Datasets
Standard Dataset
Freehand 3D ultrafast Doppler reconstruction sample dataset
- Citation Author(s):
- Submitted by:
- Luuk Verhoef
- Last updated:
- Tue, 03/18/2025 - 08:52
- DOI:
- 10.21227/kfvy-s639
- Data Format:
- License:
22 Views- Categories:
- Keywords:
Abstract
Sample dataset of a freehand 3D ultrafast Doppler scan acquired during brain tumor resection surgery. The dataset includes the beamformed ultrasound images, as well as the simultaneously acquired optical tracking data, describing the position of the ultrasound probe in space.
An example of how to reconstruct a 3D volume can be found in the attached python script. This script requires installation of the following python packages: matplotlib, sklearn, nibabel, numpy, scipy, cupy, knee, h5py. Optionally, some functions can leverage computation on a GPU for more computationally intensive steps in the pipeline. These can be set by passing a gpu=True argument.
This dataset accompanies IEEE TMI submission TMI-2024-3272.
The HDF5 file contains the following datasets:
"images/data" - complex ultrasound image data in shape (time, nx, nz)
"images/spacing" - ultrasound image pixel dimensions as (dx, dz)
"images/timestamps" - acquisition timestamps of the ultrasound images in shape (time,)
"tracking/transforms" - 4x4 rigid transforms obtained from the optical tracking system during ultrasound acquisition in shape (time, 4, 4)
"tracking/timestamps" - timestamps corresponding to transform data
