Improving real-time ultrasound spine imaging with a large-aperture array
Data files
Jul 14, 2025 version files 34.24 GB
-
CIRSphantom.zip
1.88 GB
-
humanspine_(2).zip
15.48 GB
-
README.md
6.70 KB
-
spinephantom_(2).zip
15.50 GB
-
spinephantom_(3).zip
1.23 GB
-
stats.zip
429.11 KB
-
VSXsimulation.zip
148.15 MB
Abstract
Ultrasound offers a safe, low-cost alternative to computed tomography (CT) and magnetic resonance imaging (MRI) for spinal diagnostics and intervention by enabling real-time imaging. However, the complex structure of the spine and acoustic shadowing from bones present challenges for ultrasonography. This study addresses these limitations using an 8.8-cm 384-element large-aperture array and full aperture-based imaging protocols. Volumetric scanning across multiple vertebrae was accomplished in 5 seconds using ultrafast, diverging wave acquisition. In 7 healthy volunteers, the large-aperture array and diverging wave transmission improved resolution, contrast, and visualization of the spinal canal, venous plexuses, and facet joints compared with conventional probes. A comparison between the co-registered CT and ultrasound scan confirmed the imaging accuracy. A simulated lumbar puncture demonstrated needle tip visualization throughout the trajectory into the spinal canal. The results suggest that large-aperture arrays, coupled with plane or diverging protocols, are a valuable tool for spine imaging and image-guided intervention.
Dataset DOI: 10.5061/dryad.9zw3r22sx
Description of the data and file structure
The attached dataset includes raw data, processed data, and MATLAB scripts that are used for data processing for the paper entitled "Improving real-time ultrasound spine imaging with a large-aperture array".
Overview of the raw data collection: Phantom data were collected with a standard ultrasound tissue-mimicking phantom (CIRS Model 040GSE) and a house-made tissue-mimicking spine phantom as described in the paper. Human spine images were collected from seven healthy human volunteers. All the human ultrasound imaging was performed following the protocol approved by the Institutional Review Board. Informed written consent was obtained from all 7 volunteers involved in human spine imaging.
Please see the details below for each zip file.
Files and variables
File: stats.zip
Description: Original data and statistical analysis in Prism Graphpad. The ANOVA tests correspond to results presented in Fig. 4D. The Bland-Altman tests correspond to results presented in Fig. 5C and Fig. 6C.
File: CIRSphantom.zip
Description: Standard ultrasound phantom data and the MATLAB scripts used for data processing.
Data description:
- Humimicaberration:
- Beamformed ultrasound data acquired with the CIRS phantom and the house-made humimic aberrator are named as "BFdata_acqTX_cdB.mat". For example, "BFdata_acq001_45TX_c1540_75dB_ver18.mat" refers to acquisition#001, which uses 45 transmission, and a sound speed of 1540m/s and a dynamic range of 75 dB for reconstruction.
- The transmission type is denoted in the file name of workspace.mat file, where all ultrasound parameters associated with transmission sequence and reconstruction methods are saved.
- workspace_PW_1.mat: parameter file for Acquisition #001 using plane wave (PW) transmission.
- workspace_DW_2.mat: parameter file for Acquisition #002 using diverging wave (DW) transmission.
- workspace_CW80_3.mat: parameter file for Acquisition #003 using converging wave (CW, i.e., focused beam) transmission with a focal depth of 80mm.
- workspace_CW120_4.mat: parameter file for Acquisition #004 using converging wave (CW, i.e., focused beam) transmission with a focal depth of 120mm.
- workspace_CW160_5.mat: parameter file for Acquisition #005 using converging wave (CW, i.e., focused beam) transmission with a focal depth of 160mm.
- noaberration:
- Beamformed ultrasound data acquired with the CIRS phantom. The file name follows the same fashion as those in the "humimicaberration" folder. More details of the data collection can be found in the text file in each subfolder.
MATLAB scripts description:
- analyze_results_aber.m: script that was used to measure lateral resolution as a function of depth on the CIRS phantom with aberration.
- analyze_results_noaber.m: script that was used to measure lateral resolution as a function of depth on the CIRS phantom without aberration.
- gcnr_calc.m: MATLAB function that was used to calculate generalized contrast-to-noise ratio (gCNR) as part of image quality analysis.
- .fig files: example figures produced by the MATLAB scripts.
File: VSXsimulation.zip
Description: Transmission sequence visualization and comparison using the simulated data from the Verasonics software package.
- DW_vs_ryln_fieldhomogeneity.m: Load simulated transmission pattern of diverging waves (DW) versus ray-line (RyLn) method and evaluate the field homogeneity in each case.
- fieldhomogeneity.m: Load simulated transmission pattern of diverging waves (DW), plane waves (PW), and converging waves (CW), and evaluate the transmission field homogeneity in each case.
- sim_VLA_45Angle_60degree_TXmode-1_radius-100mm_matchFOV.mat: Simulation data generated by the Verasonics software package using 45 diverging waves with virtual source depth at 100mm and a steering range of +/- 30 degrees.
- sim_VLA_45Angle_60degree_TXmode0_radius0mm_matchFOV.mat: Simulation data generated by the Verasonics software package using 45 plane waves with a steering range of +/- 30 degrees.
- sim_VLA_45Angle_60degree_TXmode1_radius80mm_matchFOV.mat: Simulation data generated by the Verasonics software package using 45 converging waves with a focal depth of 80mm and a steering range of +/- 30 degrees.
File: humanspine_(2).zip
Description: Original and processed data of human spine images, along with the MATLAB scripts used for processing.
- CF_DAS_comparison: Beamformed ultrasound data and processing script in MATLAB for comparison of the linear delay-and-sum (DAS) beamformer versus the nonlinear coherence-factor (CF)-based DAS beamformer, corresponding to results presented in Fig. 4 and Table. S2.
- US_CTcomparison: Processed ultrasound data and CT data from volunteer#2 for ultrasound-CT co-registration analysis, as well as the 3D rendering in 3D Slicer. This corresponds to results presented in Fig.5 and Fig. S4.
- volumnscans: Beamformed ultrasound data from a human volunteer volumetric scan. This corresponds to results presented in Fig. S5 and the supplementary movies.
- volunteerimages: Beamformed ultrasound data from a human volunteer single-frame scan. This corresponds to results presented in Fig. 2, Fig. S2, and Fig. S3.
File: spinephantom_(3).zip and spinephantom_(2).zip
Description: Original and processed data of the spine phantom, along with the MATLAB scripts used for processing. This corresponds to results presented in Fig. 6 and Fig.7. Reconstructed ultrasound volume data are named as "Vol_CFDAS_scan_DW10mm_lateraldeg_centerallframes.mat". These files can be directly imported into the MATLAB workspace for visualization of the ultrasound volumetric scan. "SpineSegment.m" is the MATLAB script for segmenting the bony structure of the spine from the ultrasound volumetric data. Segmented bony structures are also provided in the MATLAB data format. Detailed instructions on how to use the MATLAB scripts can be found in each script.
Code/software
- MATLAB (2020b or later)
- Prism Graphpad
- 3D Slicer
Human subjects data
The human subject data (ultrasound images of lumbar and cervical spine from 7 healthy human volunteers) were only collected to analyze image quality. No information related to human subject identification were collected except the gender, age, and race. All participants explicitly consented to publish the de-identified data in the public domain.