ZOLA-3D

ZOLA-3D (Zernike Optimized Localization Approach) is a full software package to reconstruct 3D single molecule localization images for a wide range of point spread functions. Here we provide a CUDA-enabled 3D localization plugin for ImageJ/Fiji software.

Publication: https://www.nature.com/articles/s41467-018-04709-4

A brief introduction of ZOLA-3D is given here. Please, open the User manual - HOW TO USE ZOLA.pdf file for an extended explanation of how to use ZOLA.

Example of reconstructed images can be downloaded HERE. Test data can be downloaded HERE.

License

GPLV3

The plugin is distributed under the license GPLV3 on an "as is" basis, without warranties or conditions of any kind.

Requirements

To enable fast processing on GPU, Nvidia CUDA must be installed and included in the path. We recommand the use of CUDA 10.0 that is compatible with our tools. If you encounter problems with GPU installation or if you need to make ZOLA compatible with your cuda version, please, send us an email (benoit.lelandais@pasteur.fr).

Check your version by executing nvcc --version in the terminal.

Follow instructions on https://developer.nvidia.com/cuda-toolkit to install CUDA

The plugin was tested on Ubuntu, Windows and Mac with Cuda 8.

Plugin Installation

The easiest way to install ZOLA-3D is to use the Fiji update system. The installation requires roughly 15 minutes. If you plan to benefit from GPU acceleration, plugin installation requires 15 minutes more.

Launch Fiji
Select Help -> Update
Select Manage update sites
Select IMOD-ZOLA from the list
Click Close
Click Apply changes
Restart Fiji
ZOLA should now appear in the Fiji->Plugins dropdown menu

If you prefer to use ImageJ instead, you can download ZOLA_-0.1.X-SNAPSHOT.jar plugin in Releases section.

If you want to benefit from GPU acceleration, lib/* and jar/* folders containing jcuda bindings should be copied into Fiji/ImageJ folder. We provide these files in Releases section for cuda 7.5, 8 and 9.

Usage

The full manual can be found HERE and test data can be downloaded HERE

For optimal performance, your images should be in photon counts, as usually the case with recent EMCCD cameras. images are not in photon counts, you can convert them as explained in User manual - HOW TO USE ZOLA.pdf .
Calibrate your PSF from a z-stack of fluorescent beads using Zernike polynomials (this is ideal for PSFs shaped using a cylindrical lens or deformable mirror). We recommend using z-steps of 50 or 100 nm. Open your stack, select central position of one or few beads with the Fiji Point selector tool and run ZOLA -> Calibration: PSF modeling Choose 28-45 Zernike modes.
Now that the PSF is calibrated, you can use ZOLA for 3D single molecule localization based super-resolution microscopy. Open your stack with single molecule blinking images, select region of interest and run ZOLA -> Localization.
The single molecule Localization table is saved in csv, Thunderstorm format. You can correct the drift in 3D and render 3D color-stack using ZOLA->Drift correction->3D Drift correction or any other software.

Example of 3D reconstruction of immunolabeled mitochondrial protein TOM22:

Image reconstructed in 2h using Tesla K40 GPU:


raw image sequence with blinking tetrapod PSFs	3D scatter-plot reconstruction (Visp rendering)

For detailed instructions please follow step by step guide on how to run PSF calibration and localization

Distinctive features:

User-friendly interface implemented in Fiji.
Ability to reconstruct 3D single molecule images with variable axial range (e.g. 1-5 micrometers).
Ability to model a wide range of spatially extended PSFs (including astigmatism, saddle point and tetrapod) and to detect and localize single molecules using cross-correlation and maximum likelihood algorithms.
Realistic and robust calibration of the PSF using only one or a few beads thanks to a Zernike-based PSF model. Depth-dependent aberrations due to refractive index mismatch are automatically handled using beads on the coverslip only.
All calculations can run on GPU. Main features also run on CPU, but processing on GPUs is much faster.

Source code compilation

Requirement: Java Development Kit version 8 Requirement: Cuda ToolKit version 8.0 (version 9 not yet available with maven)

Clone or download ZOLA-3D (see link on the top of the page)

To compile the source code, you need for instance Eclipse or NetBeans editors.

For Eclipse, ZOLA project can be imported using the following command: File -> Import... -> Existing Maven Project -> select "zola_source_code" folder

For NetBeans, ZOLA project can be imported using the following command: File -> Open Project -> select "zola_source_code" folder

Then, do: Run-> Build Project

In zola_sourcecode/target folder, "ZOLA-0.1.X-SNAPSHOT.jar" file is created. You can copy this file and paste it in your ImageJ/FIJI plugin folder to use the plugin.

The GPU-Cuda source code is available in src/main/ressources/kernels/ folder. In this folder, check the "ZOLA_CudaCode_ReadMe_file" to compile the file (note that compilation using nvcc can be performed only with gcc-4.8 currently).

imodpasteur / ZOLA-3D

readme