > S_par).
59
+ ```
60
+
61
+ ## Citation & license
62
+
63
+ **License:** CC-BY-4.0
64
+ **Cite:** Fick RHJ, dmrai-lab (2026), canonical WM master (synthetic).
canonical-wm/g070-f055-3T.rpk ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb57348ea115be2900cfb44eb0ad13b3ba307f1d2b967086136d9fbd23423106
3
+ size 18231174
manifest.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema": "dmipy-sim substrate-bank/1.1",
3
+ "n_entries": 2,
4
+ "entries": [
5
+ {
6
+ "id": "cactus-axons/single-bundle-g070-icvf061-3T",
7
+ "file": "cactus-axons/single-bundle-g070-icvf061-3T.rpk",
8
+ "sha256": "9a7dd9015b0a5cf5a4e0715c406c483404e7a10234572a76f0504d1142ada5d3",
9
+ "bytes": 5838772,
10
+ "license": "CC-BY-4.0",
11
+ "replay_envelope": {
12
+ "gradient": true,
13
+ "bulk_relaxation": true,
14
+ "surface_relaxivity": true,
15
+ "field": false,
16
+ "magnetization_transfer": false,
17
+ "diffusivity_fixed": true,
18
+ "acquisition": {
19
+ "b_max": 3000000000.0,
20
+ "ogse_periods": [
21
+ 1,
22
+ 2,
23
+ 3,
24
+ 5
25
+ ],
26
+ "B0_list": [
27
+ 3.0,
28
+ 7.0,
29
+ 11.7
30
+ ],
31
+ "note": "temporal band set by max OGSE period / min delta"
32
+ }
33
+ },
34
+ "fidelity": 0.0005992180903185556,
35
+ "compression": {
36
+ "method": "lowrank",
37
+ "K": 96,
38
+ "walker_preserving": true,
39
+ "channels": {
40
+ "compartment": {
41
+ "channel": "compartment",
42
+ "fractional": false,
43
+ "n_t": 81
44
+ },
45
+ "boundary_local_time": {
46
+ "channel": "boundary_local_time",
47
+ "mode": "dense",
48
+ "n_t": 81,
49
+ "scale": 3.8564205169677734e-05,
50
+ "nlevels": 127
51
+ }
52
+ }
53
+ }
54
+ },
55
+ {
56
+ "id": "canonical-wm/g070-f055-3T",
57
+ "file": "canonical-wm/g070-f055-3T.rpk",
58
+ "sha256": "bb57348ea115be2900cfb44eb0ad13b3ba307f1d2b967086136d9fbd23423106",
59
+ "bytes": 18231174,
60
+ "license": "CC-BY-4.0",
61
+ "replay_envelope": {
62
+ "gradient": true,
63
+ "rf": true,
64
+ "B0_any": true,
65
+ "orientation_any": true,
66
+ "T1T2": true,
67
+ "rho": false,
68
+ "mt": false,
69
+ "permeability": false,
70
+ "diffusivity_fixed": true,
71
+ "acquisition": {
72
+ "b_max": 3000000000.0,
73
+ "ogse_periods": [
74
+ 1,
75
+ 2,
76
+ 3,
77
+ 5
78
+ ],
79
+ "B0_list": [
80
+ 3.0,
81
+ 7.0,
82
+ 11.7
83
+ ],
84
+ "note": "temporal band set by max OGSE period / min delta"
85
+ }
86
+ },
87
+ "fidelity": 0.0015449349215161125,
88
+ "compression": {
89
+ "method": "lowrank",
90
+ "K": 64,
91
+ "walker_preserving": true
92
+ }
93
+ }
94
+ ]
95
+ }