Close

OptiJ: Open-source optical projection tomography of large organ samples

Pedro P. Vallejo Ramirez, Joseph Zammit, Oliver Vanderpoorten, Fergus Riche, François-Xavier Blé, Xiao-Hong Zhou, Bogdan Spiridon, Christopher Valentine, Simeon P. Spasov, Pelumi W. Oluwasanya, Gemma Goodfellow, Marcus J. Fantham, Omid Siddiqui, Farah Alimagham, Miranda Robbins, Andrew Stretton, Dimitrios Simatos, Oliver Hadeler, Eric J. Rees, Florian Ströhl, Romain F. Laine, Clemens F. Kaminski

Preprint posted on 2 June 2019 https://www.biorxiv.org/content/10.1101/656488v1

Article now published in Scientific Reports at http://dx.doi.org/10.1038/s41598-019-52065-0

A tool for all- OptiJ: an easy-to-assemble setup and freely available software as an incentive for widescale implementation of OPT.

Selected by Mariana De Niz

Categories: bioengineering, biophysics

Background

Optical microscopy has been fundamental for biological discovery, however, detailed imaging within tissues is limited to a few micrometers of depth due to photon scattering (reviewed in (1)). In transparent tissues, light propagates like X rays do, so that tomographic approaches can be used for volumetric image reconstruction. In 2002, a revolutionary method was developed by Sharpe et al (2), namely optical projection tomography (OPT), to produce high resolution 3D images of fluorescent and non-fluorescent, optically cleared, biological specimens. OPT enables visualization of complete biological samples relying solely on optical sectioning. Since the publication of this method in 2002, the application of OPT within biology has expanded greatly, allowing the study of multiple processes within specimens including zebrafish, embryos, and multiple organs of rodent animals. The advent of multiple optical clearing methods has contributed greatly to the use of OPT in many organs and specimens, making it currently possible to generate and visualize volumetric data from optically transparent mice (3).

Like-wise, the generation of hybrid techniques such as OPTiSPIM, which combines selective plane illumination microscopy (SPIM) and OPT (4), has expanded the types of questions that can be addressed in multiple disciplines in biology.

While substantial technical work has been done to foster the use of these methods, a limiting factor for many labs to implement this technique is the cost and the availability of specialized software and hardware. For SPIM, this challenge was addressed by the creation of OpenSPIM in 2013 (5), a platform dedicated to making light sheet microscopy widely accessible. Furthermore, the rationale was also to allow higher throughput by making the setups more affordable, providing the possibility of using arrays to do parallel imaging. Furthermore, OpenSPIM encouraged the creation of open source solutions for SPIM image processing (6).  With OptiJ, the authors address in a similar manner, the challenge of affordability and accessibility of OPT, and promote the use and further development of OPT for research by the scientific community.

Key findings and developments

  • The authors have developed OptiJ, a low-cost, open-source hardware and software implementation of OPT for investigation of whole organs in 3D at near cellular resolution (Figure 1).
  • This platform overcomes a key challenge: that most OPT applications require advanced technical expertise, and expensive software/hardware.
  • The overall goal was to develop high-end technology that incentivises further deployment and development of OPT by the research community.
Figure 1. Left panel: a) OptiJ framework showing the workflow from sample acquisition, hardware: assembly at stage, and software: calibration and reconstruction algorithms. b) Picture of the OptiJ hardware setup including the rotation and translation stages, the magnetic mount, the sample chamber, the LEDs, and camera. c) Schematic of top view of OptiJ hardware. Right panel: reconstructions from murine lungs in xy, and 3D projections showing secondary and tertiary bronchi.

Hardware

  • The OPT principle relies on the rotation of a sample to acquire 2D projections at different angles. Axis of rotation and alignment were main considerations for the setup design.
  • Other main criteria guiding the component choice were: ease of access, widespread availability and low cost. With this in mind, the components developed for the setup are:
  • A monolithic 3D-printed stageadapted from the Flexscope design, to accomplish the movement necessary for linear alignment and rotation of the sample with low cost motors that achieve sub-micron steps.
  • A telecentric relay lens, with low numerical aperture. The telecentricity of the lens is key for an efficient filtered back projection (FBP) reconstruction approach afterwards.
  • A camera. In this preprint, an Andor CLARA camera with 6.45 x 6.45 um2pixels was used, however, lower cost cameras can be used instead.
  • Two broadband LEDs, emitting over a broad spectral range. A custom circuit board was designed to minimize output flicker.
  • Fluorescence excitation and emission filters.
  • Collimating and diffusing optics to ensure uniform illumination.

 

Software

  • OptiJ includes a set of freely available ImageJ/Fiji plugins to process OPT data. This set of plugins is offered as an improvement over the Radon Transform (currently available in Fiji/ImageJ), as it includes calibration and accelerated reconstruction algorithms.
  • Main plugins developed by the authors include:

A) Calibration plugins

  • The Beer-Lambert correction plugin, which divides each transmission OPT projection (from absorption or scattering from the sample) by average brightfield image, to obtain linear attenuation coefficients corrected for non-uniform pixel intensities.
  • The Estimate Tilt and Offset plugin,which tracks a fiducial marker in the projections to determine if the axis of rotation is parallel to the centre of the FOV, and produces correction values for the projection stack if not.
  • The Create Sinogram plugin, which displays a Radon Transform of the projections, accounting for tilt and offset corrections. It relaxes the need for precise alignment prior to acquisitions. The output of this plugin is a sinogram.
  • The dynamic offset correction plugin, which calculates a sinusoidal fit of the motion of a fiducial marker and uses the difference between the ideal fit coordinates and the actual motion in the corrected image.

B) A reconstruction plugin:

  • The 2D reconstruction pluginimplements a Filtered Back-Projection algorithm to reconstruct a 3D cross-sectional stack of the original object. The plugin allows for GPU-enabled acceleration using OpenCL. This downscales the processing time from hours, to minutes.
  • Implementation for the first time of the Fourier Ring Correlation (7) as a resolution measure for OPT datasets.

Proof of principle

  • OptiJ allowed 3D exploration of the tertiary airways, bronchioles, and alveolar sacs in complete murine lungs, in the context of COPD.

What I like about this paper

  • a) The main aim of the paper, which is to make science widely available to everyone. This is a philosophy that could permeate across many disciplines and techniques so that cost is not a limiting factor for answering research questions with the best possible technologies.
  • b) That the protocols are carefully described in enough detail to allow consistent reproducibility. They consider various possible artefacts and account for them both in hardware and software.
  • c) The origin of this project. It is very consistent with a philosophy of open and inclusive science, as well as fostering independence among young scientists since their early career. The work performed by the authors is the product of collaborative effort of the 14 PhD students who drove the project, as part of the programme EPSRC Centre for Doctoral Training in Sensor Technologies and Applications.

 

Open questions

  1. You have used OptiJ to image lungs as a proof of principle on the application of the setup. Previous publications have shown that various organs differ not only in optical clearance time, but also in the final clearance achievable. Do you envisage that the hardware and software setups you have implemented to calibrate the system will be equally optimal for all tissues and specimens? Under these lines, do you envisage that different mounting procedures, or optical clearance methods will influence the performance of the system?

 

  1. Given the creation of the OPTiSPIM hybrid, and the existence of OpenSPIM, is a joined low-cost platform, something you would envisage?

 

  1. Regarding the Radon Transform plugin, what are the major advances brought by OptiJ?

 

  1. You developed the Dynamic Offset correction plugin, and mention its usefulness due to the possible jitter introduced by low-cost stepper motors used for sample rotation. Can you expand further on the advantages of this method and how it works?

 

  1. You introduced the Fourier Ring Correlation in OPT. Can you expand briefly for the general readership, on how FRC has been used in various microscopy settings, as a measure of resolution without dependence on a priori calibration, and why this is novel and useful in OPT?

 

  1. For SPIM, OpenSPIM Wiki was the chosen platform to disseminate the know-how of the technique in various respects. How will you best envisage reaching your goal of a widespread access to OptiJ in the future?

References

  1. Ntziachristos V, Going deeper than microscopy: the optical imaging frontier in biology, Nature Methods, 2010, 7(8): 603-614.
  2. Sharpe J, Ahlgren U, Perry P, Hill B, Ross A, Hecksher-Sorensen J, Baldock R, Davidson D, Optical Projection tomography as a tool for 3D microscopy and gene expression studies, Science, 2002, 296(5567): 541-545
  3. Yang B, Treweek JB, Kulkarni RP, Deverman BE, Chen CK, Lubeck E, Shah S, Cai L, Gradinaru V, Single cell phenotyping within transparent intact tissue through whole body clearing, Cell, 2014, 158(4): 945-958
  4. Mayer J, Robert-Moreno A, Danuser R, Stein JV, Sharpe J, Swoger J, OPTiSPIM: integrating optical projection tomography in light sheet microscopy extends specimen characterization to non-fluorescent contrasts, Opt Lett., 2014, 39(4): 1053-1056
  5. Pitrone PG, Schindelin J, Stuyvenberg L, Preibisch S, Weber M, Eliceiri KW, Huisken J, Tomancak P, OpenSPIM: an open-access light-sheet microscopy platform, Nature Methods, 2013, 10(7): 598-599
  6. Schmied C, Stamataki E, Tomancak P, Open-source solutions for SPIMage processing, Methods Cell Biol, 2014, 123:505-529
  7. Banterle N, Bui KH, Lemke EA, Beck M, Fourier ring correlation as a resolution criterion for super-resolution microscopy, Struct Biol, 2013, 183(3): 363-367

 

Posted on: 14 June 2019 , updated on: 24 June 2019

doi: https://doi.org/10.1242/prelights.11202

Read preprint (No Ratings Yet)

Author's response

Pedro Vallejo Ramirez and Romain Laine shared

1. You have used OptiJ to image lungs as a proof of principle on the application of the setup. Previous publications have shown that various organs differ not only in optical clearance time, but also in the final clearance achievable. Do you envisage that the hardware and software setups you have implemented to calibrate the system will be equally optimal for all tissues and specimens? Under these lines, do you envisage that different mounting procedures, or optical clearance methods will influence the performance of the system?

The lens and sensor combination we used was optimized for the adult mouse lungs we examined so in principle any organ sample of a similar scale can be scanned with the OptiJ set up. For a smaller sample, a higher NA objective can be used to boost the resolution (trading for a smaller field of view). Since OPT works better in completely cleared samples, a better clearing method (3DISCO or CLARITY) would likely improve the quality of the reconstructions as well.

 

2. Given the creation of the OPTiSPIM hybrid, and the existence of OpenSPIM, is a joined low-cost platform, something you would envisage?

Absolutely! It’s great that there are other open-source platforms, users can choose the best features of each which best suit their applications. We could integrate some of the OptiJ improvements, such as the open-source plugin library or the telecentric objective, into existing implementations.

 

3. Regarding the Radon Transform plugin, what are the major advances brought by OptiJ?

As a plugin in Fiji, it’s very easy to access and to use on tomography data. The slice-by-slice reconstruction of tomography data via filtered back-projection can be a time-consuming process, so the CPU/GPU acceleration offered by the OptiJ plugins allow for quick reconstructions to troubleshoot potential reconstruction artefacts.

 

4. You developed the Dynamic Offset correction plugin, and mention its usefulness due to the possible jitter introduced by low-cost stepper motors used for sample rotation. Can you expand further on the advantages of this method and how it works?

Using low-cost stepper motors without microstepping (which controls the initial and final acceleration of each rotation step, to allow for smoother move-and-stop transitions) leads to jitter in the raw projections, which shows up as jagged edges in the sinogram. These can be avoided by either using more expensive motors, a compensating bearing system between the motor and the sample, or using a sinusoidal fit of the ideal sinogram shape to shift the jagged edges back to the ideal rotation position. This effectively reduces the jitter in the raw projections caused by abrupt movement in the motors.

 

5. You introduced the Fourier Ring Correlation in OPT. Can you expand briefly for the general readership, on how FRC has been used in various microscopy settings, as a measure of resolution without dependence on a priori calibration, and why this is novel and useful in OPT?

FRC is a resolution metric routinely used in electron microscopy reconstructions that measures the similarity of two independent reconstructions of the same object, and determines the highest spatial frequency at which both reconstructions are consistent. It has been recently adopted in super-resolution microscopy (Nieuwenhuizen et al., 2013) as a measure of the smallest reproducible feature of the sample after reconstruction (e.g. in single molecule localisation). We proposed to implement it in OPT to quantify the smallest reproducible detail in a tomographic reconstruction, as an unbiased, complementary metric to measuring cross sections of small features in individual slices of a reconstructed stack. It’s useful as a metric to check how resolution varies in OPT as a function of radius away from the axis of rotation of the sample.

 

6. For SPIM, OpenSPIM Wiki was the chosen platform to disseminate the know-how of the technique in various respects. How will you best envisage reaching your goal of a widespread access to OptiJ in the future?

Initially, we have put all our files in a GitHub repository (https://lag-opt.github.io/) as an easy way to allow contributions from external collaborators. All the components necessary for the building and implementation and use of OptiJ is described either on our GitHub page or in our supplementary file on bRx. But by all means, if anyone is interested in building one and encounters any problems, do get in touch!

 

Have your say

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Sign up to customise the site to your preferences and to receive alerts

Register here
Close