AutoDeconJ: a GPU-accelerated ImageJ plugin for 3D light-field deconvolution with optimal iteration numbers predicting

Changqing Su, Yuhan Gao, You Zhou, Yaoqi Sun, Chenggang Yan, Haibing Yin, Bo Xiong

Light-field microscopy (LFM) is a compact solution to high-speed 3D fluorescence imaging. Usually, we need to do 3D deconvolution to the captured raw data. Although there are deep neural network methods that can accelerate the reconstruction process, the model is not universally applicable for all system parameters. Here, we develop AutoDeconJ, a GPU-accelerated ImageJ plugin for 4.4× faster and more accurate deconvolution of LFM data. We further propose an image quality metric for the deconvolution process, aiding in automatically determining the optimal number of iterations with higher reconstruction accuracy and fewer artifacts.

Installation

Cloning the repository

Create a local clone of this project by calling

git clone https://github.com/Onetism/AutoDeconJ.git

Build PTX

AutoDeconJ requires the NVIDIA cards support by CUDA8.0 or later. See https://developer.nvidia.com for more details about CUDA. Please make sure that CUDA is properly installed and the path enviroment is configured correctly. In order to build the ptx file for java corresponding to the version of CUDA, change into the root directory of the project and execute

On linux

./build_cuda.sh

On Windows

./build_cuda.bat

Building the plugin JAR and install in ImageJ

The cloned project can be opened and edited in any IDE (e.g. Eclipse, NetBeans, IntelliJ...). However, the preferred way to build the final plugin JAR is via Apache Maven.

In order to build the plugin JAR, change into the root directory of the project and execute

mvn clean package

Note that the resulting JAR file has a name that is different from the default name that Maven would assign to it: In order to properly be recognized as an ImageJ plugin JAR, it is named AutoDeconJ_Plugin-jar-with-dependencies.jar. Copy the resulting /target/AutoDeconJ_Plugin-jar-with-dependencies.jar file into the Fiji.app/plugins directory of your Fiji/ImageJ installation. Of course, AutoDeconJ_Plugin-jar-with-dependencies.jar is also a execution file, you can execute directly.

Using the plugin

After the JAR files for the plugin have been added, it may be used inside ImageJ: Start ImageJ, load an image, and select

Plugins>AutoDeconJ, "Run Auto_LF_Deconvolution...", Auto_LF_Deconvolution("run")
Plugins>AutoDeconJ, "Run ImageRectification...", ImageRectification("run")

from the menu bar.

Test Data

There is a test images in /test/resources, corresponding to the default parameters in the program: objective magnification 40, NA 0.8, microlens pitch size 150 um, microlens focal length 3000um, wavelength 520um, n 1.0, OSR 3, Nnum 15, z-spacing 2um, z-max 0um, and z-min -26um.

Citation

@article{AutoDeconJ,
    author = {Su, Changqing and Gao, Yuhan and Zhou, You and Sun, Yaoqi and Yan, Chenggang and Yin, Haibing and Xiong, Bo},
    title = "{AutoDeconJ: a GPU-accelerated ImageJ plugin for 3D light-field deconvolution with optimal iteration numbers predicting}",
    journal = {Bioinformatics},
    year = {2022},
    month = {11},
    issn = {1367-4803},
}