ppcnet-dataset commited on
Commit
9db47cd
·
verified ·
1 Parent(s): 1070bc2

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +11 -15
README.md CHANGED
@@ -16,8 +16,6 @@ size_categories:
16
  - 1K<n<10K
17
  ---
18
 
19
- <div align="center">
20
-
21
  # PPCNet Dataset
22
  ## Biplanar DRRs + Projection Matrices + Ground-Truth Point Clouds
23
 
@@ -28,11 +26,9 @@ size_categories:
28
 
29
  *A curated lumbar spine dataset for 3D point cloud reconstruction from biplanar radiographs.*
30
 
31
- </div>
32
-
33
  ---
34
 
35
- ## 📋 Overview
36
 
37
  This dataset provides paired biplanar DRRs, calibrated 3×4 projection matrices, and dense ground-truth point clouds for **1,037 patients** with complete L1–L5 lumbar vertebrae, derived from the publicly available **VerSe'19**, **VerSe'20**, and **CTSpine1K** collections.
38
 
@@ -55,7 +51,7 @@ Each of the 1,037 patient folders contains:
55
 
56
  ---
57
 
58
- ## 📊 Data Split
59
 
60
  Fixed split in `dataset_split.json` (seed=42):
61
 
@@ -68,7 +64,7 @@ Fixed split in `dataset_split.json` (seed=42):
68
 
69
  ---
70
 
71
- ## 🔬 DRR Generation
72
 
73
  DRRs are generated using **Plastimatch's ray-casting algorithm** with the following parameters:
74
 
@@ -86,15 +82,15 @@ Each DRR comes with a calibrated **3×4 perspective projection matrix** encoding
86
 
87
  ---
88
 
89
- ## 🎯 Ground-Truth Generation
90
 
91
  The ground-truth point clouds are constructed through a five-step pipeline:
92
 
93
- 1. 🦴 **Surface extraction** — marching cubes on per-vertebra segmentation masks
94
- 2. 📐 **Coordinate transform** — vertices mapped to physical (mm) space via NIfTI affine
95
- 3. 🔗 **Concatenation** — all five vertebrae (L1–L5) merged into a single cloud
96
- 4. 🎯 **Alignment** — rigid ICP with four flip initialisations to match DRR projection space
97
- 5. ✂️ **Sampling** — farthest-point sampling to a uniform **5,120-point** representation
98
 
99
  ---
100
 
@@ -120,7 +116,7 @@ huggingface-cli download ppcnet-dataset/PPCNet --repo-type dataset --local-dir .
120
 
121
  ---
122
 
123
- ## 🔗 Source Datasets
124
 
125
  This dataset builds upon:
126
 
@@ -129,7 +125,7 @@ This dataset builds upon:
129
 
130
  ---
131
 
132
- ## 📜 License
133
 
134
  This dataset is released under the [MIT License](LICENSE).
135
 
 
16
  - 1K<n<10K
17
  ---
18
 
 
 
19
  # PPCNet Dataset
20
  ## Biplanar DRRs + Projection Matrices + Ground-Truth Point Clouds
21
 
 
26
 
27
  *A curated lumbar spine dataset for 3D point cloud reconstruction from biplanar radiographs.*
28
 
 
 
29
  ---
30
 
31
+ ## Overview
32
 
33
  This dataset provides paired biplanar DRRs, calibrated 3×4 projection matrices, and dense ground-truth point clouds for **1,037 patients** with complete L1–L5 lumbar vertebrae, derived from the publicly available **VerSe'19**, **VerSe'20**, and **CTSpine1K** collections.
34
 
 
51
 
52
  ---
53
 
54
+ ## Data Split
55
 
56
  Fixed split in `dataset_split.json` (seed=42):
57
 
 
64
 
65
  ---
66
 
67
+ ## DRR Generation
68
 
69
  DRRs are generated using **Plastimatch's ray-casting algorithm** with the following parameters:
70
 
 
82
 
83
  ---
84
 
85
+ ## Ground-Truth Generation
86
 
87
  The ground-truth point clouds are constructed through a five-step pipeline:
88
 
89
+ 1. **Surface extraction** — marching cubes on per-vertebra segmentation masks
90
+ 2. **Coordinate transform** — vertices mapped to physical (mm) space via NIfTI affine
91
+ 3. **Concatenation** — all five vertebrae (L1–L5) merged into a single cloud
92
+ 4. **Alignment** — rigid ICP with four flip initialisations to match DRR projection space
93
+ 5. **Sampling** — farthest-point sampling to a uniform **5,120-point** representation
94
 
95
  ---
96
 
 
116
 
117
  ---
118
 
119
+ ## Source Datasets
120
 
121
  This dataset builds upon:
122
 
 
125
 
126
  ---
127
 
128
+ ## License
129
 
130
  This dataset is released under the [MIT License](LICENSE).
131