Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Sliding Gaussian ball adaptive growth (SlingBAG): point cloud-based iterative algorithm for large-scale 3D photoacoustic imaging

Shuang Li, Yibing Wang, Jian Gao, Chulhong Kim, Seongwook Choi, Yu Zhang, Qian Chen, Yao Yao, Changhui Li

TL;DR

A point cloud-based IR algorithm that reduces memory consumption by several orders, wherein the 3D PA scene is modeled as a series of Gaussian-distributed spherical sources stored in form of point cloud, which enables high-quality large-scale 3D PA reconstruction with fast iteration and extremely low memory usage.

Abstract

Large-scale 3D photoacoustic (PA) imaging has become increasingly important for both clinical and pre-clinical applications. Limited by cost and system complexity, only systems with sparsely-distributed sensors can be widely implemented, which desires advanced reconstruction algorithms to reduce artifacts. However, high computing memory and time consumption of traditional iterative reconstruction (IR) algorithms is practically unacceptable for large-scale 3D PA imaging. Here, we propose a point cloud-based IR algorithm that reduces memory consumption by several orders, wherein the 3D PA scene is modeled as a series of Gaussian-distributed spherical sources stored in form of point cloud. During the IR process, not only are properties of each Gaussian source, including its peak intensity (initial pressure value), standard deviation (size) and mean (position) continuously optimized, but also each Gaussian source itself adaptively undergoes destroying, splitting, and duplication along the gradient direction. This method, named the sliding Gaussian ball adaptive growth (SlingBAG) algorithm, enables high-quality large-scale 3D PA reconstruction with fast iteration and extremely low memory usage. We validated SlingBAG algorithm in both simulation study and in vivo animal experiments. The source code and data for SlingBAG, along with supplementary materials and demonstration videos, are now available in the following GitHub repository: https://github.com/JaegerCQ/SlingBAG.

Sliding Gaussian ball adaptive growth (SlingBAG): point cloud-based iterative algorithm for large-scale 3D photoacoustic imaging

TL;DR

A point cloud-based IR algorithm that reduces memory consumption by several orders, wherein the 3D PA scene is modeled as a series of Gaussian-distributed spherical sources stored in form of point cloud, which enables high-quality large-scale 3D PA reconstruction with fast iteration and extremely low memory usage.

Abstract

Large-scale 3D photoacoustic (PA) imaging has become increasingly important for both clinical and pre-clinical applications. Limited by cost and system complexity, only systems with sparsely-distributed sensors can be widely implemented, which desires advanced reconstruction algorithms to reduce artifacts. However, high computing memory and time consumption of traditional iterative reconstruction (IR) algorithms is practically unacceptable for large-scale 3D PA imaging. Here, we propose a point cloud-based IR algorithm that reduces memory consumption by several orders, wherein the 3D PA scene is modeled as a series of Gaussian-distributed spherical sources stored in form of point cloud. During the IR process, not only are properties of each Gaussian source, including its peak intensity (initial pressure value), standard deviation (size) and mean (position) continuously optimized, but also each Gaussian source itself adaptively undergoes destroying, splitting, and duplication along the gradient direction. This method, named the sliding Gaussian ball adaptive growth (SlingBAG) algorithm, enables high-quality large-scale 3D PA reconstruction with fast iteration and extremely low memory usage. We validated SlingBAG algorithm in both simulation study and in vivo animal experiments. The source code and data for SlingBAG, along with supplementary materials and demonstration videos, are now available in the following GitHub repository: https://github.com/JaegerCQ/SlingBAG.
Paper Structure (10 sections, 5 equations, 8 figures, 1 table, 1 algorithm)

This paper contains 10 sections, 5 equations, 8 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Results
  3. 3D PA image reconstruction under extremely sparse planar array
  4. 3D PA image reconstruction of in vivo animal studies
  5. Computing memory consumption analysis and comparison
  6. Discussion
  7. Methods
  8. Differentiable rapid radiator based on multi-order spherical discretization of Gaussian-distributed spherical PA source
  9. Sliding Gaussian ball adaptive growth point cloud model
  10. Physically-based point cloud to voxel grid shader

Figures (8)

  • Figure 1: Comparison of 3D photoacoustic reconstruction results under extremely sparse planar arrays. (a) Top-view maximum amplitude projection, front-view maximum amplitude projection, and the cross-section slice at the green dashed line marked in the top-view-MAP of the acoustic source. (b-f) Top-view maximum amplitude projection, front-view maximum amplitude projection, and the cross-section slice at the green dashed line marked in the top-view-MAP of the UBP reconstruction result with 49, 196, 576, 4900 and 122,500 sensors. (g-j) Top-view maximum amplitude projection, front-view maximum amplitude projection, and the cross-section slice at the green dashed line marked in the top-view-MAP of the SlingBAG reconstruction result with 49, 196, 576 and 4900 sensors. (k) Imaging setup. (l) Intensity distribution curves along the yellow dashed line in the vertical coordinate of the slices from the acoustic source, 576-sensor reconstruction, and 4900-sensor reconstruction. (Scale: 10 mm.)
  • Figure 2: Assessment for image quality of 3D photoacoustic reconstruction results under extremely sparse planar arrays. (a) Structural similarity index (SSIM) of the Top-view maximum amplitude projection of the reconstruction results. (b) Structural similarity index (SSIM) of the slice of the reconstruction results. (c) The signal-to-noise ratio (SNR) of the slice of the reconstruction results. (d) The contrast-to-noise ratio (CNR) of the slice of the reconstruction results.
  • Figure 3: Visualization of SlingBAG iterative results. (a) Visualization of initialization point cloud, coarse reconstruction point cloud, fine reconstruction point cloud and the final voxel grid reconstruction result. (b) 3D visualization of point clouds at various stages of the entire reconstruction process, including random initialization, coarse reconstruction stage, and fine reconstruction stage.
  • Figure 4: 3D PA reconstruction of a mouse brain. (a) XY Plane-MAP, XZ Plane-MAP, YZ Plane-MAP and the cross-section slice at green dashed line marked in XY Plane-MAP of the UBP 3D reconstruction results using 1024 sensor signals. (b) XY Plane-MAP, XZ Plane-MAP, YZ Plane-MAP and the cross-section slice at green dashed line marked in XY Plane-MAP of the SlingBAG 3D reconstruction results using 1024 sensor signals. (Scale: 2 mm.) (c) Schematic diagram of the hemispherical array imaging the mouse brain. (d) SlingBAG reconstruction result.
  • Figure 5: 3D PA reconstruction results of a rat kidney. (a) XY Plane-MAP, XZ Plane-MAP, YZ Plane-MAP and the cross-section slice at green dashed line marked in XY Plane-MAP of the UBP 3D reconstruction results using 1024 sensor signals. (b) XY Plane-MAP, XZ Plane-MAP, YZ Plane-MAP and the cross-section slice at green dashed line marked in XY Plane-MAP of the SlingBAG 3D reconstruction results using 1024 sensor signals. (Scale: 2 mm.) (c) Schematic diagram of the imaging area. (d) SlingBAG reconstruction result.
  • ...and 3 more figures