File size: 3,907 Bytes
a6dce42
68f3d02
 
 
 
 
 
 
 
a6dce42
 
 
 
68f3d02
 
 
 
 
 
 
a6dce42
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
68f3d02
a6dce42
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
---
pretty_name: "OpenH-RF eSAF — Raw Source Data Archive"
license: cc-by-4.0
tags:
  - ultrasound
  - rf
  - openh-rf
  - provenance
  - raw-data
---

# OpenH-RF eSAF — Raw Source Data Archive

**This is a provenance/reproducibility archive — not itself the OpenH-RF
submission.** The submitted dataset (195 zea-format `.hdf5` acquisitions + data card
+ reconstruction pipeline) is at
**<https://huggingface.co/datasets/RyoMurakami/OpenH-RF-eSAF>**. This archive
preserves the raw, upstream acquisition/simulation outputs that were converted into
that packaged submission — released under the same **CC BY 4.0** license (see
`LICENSE`).

Code and full documentation for how this raw data was collected and processed
live in the GitHub repo `RyoMurakami-MedRobo/OpenH-RF-eSAF`
(`experiment/`, `sim/`, `matlab/saf/`).

## What's here

| File | Contents |
|---|---|
| `raw/experiment_10mm.zip` | Raw Verasonics per-line channel-RF capture (`RFdata_line*.mat`, one file per acquired line) for the 10 mm nominal-depth measured phantom scan, plus `acq_meta.mat`, `acq_log.mat`, `EncoderInfo/` (motor angle log), and working QC files from this scan folder (depth-identification figures, intermediate eSAF `.mat`/`.png`). This is the data **before** time-tag frame-to-angle synchronization and before centre-plane-wave extraction — i.e. upstream of `experiment/sync_channel_rf.m`, which produces the packaged single-plane-wave-per-angle raw channel data delivered in the public submission as `data/experiment__acq_exp_10mm.hdf5`. |
| `raw/experiment_20mm.zip` / `_30mm` / `_40mm` / `_45mm` | Same, for the 20/30/40/45 mm nominal-depth measured scans (source: `20260617/scan_data_JP_US_Angle_3D_<depth>mm/` on lab storage). |
| `raw/sim_dataset_out.zip` | Per-case MATLAB `.mat` + reference `.png` for all 190 simulated (probe × target) cases from `sim/batch_sim_probe_target.m`: raw per-element channel RF, the in-plane DAS beamformed result, full metadata, and the eSAF-reconstructed output (`safrot_backproj`), plus the `dataset_overview_r4.png` FWHM-vs-depth summary. This is the MATLAB-side counterpart of the zea `.hdf5` files delivered in the public submission. |

## Why this archive exists

The public submission ships the already-converted, packaged zea `.hdf5` format
(raw channel RF + scan metadata + paired SAF-volume label) needed by OpenH-RF. This
archive preserves the raw, upstream acquisition/simulation outputs that were
converted into that packaged form, in case of a need to re-derive, re-verify, or
re-process from source (different frame-to-angle sync window, different plane-wave
selection, alternate SAF parameters, etc.) beyond what the packaged form supports.

## Reproducing the packaged data from this archive

1. Simulated: `sim/mat_to_zea.py` or `sim/sim_dataset_to_zea.py` converts the
   `sim_dataset_out/<probe>/<target>.mat` files here into the zea `data/*.hdf5`
   format in the public submission.
2. Measured: `experiment/sync_channel_rf.m` reads the raw per-line `RFdata_line*.mat`
   + `EncoderInfo/` here, performs time-tag frame→angle synchronization and
   centre-plane-wave extraction, and writes the packaged channel-RF used by
   `sim/sim_dataset_to_zea.py` to produce `data/experiment__acq_exp_<depth>mm.hdf5`.

## Known caveats

- Not all raw acquisitions collected during this project are archived here — only
  the ones that back the 5 measured acquisitions actually delivered in the public
  submission (10/20/30/40/45 mm). Earlier troubleshooting captures (50–130 mm and
  several re-acquisition variants used to diagnose a depth-dependent eSAF focus
  issue) are not included; they remain on lab storage if needed later.
- Zips use standard DEFLATE compression (`zip -9`); raw ultrasound RF data
  compresses modestly (~45–55% of original size), not dramatically — it isn't
  incompressible, but don't expect an order-of-magnitude reduction.