OpenSpin – an open source integrated microscopy platform

Emilio J Gualda, Tiago Vale, Pedro Almada, José A Feijó, Gabriel Gonçalves Martins & Nuno

Moreno

Supplementary figures and text:

Supplementary Figure 1 Micromanager OpenSpinMicroscopy plugin front panel

Supplementary Figure 2 Arduino microcontroller shield for galvo control

Supplementary Figure 3 Schematic of the SPIM/DSLM/OPT setup

Supplementary Figure 4 Schematic of a dedicated OPT setup

Supplementary Figure 5 Light-sheet characterization

Supplementary Figure 6 Multimode SPIM/DSLM images of a Drosophila embryo

Supplementary Table 1 Major parts list of the SPIM/DSLM/OPT setup

Supplementary Note 1 Micromanager and Arduino

Supplementary Note 2 Sample preparation

Supplementary Note 3 Preparation of the OPT dataset and tomographic

reconstruction

Note: Supplementary Videos 1–3 are available on the Nature Methods website.

Nature Methods: doi:10.1038/nmeth.2508

Supplementary Figure 1 | Micromanager OpenSpinMicroscopy plugin front panel

Supplementary Figure 1 | Micromanager OpenSpinMicroscopy plugin front panel

We have designed a custom Java plugin for fully controlling DSLM/SPIM/OPT microscopes

with a single window. It allows the capture of time lapse sequences, multicolor, z stacks and

multi-view recordings in an easy and intuitive manner. The shutter state and the

excitation/emission filters can be also controlled with this plugin. The Stage Control and the

Rotation Control panels help with sample positioning in z and view angle. Two different modes,

i.e. DSLM/SPIM or OPT, can be chosen using the Mode Panel. In OPT mode only rotation and

time lapse are allowed while in SPIM/DSLM mode the different imaging options (Stack of

Images, Channels, Rotation, Time lapse) can be selected through the checkboxes. In the DSLM

panel we are able to define the amplitude and speed of a triangular wave (Move button) or

define a specific position (Set button) of the galvanometric mirrors. The latter option is useful

for alignment purposes, although only positive values are possible. The camera parameters

(exposure time and binning) are controlled with the Micromanager main control, except when

the channel option is selected. Then, it can be edited the exposure time, as well as the channel

name and the excitation/emission filter combination, for each channel.

The plugin and the source code can be found at http://uic.igc.gulbenkian.pt/micro-dslm.htm.

Nature Methods: doi:10.1038/nmeth.2508

Supplementary Figure 2 | Arduino microcontroller shield for galvo control

Supplementary Figure 2 | Arduino microcontroller shield for galvo control

In order to create a light sheet, DSLM version scans a beam over the axis using a galvanometric

mirror. This mirror changes its deflection angle depending on the voltage applied. However

Micromanager does not provide a straightforward way to control galvo devices. We have design

a solution using an Arduino board modifying both, the Micromanager Device Adaptor for

Arduino and the Micromanager firmware loaded on the Arduino board. Instead of the original

Micromanager Device Adaptor, the modified one sends two words (16 bits) one containing the

amplitude information and the other one the frequency codified in a 12 bits base. The Arduino

firmware reads those values, calculates the corresponding triangular wave and then sends the 12

bits of each point in the triangular wave using the modified “analogOut” function in a

continuous loop through a 12 bit digitat-to-analog (DAC) chip (MCP4921) to the galvanometric

mirror. A defined delay between these points controls the wave frequency. In order to exit the

loop to stop the galvos or modify the triangular wave parameters an external signal is needed.

This signal will be connected from pin12 on the Arduino board that controls the sample rotation

to pin2 on the Arduino board that controls the galvanometric mirror.

Arduino provides signals between 0 and 5 V, however galvos are optimally designed to oscillate

around the 0 V with positive and negative voltages. To obtain this behaviour we implemented

the shield circuit shown in the figure. This circuit is based in an OP07 operational amplifier

which is used as an offset regulator. The maximum range of output is ruled by:

,.

The amplification of the input signal can also be controlled, by changing R1 and R2. Part of the

circuit is an inverting amplifier, so the gain can be calculated by

.

In that way we are able to obtain triangular waves with amplitudes up to ±7.5V and frequencies

up to 300 Hz for 200 points and 500 Hz for 100 points.

The Micromanager Device Adaptor and the Arduino firmware code could be checked on:

http://uic.igc.gulbenkian.pt/micro-dslm.htm.

Nature Methods: doi:10.1038/nmeth.2508

Supplementary Figure 3 | Schematic of the SPIM/DSLM/OPT setup

Supplementary Figure 3 | Schematic of the SPIM/DSLM/OPT setup

The SPIM/DSLM microscope can work in two different imaging regimes depending on the

laser (L). The linear mode uses an Argon/Krypton laser providing excitation wavelengths of

488 nm, 568 nm and 647, while the non-linear mode uses a Ti:Sapphire laser with tunable

excitation between 750 and 900 nm. The switch between the two imaging modes is performed

manually through a flip mount mirror. In the case of the linear mode the excitation laser lines

are selected using a filter wheel (FW1) with four different filters. The laser power is controlled

using a varying neutral density filter. A shutter (S) is used to block the laser beam and control

the light dosage applied to the sample for long time lapse recording when is only opened during

image acquisition. In order to create the light sheet on the sample plane we can also choose two

different modalities, DSLM, scanning in the vertical axis the laser beam using a galvanometric

mirror (GM), or classical SPIM, stopping the galvanometer at 0 V and using a 50 mm focal

length cylindrical lens (CL). In both cases the optical plane of the galvanometer is conjugated

with the back focal aperture of the excitation objective (EO) lens using a 3.5x or 8x lens

telescope (LT). In SPIM mode, the cylindrical lens is inserted in the optical path in such a way

that the horizontal axis of the beam is focused on the back aperture of the excitation objective,

while the vertical axis fills the aperture. In DSLM mode the cylindrical lens is replaced with a

hollow tube lens to maintain the distance between optics. The light sheet created either by the

cylindrical lens or the scanning galvo creates fluorescence at the focal plane of detection

objective (DO), which is imaged at the camera (C) by a tube lens (TL). After the objective, the

excitation wavelengths are rejected using different emission filters placed in infinity space

before the camera. Those filters are mounted in a custom made motorized filter wheel (FW2).

Different planes are collected by moving the sample using a translational stage (TS) through the

light sheet. Multi-view images are obtained by rotating the sample with a stepper motor (SM).

This easy and cheap solution allows up to 0.225 degrees steps and is controlled with an Arduino

UNO board (AB2). Sample centering (X/Y axis) on the field of view of the camera is performed

manually with two linear translation stages. The same system can be used for OPT microscopy.

The OPT microscope uses the detection arm of the SPIM/DSLM system with frontal

illumination performed with a LED system for transmitted light OPT and lateral illumination

with a blue LED. Different chambers (SC) have been designed for water immersion and air

detection objectives.

Nature Methods: doi:10.1038/nmeth.2508

Supplementary Figure 4 | Schematic of a dedicated OPT setup

Supplementary Figure 4 | Schematic of a dedicated OPT setup

This dedicated OPT scanner is a much simpler, easier (and inexpensive) to implement setup

than a complete OpenSPIN setup with DSLM/SPIM/OPT. It uses two sources of illumination; a

white transmitted light source (TL), for “see-through” projections. A simple and effective white

light source is a common fluorescent light bulb, and a diffuser glass (di), placed behind the

sample so that light goes through it while imaging. Incident illumination (IL) can be used for

fluorescence excitation, consisting of a blue LED, LED optics to produce a more focused beam

and an excitation filter (bandpass GFP excitation filter or a 500nm short-pass) placed in front of

the LED optics to minimize reflection spots during fluorescence image acquisition. The LED

power can be controlled by the electronics (E2) which consist of a LED driver + potentiometer.

The IL is positioned at an angle of approximately 45º facing the sample frontwards. A sample-

holder assembly (SH), consisting of 3 dovetail rack&pinion travel stages hold the stepper motor

(SM) in position over the sample chamber (SC; =cuvette), and allow XYZ adjustments of motor

position. Connecting the motor to the stage assembly is a kinematic mount (ki) which allows tilt

adjustments. The sample agarose block (containing the specimen) is first glued to a large metal

washer, and then mounted on the stage by connecting the washer to the magnet (ma) attached to

tip of the motor axis. While imaging, the sample block is submerged in the sample chamber

(SC) filled with immersion liquid (BABB). Images are captured by a camera (ca) coupled to the

detection optics (DO) which consist of low powered infinity corrected objectives (0.5x - 4x)

coupled via a tube lens. Because of the infinity corrected optics, it is possible to place an

emission filter (em) in the detection path to minimize reflections (which produce “star-like”

artifacts in the final reconstruction). During acquisition, the motor is controlled via Arduino

(E1) and synchronized with image capture. (See Supplementary Table 1 for detailed part

numbers and the OpenSPIN wiki for detailed instructions on how to assemble the different

parts).

Nature Methods: doi:10.1038/nmeth.2508

Supplementary Figure 5 | Light-sheet characterization

Supplementary Figure 5 | Light-sheet characterization

We performed the measurement of the dimensions of the fluorescent line excited by the light

beam to characterize the imaging modalities available in our DSLM system, linear (a, b) and

non-linear operation (c, d). This characterization was performed by using a Coumarin solution,

the GM was set to be static, a Plan Fluor 4x 0.13 (WD 17.4mm) objective lens was used to

excite the sample and the fluorescent line generated into the sample was imaged on the CCD

camera with a LWD 16x 0.8W (WD 3mm) objective lens. The lateral (e, g) and transversal (f,

h) profiles are also shown. For this work, we use the full-width-at-half-maximum (FWHM) of

the normalized intensity profiles along the axis of interest and centered at the point of

maximum intensity to characterize the dimensions of the different intensity distributions (Δx:

lateral and Δz: transversal).

We have two different configurations on the illumination side using either 8x (a, c) or 3.5x (b,

d) lens telescopes, to expand the laser beam. The first one is composed by a 50 mm and a 175

mm plano-convex lens while the second uses a 25 mm and a 200 mm plano-convex lens. In that

way we obtain different magnifications at the same time that the conjugated planes lie on the

same position. At first glance it is possible to see that when using the nonlinear regime (2p

DSLM) an important decrease of the length of the usable field of view happens compared with

linear DSLM due to the confined nature of the nonlinear excitation. Nonetheless, for the linear

case a considerable amount of background is added from the fluorescence excited outside the

Rayleigh range of the beam. As expected the 8x configuration, that overfill the objective back

focal aperture, provides a sharper transversal section at the focus in both modes, linear

(FWHM: 3.7 µm) and non-linear (3.8 µm) compared with the 3.5x configuration (with a

transversal section of 6.6 and 5.4 µm, respectively). However for linear mode the 8x light sheet

thickness expands faster at the edges of the field of view than the 3.5x one. Normally a

compromise is needed between sectioning sharpness and field of view. For that reason, although

less accurate in z, the 3.5x configuration is better suited for big samples using the maximum

field of view available. Scale bar: 100 µm.

Nature Methods: doi:10.1038/nmeth.2508

Supplementary Figure 6 | Multimode SPIM/DSLM images of a Drosophila embryo

Supplementary Figure 6 | Multimode SPIM/DSLM images of a Drosophila embryo

We tested the performance of the three different light sheet microscopy (LSM) techniques

available in our system and its multichannel acquisition capabilities imaging Drosophila

melanogaster fly embryos (w P{w+ asl:YFP};;G147 His:RFP/TM3). Fig. 5 shows a plane 90

µm deep in (a, b) SPIM mode, (c, d) linear DSLM mode (YFP and RFP channels respectively)

and (e) two-photon DSLM mode (yellow channel). Finally in (f) we show a merged image of

the two channels (YFP in green and RFP in magenta) in linear DSLM mode. Comparing the

SPIM and linear DSLM mode we observe the same biological features. However, as expected,

the signal is much brighter in DSLM mode for the same excitation power, since all the laser

intensity is constrained to a line, instead of it been distributed in a whole plane as in SPIM

mode. That makes a more efficient use of the illumination and will prevent for sample damage

since less light is needed. Two-photon DSLM, on the other hand, shows much lower

autofluorescence background, increasing the penetration depth, and brighter signal in the

ectoderm although the overall image intensity is lower than both linear imaging modes.

Moreover, as shown in Fig. 4, the fluorescent excited area is smaller (148 µm FWHM),

providing a non-uniform illumination of the sample. This problem can be solved with the use of

two side illumination or Bessel beams1. Scale bar: 100 µm.

Nature Methods: doi:10.1038/nmeth.2508

Supplementary Table 1 | Major parts list of the SPIM/DSLM/OPT setup

Parts in Figure 3 Vendor Part Number

Lasers L Melles Griot

Coherent Inc.

35 LTL 835-230 (Argon/Kryton laser)

Milllenia (Ti:Sapphire laser)

LED +driver +optics*

+

Excitation filter*

eg LuxeonStar LED

Thorlabs

MR-B0040-20S + A011-D-V-700

+ FLP-N4-RE-HRF*

MF469-35*

Shutter S Uniblitz electronics LS3T2

Objective EO Nikon Plan Fluor 4x 0.13 WD 17.4mm

Plan Fluor 10x 0.30 WD 16 mm

Objectives DO

Nikon

Infinity*

Plan Fluor 4x 0.13 WD 17.4mm

LWD 16x 0.8W WD 3mm

InfiniTube + 0.5x/1x/2x IF series lenses*

Galvanometer GM Cambridge

Technologies 6210H

Translational Stage

TS Thorlabs

MTS50A-Z8 with servo controller

TDC001 (TH)

Stepper Motors SM Astrosyn 9598642

Camera C Hamamatsu Orca-Flash4

Galvanometer AB1,

Stepper Motor AB2

and

Filter Wheel

controllers AB3

Arduino UNO

Excitation Filter

Wheel FW1

Custom made with

Chroma filters

Filters D488/10, 568/10, 488/568DBX

and D647/10

Emission Filter

Wheel FW2

Custom made with

Chroma filters

Filters HQ 405/30m-2p, HQ 430/50m-2p,

ET 480/40m-2p, HQ 535/70m-2p, HQ

580/25m-2p, HQ 620/90m-2p and HQ

640/25m-2p.

LP520*

Lens Telescope LT Thorlabs

3.5x: 50mm + 175mm plano-convex

lenses

8x: 25mm + 200mm plano-convex lenses

Cylindrical lens CL Thorlabs 50mm plano-convex cyl lens

(LJ1695RM)

Tube lens TL

Thorlabs

Infinity*

200mm Best Form Lens (LBF254-200-A)

Infinitube standard system (#210200)*

Sample chamber SC Custom made or

Hellma*

704.003-OG*

* Components used exclusively for acquisition of OPT datasets.

A more detailed list of the parts of the system can be found on our wiki webpage.

Nature Methods: doi:10.1038/nmeth.2508

Supplementary Note 1 | Micromanager and Arduino

The aim of this job is the design of new open source interfaces, plugins and hardware control

that make possible the use of the SPIM/DSLM/OPT microscopy in an cheap easy manner,

helping bringing this technology to any imaging facility with minimum technical capabilities

and adapting it to the user’s needs. All the image acquisition and image analysis is based on free

software such as Micromanager and open source hardware such as Arduino microcontrollers.

Micromanager is an open source general purpose microscope control software built on the

ImageJ framework. It allows the direct configuration and control of regular commercial and

custom made microscope setups with a great variety of peripheral devices (camera, stage, focus

control, filter wheels, etc.). Any of those devices needs a specially designed driver, called

device adaptor, to perform the communication between Micromanager and the device.

Micromanager also allows the use of macros, scripts and plugins to automate measurements or

to create custom user interfaces. Arduino is a popular, open-source hardware prototyping board

with an ATmega328 microcontroller. This low-cost (less than 20 $) programmable digital IO

board was primarily designed to extend the use of electronics to areas beyond the ones used to

work with this kind of technologies, providing six analogue input pins and fourteen digital

input/output pins. These devices are sold pre-assembled, in a variety of models, depending on

the main purpose. For our work, we choose Arduino Uno, mainly because of its compatibility

with Micromanager, which has a specific adaptor and firmware for this kind of board. We used

three Arduino boards with modified firmwares to control the shutter, the galvo for DSLM and

the stepper motors for rotation of the sample and the filter wheels. The communication between

Micromanager and Arduino boards is performed through a serial port at 57600 bauds.

Micromanager's Arduino adaptor sends a byte of information (“Case bit”) through this serial

port which is interpreted by the firmware loaded in the Arduino memory as the function to be

implemented, and different extra parameters depending on the selected function. When the

execution is successful the controller sends a confirmation byte. We mainly use “Case 1” (Set

digital out command) for digital output signals and “Case 3” (Set analogue out command) for

analogue output signals through digital to analogue circuits (DAC).

In order to operate the galvo we use an Arduino board (AB1), as explained in Supplementary

Fig. 2. To control the stepper motors of the filter wheels and the sample rotation, we used

another two Arduino boards (AB2, AB3) as a State device controller. For AB2 we use the

Motor Shield Kit v1.1 from Adafruit Industries while for AB3 we can either use SparkFun’s

EasyDriver v4.4 or the Adafruit shield. The original firmware/adaptor pair of these boards is

programmed to use six digital output pins (8, 9, 10, 11, 12 and 13). After the “Case 1” byte, the

Arduino board receives a 6-bit number which is read by the microcontroller to control the

output pins. Byte 1 is associated with the pin 8, byte 2 with pin 9, 3 with pin 10, 4 with pin 11, 5

with pin 12 and 6 with pin 13, in a total of 64 different states. The byte is sent by Micromanager

through selecting an “Arduino State” and turning the defined “Arduino Shutter” on or off,

turning the chosen input/output pin high or low level state. The Shutter normally uses the pin 13

of the board. However this operation is not optimal for big rotation angles in a stepper motor,

since rotation will be performed step by step, by activating the different motor windings. For

every step four bytes should be sending therefore limiting the speed of the serial communication

between the computer with Micromanager and the Arduino device. For that reason we have

changed Arduino's firmware to diminish the time used in serial communication (see

http://uic.igc.gulbenkian.pt/micro-dslm.htm). When Micromanager sends a specific State to the

board, it moves the motor a predefined number of steps achieving 0.225, 0.45, 0.9, 1.8, 9, 18,

45, 90, 180 or 360 degrees rotation angles, clockwise or counter-clockwise. The “move”

function controlling the windings activation order was programmed inside the Arduino

firmware. By doing so, a limited number (20) of states is defined, but the autonomy and speed

are increased.

Nature Methods: doi:10.1038/nmeth.2508

Supplementary Note 2 | Sample preparation

Three different samples were used to test the performance of the SPIM/DSLM microscope. The

performance of the different imaging modes (multicolor SPIM, DSLM, 2p-DSLM) and the

ability of multiview imaging was tested on Drosophila melanogaster fly embryos (w P{w+

asl:YFP};;G147 His:RFP/TM3) expressing YFP-tagged asterless and RFP-tagged histones

kindly provided by Cayetano Gonzalez and Monica Bettencourt-Dias (Fig. 1a and

Supplementary Fig. 5). The fly embryos were dechorionated by collecting in 0.1% Tween in

water, bathing for 5 min in 50% bleach and washing 3 times for 5 minutes in distilled water.

Time lapse recordings (Fig. 1b and Supplementary Video 1) were performed on Transgenic

Tg(β-actin:HRAS-EGFP) zebrafish embryos2 constitutively expressing membrane-bound GFP

were kindly provided by António Jacinto and Ana Cristina Borges. Both samples were later

embedded in 1% low-melting-temperature agarose (LMA) in PBS for imaging. Multiview

fusion (Fig. 1c) was tested on a transgenic FVB/N.Tie II-GFP mouse3, expressing GFP in the

blood vessels endothelial cells under direction of the endothelial specific receptor tyrosine

kinase, developed until E10.5 was fixed overnight in 4% PFA, washed in PBS and cleared using

Scale solution4. In order to fit the field of view, the mouse was cut in half transversally by the

gut and the tail section was embedded in 1% (LMA) with a 1:10000 diluted TetraSpeck 0.5um

beads (Invitrogen) and placed inside a syringe to solidify. This sample was kindly provided and

prepared by Moisés Mallo and Arnon Jurberg.

OPT images of a mouse embryo E14.5 are shown in Fig. 1d,e following standard sample

preparation protocols5,6,7

. Briefly, after fixation (and bleaching and staining if necessary), the

sample is embedded in 1% LMA, a round block is cut and then gradually dehydrated into

methanol. The block is then embedded in Benzyl Alcohol+Benzyl Benzoate 1:2 (BABB), and

glued using a cyanoacrylate-based glue to a 30mm diameter metal washer which is connected

via a magnetic plate attached to the stepper motor of the OPT scanner.

Nature Methods: doi:10.1038/nmeth.2508

Supplementary Note 3 | Preparation of the OPT dataset and tomographic reconstruction

Axis of rotation of the sample must be perfectly aligned with the mid-vertical axis of the final

image. To accomplish this using the OpenSpinMicroscopy plugin, activate the button which

draws the crosshairs (“lines”), make sample rotate continuously (using a Time lapse acquisition

with 360º rotation angles) and use crosshairs as guides while adjusting the motor position using

the stage holder assembly, so the field-of-view (FOV) and motor axis are concentric. Note that,

for samples larger than wider (eg most late stage embryos), it is preferable to rotate the camera

90º so that the AP axis of the embryo is positioned along the widest dimension of the camera

chip. In this case the axis of rotation will be horizontal; tilt the motor by adjusting the kinematic

mount or by slightly rotating the camera. Note that if the images were acquired with the sample

rotating horizontally, the OPT dataset stack will have to be rotated so that the axis of rotation is

vertical, before back projection reconstruction. Adjust camera settings to avoid saturated pixels

(avoid intensities beyond 90% of the dynamic range), as these will produce artifacts in the final

reconstruction. Make sure you adjust the camera settings while imaging through the thickest

side of the sample as this will produce the brightest images (for example, for an embryo rotate

so either the left or the right side faces the camera); adjust camera settings to avoid saturation

and maximize dynamic range.

Acquire a full 360º set of projections, with 800 or 1600 angles. Less than 800 projections will

produce a poorly resolved reconstruction and more than 1600 projections will be difficult to

process without a powerful workstation.

Open full dataset in ImageJ or FIJI, if necessary rotate 90º to ensure axis of rotation is vertical,

and remove outliers (1 pixel radius and 1 gray level threshold, both bright and dark) which

would produce “ringing” artifacts in the reconstructed axial slices.

Even if sample was physically aligned, minute misalignments might still be present and will

produce noticeable artifacts (“ghost” images) and/or compromise resolution in the final

reconstruction. To confirm this (and re-adjust à posteriori), make a stack projection (StDev) -

which we will call a “Rorchach” image - and draw a line across the middle of the Rorchach to

find its vertical angle. If it is not 90º then the dataset stack needs to be “rotated” so the mid-line

is perfectly vertical. Even if the misalignment is as little as 0.5º it will be noticeable in the

reconstructed slices as “ghost” images. Sometimes it may be easier to find horizontal trails of

sharp details in the Rorchach image and measure the misalignment from those.

Confirm that the vertical line-of-symmetry of the Rorchach is concentric with the FOV,

otherwise re-center dataset by drawing a selection box around the Rorchach making sure the

middle horizontal ticks of the selection box coincide with the mid-line of the Rorchach. Switch

to the OPT dataset, restore selection, and crop.

Now the 0º and 180º projection images from the dataset should be fairly well aligned, ie,

coincident. To confirm this, create a new 2 slice stack and paste 0º projection in slice 1 and 180º

projection in slice 2, flip horizontally the 180º projection and compare them to make sure there

are no displacements or misalignments. Note that because of off-axis illumination or because of

signal attenuation in depth the two images may not be perfectly similar in brightness and details,

but they should be perfectly aligned! If there are still misalignments go back and realign in

ImageJ/FIJI, otherwise proceed to reconstruct slices using the Radon_Transform plugin, or your

back projection reconstruction algorithm/package of choice.

Reconstructing slices with the RadonTransform plugion for ImageJ/FIJI:

This processing step will reconstruct the tomographic slices from the full OPT dataset, which

are later used by 3D image reconstruction software for performing 3D image reconstruction,

rendering and analysis. There are several algorithms to do this, however here we will present a

simple method based on an open-source plugin for ImageJ/FIJI known as Radon_Transform

which can be obtained here: http://rsbweb.nih.gov/ij/plugins/radon-transform.html

Nature Methods: doi:10.1038/nmeth.2508

Supplementary Note 3 | Preparation of the OPT dataset and tomographic reconstruction

The Radon_Transform plugin uses only 180º projections, so half of the projections of the full

360º revolution dataset need to be discarded (notice that other algorithms may be able to use a

full rotation dataset!). If the projections are perfectly aligned it may be possible to combine

them into a single 180º dataset by averaging angles from 0º-180º with angles 180º-360º. Also,

Radon_Transform will only work properly with 8bit power-of-2 square box images (eg

512x512; 1024x1024, 2048x2048…) and with a dataset with multiples of 180 steps (eg, 180x1º

steps, 360x0.5º steps…). Micro-stepping motors typically produce multiples of 200 steps, so it

is necessary to change the OPT dataset canvas size and scale Z to the nearest multiple of 180.

The procedure “bloats” data, so expect to have 7-8x the size of the OPT dataset for the

reconstruction process and 5x the dataset size of RAM for ImageJ. As a reference, a 1600 angle

dataset acquired with a 2Mpixel 8 bit camera produces ~3Gb of projection images which will

require ~24Gb of Hard Disk space and >16Gb of RAM to process.

Open Radon_Transform plugin. Switch to the OPT dataset and re-slice from top and rotate

90degrees (to produce a “sinogram stack”) and save as “sinogram.TIF”.

From the Radon_Transform plugin window, import “sinogram.TIF” and input parameters in

plugin window accordingly (angle per step and total of projections).

“Save Data” into a text file named “projdata.txt” (wait a VERY long time… an 8 bit dataset will

be enlarged ~5x) and then “Open data” the “projdata.txt”, choose filtering method and

reconstruct stack (very long processing time, even in fast workstations). Save stack for

rendering and analysis with 3D reconstruction software.

In the case of embryos or animal organs it is a good procedure to reposition the reconstructed

tomogram by rotating it according to a radiological convention so that in the final tomogram the

series contains axial sections, starting caudally, with the embryo’s ventral side facing upwards

and left side facing rightwards on the slices. This often allows cropping of a significant part of

the volume to produce a smaller final tomogram. Typically the final axial tomogram will be

~1Gb.

There are several other open-source back-projection algorithms for MatLab or Octane (eg,

http://sourceforge.net/projects/niftyrec/,

http://octave.sourceforge.net/image/function/iradon.html or http://www.u-bordeaux1.fr/crazybiocomputing/ij/tomography.html) and a popular but non-open source

alternative is offered free by SkyScan Inc. known as NRecon.

Recommended open-source 3D rendering software:

FIJI and the 3D viewer plugin can be used to view OPT tomograms; though this solution is

convenient (runs directly off of ImageJ/FIJI) and it has several interesting features and a

relatively easy method to produce 3D animations, the quality of the rendering is still somewhat

limited, a problem common to other plugins for the ImageJ platform. Higher quality and faster

renderings can be obtained with any of the following (all free, most open-source as well):

VolViewer, Volview, Osirix (Mac only), Drishti, 3D slicer, FreeSurfer, Voreen or MevisLab, by

order of difficulty for a novice user.

Nature Methods: doi:10.1038/nmeth.2508

Supplementary References

1. Olarte, O. E. et al. Biomed. Opt. Express 3, 1492–505 (2012).

2. Cooper, M. S. et al. Dev. Dyn. 232,359 -368 (2005).

3. Motoike, T. et al. Genesis 28, 75-81 (2000).

4. Hama, H. et al. Nat. Neurosci. 14, 1481-1490 (2011).

5. Sharpe, J. et al. Science 296, 541–545 (2002).

6. Bryson-Richardson & R.J., Currie, P.D. Methods Cell. Biol. 76, 37-50 (2004).

7. Mouse Atlas Project:

http://www.emouseatlas.org/emap/ema/protocols/embryo_collection/ec_agarose.html

Nature Methods: doi:10.1038/nmeth.2508