waterloo-muscle: sync data card, pipeline grid and figures with GitHub

waterloo-muscle/README.md

@@ -1,4 +1,5 @@
 ---
+name: waterloo-muscle
 pretty_name: UW-MuscleRF
 license: cc-by-4.0
 task_categories:
@@ -20,23 +21,25 @@ size_categories:
 
 # UW-Muscle RF
 
-Dataset consisting of raw RF data and speed of sound measurements acquired in an in vivo speed of sound study conducted by LITMUS @ University of Waterloo.
+![Reconstructed cineloop from Acq_p35_Calf_left_calf_lateral_longitudinal_relaxed_pressure.hdf5](assets/Acq_p35_Calf_left_calf_lateral_longitudinal_relaxed_pressure.gif)
+
+*Cine loop of the relaxed left calf (lateral, longitudinal), [`data/Acq_p35_Calf_left_calf_lateral_longitudinal_relaxed_pressure.hdf5`](https://huggingface.co/datasets/nvidia/OpenH-RF/blob/main/waterloo-muscle/data/Acq_p35_Calf_left_calf_lateral_longitudinal_relaxed_pressure.hdf5), reconstructed from the raw channel data with the `pipeline.yaml` in this folder.*
 
 ## Dataset Description
 
-This is a dataset consisting of 580K raw RF frames (plane wave) and associated SoS measurements using a programmable research scanner configured for high frame rate imaging. We used a rigorous image collection protocol based on landmarking according to bone markers to ensure image consistency. This data was collected as part of the following study:
+18,720 raw RF frames (plane wave; 580,320 frame-transmits) with associated speed-of-sound (SoS) measurements, acquired in vivo by LITMUS at the University of Waterloo with a programmable research scanner configured for high-frame-rate imaging. Images were collected with a rigorous protocol based on landmarking according to bone markers to ensure image consistency. This data was collected as part of the following study:
 
 D. Xiao, P. De La Torre, M. Saif El Nasr, A. J. Y. Chee, M. Mourtzakis, and A. C. H. Yu, “LivePulse-Echo Speed-of-Sound Estimation for Quality Assessment of Large Muscles in Humans,”Ultrasound in Medicine & Biology, vol. 51, no. 11, pp. 1925–1935, Nov. 2025.
 
-
 ## Dataset Contributor(s)
 
-Hassan Nahas, Di Xiao, Pat de la Torre, Adrian J.Y. Chee, Marina Mourtzakis, Alfred C.H. Yu
-
-Correspondence emails:
-hassan.nahas@uwaterloo.ca
-di.xiao@uwaterloo.ca
-alfred.yu@uwaterloo.ca
+- Hassan Nahas <hassan.nahas@uwaterloo.ca>
+- Di Xiao <di.xiao@uwaterloo.ca>
+- Pat de la Torre
+- Adrian J.Y. Chee
+- Marina Mourtzakis
+- Alfred C.H. Yu <alfred.yu@uwaterloo.ca>
+- LITMUS, University of Waterloo
 
 ## Dataset Creation Date
 
@@ -44,10 +47,7 @@ alfred.yu@uwaterloo.ca
 
 ## License / Terms of Use
 
-[Creative Commons Attribution 4.0 International (CC BY 4.0)](https://creativecommons.org/licenses/by/4.0/legalcode.en).
-
-This human study was approved by the University of Waterloo’s Human Research Ethics Board (ORE #44778). All included data was acquired from participants who provided both written and verbal consent prior to participating in the study regarding public data sharing.
-
+[Creative Commons Attribution 4.0 International (CC BY 4.0)](https://creativecommons.org/licenses/by/4.0/legalcode.en). Retain attribution and identify modifications when reusing the data.
 
 ## Intended Usage
 
@@ -63,13 +63,31 @@ The algorithm for global speed of sound estimation can be found here:
 D. Xiao, P. D. l. Torre and A. C. H. Yu, "Real-Time Speed-of-Sound Estimation In Vivo via Steered Plane Wave Ultrasound," in IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, vol. 71, no. 6, pp. 673-686, June 2024, doi: 10.1109/TUFFC.2024.3395490.
 
 
-- **Acquisition system:**
-Raw RF data was acquired from the US4R-Lite research scanner (US4US, Warsaw, Poland), equipped with an L14-5 linear array. For a subset of acquisitions, a through-transmission SoS estimation was made using a custom setup consisting of two single-element Olympus transducers (C567; Olympus; Tokyo, Japan).
+- **Acquisition system:** Raw RF data was acquired from the US4R-Lite research scanner (US4US, Warsaw, Poland), equipped with an L14-5 linear array. For a subset of acquisitions, a through-transmission SoS estimation was made using a custom setup consisting of two single-element Olympus transducers (C567; Olympus; Tokyo, Japan).
+
+## Processing the Dataset
+
+The acquisitions can be processed with the `pipeline.yaml` definition in this folder and the [zea library](https://github.com/tue-bmd/zea).
+
+`zea` streams the data from the Hugging Face Hub and processes it according to the pipeline. You can try it out with the following command:
+
+```bash
+zea process \
+  --dataset hf://nvidia/OpenH-RF/waterloo-muscle/data/Acq_p35_Calf_left_calf_lateral_longitudinal_relaxed_pressure.hdf5 \
+  --config hf://nvidia/OpenH-RF/waterloo-muscle/pipeline.yaml \
+  --n-frames 1 \
+  --save-as png
+```
+
+Alternatively, you can use the `reconstruct.py` [script](https://github.com/open-h/OpenH-RF/blob/main/datasets/waterloo-muscle/reconstruct.py) as provided in the [OpenH-RF GitHub repository](https://github.com/open-h/OpenH-RF).
+
+Swap `--n-frames 1 --save-as png` for `--save-as gif` to get the cine loop. In the script, `ZEA_FILE` and `FRAME` at the top select what is reconstructed.
 
 ## Dataset Format
 
-Submitted in the [`zea` file format](https://zea.readthedocs.io/en/latest/)
-(one HDF5 file per acquisition).
+[zea v0.1.6](https://github.com/tue-bmd/zea)
+
+Submitted in the [`zea` file format](https://zea.readthedocs.io/en/latest/) (one HDF5 file per acquisition).
 
 Per-sample contents of the converted HDF5:
 
@@ -89,14 +107,14 @@ Per-sample contents of the converted HDF5:
 | `custom/baecke_sport_index` | `[1]` | float32 |  /5  | Baecke sport activity score |
 | `custom/baecke_leisure_index` | `[1]` | float32 |  /5  | Baecke leisure activity score |
 | `custom/baecke_score` | `[1]` | float32 |  /15  | Total baecke score |
-All `coordinates` arrays are per-pixel Cartesian positions in metres, last axis
-`[x, y, z]` (y = 0 for these 2-D maps).
+
+All `coordinates` arrays are per-pixel Cartesian positions in metres, last axis `[x, y, z]` (y = 0 for these 2-D maps).
 
 ## Dataset Quantification
 
 **Current OpenH-RF release:** 1,248 HDF5 files; 576.19 GB (576,186,417,152 bytes) stored; root `zea_version` **0.1.6**. Sizes include all HDF5 contents and use decimal units (MB = 10^6 bytes, GB = 10^9 bytes, TB = 10^12 bytes), not decoded-array memory or original-source download sizes.
 
-Data was collected from 39 participants, each with 32 unique images spanning calf/bicep/quad, axis and muscle state. For each imaging location, 15 frames were made per limb under minimal contact and with pressure. Given that each frame consisted of 31 steered plane waves, our protocol yielded a total of 15 repeats × 31 frames × 8 views × 2 sides × 2 pressure settings = 14880 raw RF frames per participant. In total, our dataset is expected to contain 580k frames of raw RF data.
+Data was collected from 39 participants, each with 32 unique images spanning calf/bicep/quad, axis and muscle state. For each imaging location, 15 frames were made per limb under minimal contact and with pressure. Given that each frame consisted of 31 steered plane waves, our protocol yielded a total of 15 repeats × 31 transmits × 8 views × 2 sides × 2 pressure settings = 14,880 frame-transmits per participant, i.e. 480 stored frames per participant. In total the dataset contains 18,720 stored frames and 580,320 frame-transmits (see the table below).
 
 All 1,248 HDF5 files are uploaded; current stored size and format version are reported above.
 
@@ -116,13 +134,7 @@ All 1,248 HDF5 files are uploaded; current stored size and format version are re
 
 ## Data Validation
 
-[`reconstruct.py`](reconstruct.py) builds a `zea.Pipeline` of DAS beamforming →
-envelope detection → normalization → log-compression **in code** and reconstructs
-a B-mode directly from `raw_data` — showing the raw-to-image flow without any
-config file. It also saves the pipeline to [`pipeline.yaml`](pipeline.yaml) as a
-shareable recipe. Comparing the reconstruction against the stored B-mode is a
-sanity check that the acquisition parameters and probe geometry are recorded
-correctly, and serves as a reproducible reference reconstruction.
+`reconstruct.py` builds a `zea.Pipeline` of DAS beamforming → envelope detection → normalization → log-compression **in code** and reconstructs a B-mode directly from `raw_data` — showing the raw-to-image flow without any config file. It also saves the pipeline to `pipeline.yaml` as a shareable recipe. Comparing the reconstruction against the stored B-mode is a sanity check that the acquisition parameters and probe geometry are recorded correctly, and serves as a reproducible reference reconstruction.
 
 ## Known Issues
 - Scan.sound_speed uses default 1540 m/s used for computing tx delays as was done during acquisition. This is different from the estimated global speed of sound which is currently stored as a custom element.

waterloo-muscle/pipeline.yaml

@@ -17,3 +17,15 @@ pipeline:
             - 0.0
             - 1.0
     - log_compress
+parameters:
+    xlims:
+    - -0.019
+    - 0.019
+    zlims:
+    - 0.0
+    - 0.08
+    grid_size_x: 381
+    grid_size_z: 801
+    dynamic_range:
+    - -60
+    - 0