Reconstruction Algorithm and Post-Processing for 3D Colloidal Assemblies
This repository features the CS-DART (Compressed Sensing Discrete Algebraic Reconstruction Technique) algorithm and associated post-processing codes. These are central to the research presented in our submission to Nature Communications. The details are as follows:
@article{arenas2023liquid,
title={Liquid phase fast electron tomography unravels the true 3D structure of colloidal assemblies},
author={Arenas Esteban, Daniel and Wang, Da and Kadu, Ajinkya and Olluyn, Noa and S{\'a}nchez Iglesias, Ana and Gomez Perez, Alejandro and Gonzalez Casablanca, Jesus and Nicolopoulos, Stavros and Liz-Marz{\'a}n, Luis M and Bals, Sara},
journal={arXiv e-prints},
pages={arXiv--2311},
year={2023}
}
CS-DART combines compressed sensing with discrete tomographic principles to improve the 3D reconstruction of colloidal particles in liquid environment under HAADF-STEM imaging, despite the high electron dose sensitivity and significant missing wedge issues faced due to liquid-cell holder.
Feature | Description |
---|---|
Advanced 3D Reconstruction | Enhances the resolution and detail of 3D structures in liquid conditions. |
Complex Structures | Reveals intricate geometries such as tetrahedrals and polyhedra for particle counts N = 4, 5, 6. |
Quantitative Analysis | Employs alpha shapes for detailed morphological and spatial metrics. |
Experimental Validation | Proven against both synthetic and experimental datasets in demanding liquid environments. |
- MATLAB: Tested with MATLAB R2023a.
- Operating Systems: Windows 10.
- Image Processing Toolbox
- Statistics and Machine Learning Toolbox
- Computer Vision Toolbox (optional for some functions)
Component | Specification |
---|---|
RAM | 32 GB Minimum |
CPU | Intel(R) Core(TM) i7-8700 @ 3.20GHz or equivalent |
GPU | Nvidia RTX 2070, 8 GB |
- ASTRA-Toolbox v2.1.0: A framework for tomographic operators. More Info
- SPOT: Simplifies modeling of linear operators in tomography. More Info
- MinConf Optimization Package: Optimizes objectives including CS-DART. More Info
-
Clone the Repository
git clone https://github.com/ajinkyakadu/LiquidET_NatComm2024.git cd LiquidET_NatComm2024
-
Set Up MATLAB Environment
run('setup.m');
- Typical setup time is under 1 minute on a standard desktop.
-
Prepare the Environment
cd examples
-
Run Reconstruction Script
ex01_step01_N4Liquid
-
Execute Post-Processing
ex01_step02_N4Liquid
- Reconstructed Volume: Saved as
csdart_reconstructed_volume.rec
in thedata
folder. - Quantitative Descriptors: Stored as
quant_descriptors_NP.mat
in thedata
folder.
- Approximately 30 minutes on a standard desktop.
Access the dataset for different colloidal systems (N = 4, 5, 6) at Zenodo. Zenodo DOI: 10.5281/zenodo.11175299
To adapt the software for your data:
-
Data Preparation Ensure your data is formatted correctly and placed in the
data
directory. -
Script Adjustments
- Modify
ex01_step01_N4Liquid.m
to reference your dataset and adjust parameters for CS-DART reconstruction scheme. - Adjust parameters like
cropRadius
andminArea
inex01_step02_N4Liquid.m
to fit your data.
- Modify
-
Execute the Analysis Follow the demo steps with your data specifics.