The CHEap mESOscopic Selective Plane Illumination Microscope

aka

The cheesoSPIM

A $500 fluorescence light sheet microscope that you can build at home. Plus a protocol for clearing wood for use in fluorescence microscopy that you can also do at home. Design and build guide below.

Pics up front

<p> <img src="https://github.com/PRNicovich/cheesoSPIM/blob/main/output/otherEndView.png" width="300"> <br> <em>Image of cleared balsa wood captured on the cheesoSPIM. Pores (dark holes) and tooling marks (white lines) are visible. </em> </p> <p> <img src="https://github.com/PRNicovich/cheesoSPIM/blob/main/output/swirlyDye.jpg" width="300"> <br> <em>A partially mixed dye solution illuminated by the cheesoSPIM light sheet. </em> </p>

Motivation

1. How cheap could you make an AIBSOPT?

   At an old job I designed and built this, a low-cost microscope for rapid 3D transmitted light and fluorescence imaging. The system was designed to do optical projection tomography in which a series of 2D images are collected and computationally reconstructed into a 3D image of the sample. The principle and math behind the technique is identical to computed tomography used in medical imaging. OPT uses visible light, not X-rays, and the sample is rotated between each image collected, where CT scanners rotate the detector around the patient.

   After a few emails around the topic I was thinking on how the design could be made cheaper. One of its advantages is its cost, which is cheap for lab equipment. But that's compared to, like, a confocal microscope and not to, say, a normal human number of dollars. Taking a closer look at the two most expensive pieces of the design - the camera and its lens - would make the most sense. OPT imaging is rarely starved for photons; a cheap(ish) USB webcam is plenty for this application. That leaves the lens.

   The images collected in an OPT are ideally an optical projection of the specimen onto the camera chip. The system approximates the detector on a CT scanner, in which there are no optics. The image is an orthographic projection formed by a specimen in the path of a collimated X-ray beam. The optical equivalent is imaging with a telecentric lens. These lenses are designed to form an image from a bunch of parallel rays of light. The result is the optical orthographic projection we require. Conveniently the same requirements that work for an OPT imaging lens work for a light sheet, though often only the sheet excitation plane is projected onto the camera rather than the entire specimen.

   Telecentric lenses are adventageous in metrology and industrial machine vision because the size of an object does not vary with distance to the lens. Surplus lenses are available in a variety of specifications. These lenses fit a profile of unexpectedly common (thanks to industrial applications) and highly specific (so unlikely to be reused in new work) that makes them favorable for buying surplus. We can also afford to be a little flexible here on magnification and other specs.

   I found an 0.5x C mount telecentric lens for $86, shipped, on eBay.

2. Could you use a commercial motorized lens to tile a light sheet?

   Axial scanning is hard in microscopy. It usually means moving something around precisely. Precision is always tough and tough usually means expensive. A shortcut around expensive can sometimes be found in precision someone else paid for. That can mean surplus equipment that still works or can be repaired, or it can mean mass production where the precision is in the tooling, amorated over thousands of replicates. The best is when both happen - surplus, mass-produced, precision equipment. Consumer optical equipment is a good example of this intersection. Maybe there's a product out there already doing what we need for axial scanning?

   The MesoSPIM uses a pair of Nikon DSLR lenses as the objective lens in each of the mirrored excitation light paths. There those lenses are static and used because they have the appropriate focal length and field size with good field flatness for forming the excitation sheet. In the MesoSPIM a tunable lens upstream from the DSLR lens is responsible for axially scanning the light sheet. DSLR lenses these days are all motorized - could that work for scanning the sheet on its own?

   Canon DSLR lens communication documentation or at least hacked versions thereof is available and works through an Arduino. Lenses using the EF protocol are a little outdated electronically but the optics still work just the same. A lens with about a 50 mm focal length is probably right, and a zoom wouldn't hurt for testing. The EF protocol is a requirement but there are something like 100 million of that type lens in the world.

   I found a Canon 35mm - 105mm f4.5 EF zoom lens for $36 on eBay.

3. Dramatic dichroics

   If the telecentric lens is the collection side and the DSLR lens is the excitation side, then this is a fluorescence imaging system. The remaining piece is how to separate the excitation light from the laser from the emitted fluorescence from a specimen. In a typical fluorescence microscope this is done with dichroic mirrors and interference filters. Here I'd need one large enough to cover the front element of the telecentric lens - 50 mm or so - which could mean upwards of $1000 from a scientific optical filter supplier for a single part. But they're not the only ones that sell dichroics! The film and stage lighting industries use interference filters for plenty. Would any of those work?

   Rosco's Permacolor line has many options for colors (aka pass bands) of filter to choose from. Spectra are a little tricky to find but the Industrial Green P1086 looked to line up with GFP or FITC emission reasonably well. A good place to start.

   I ordered 2 x 1.95" P1086 filters from Musson Theatrical for $70, shipped.

4. Home clearing

   A microscope needs a specimen. Fluorescence microscopy and OPT both greatly benefit from an optically clear specimen. Makes sense if you want to see all the way through something it's easier if it's see-through. Unfortunately most of the options for tissue clearning involve a system of witches brews and potions and incantations that are stinky, sticky, sickening, or all three.

   This unsanctioned project began to coalesce into an overheated science fair project and seemed appropriate to include a specimen that could be prepared at home. The item should be reasonably easy to obtain from a commercial source and clear with (reasonably) safe protocols. Turns out cleared wood is a thing as a bio-friendly building material. There the process is delignification (bleaching) followed by impregnation with plastic (index matching).

   Starting with balsa wood, bleaching, and index matching with a proper solvent or mixture seemed like a reasonable way to go. This omits the impregnation step above for the more typical approach in fluorescence microscopy of soaking the bleached and delipidated specimen in an index-matching fluid.

   Chemicals required are avialable for < $100 in quantities between sufficient and 'lifetime supply' from Amazon.

5. LOLs

   The completion of this project became a challenge in demonstrating a working light sheet + OPT microscope and specimen to image using what is available to the general public. Reimagining a complex machine in the cheapest terms possible is a useful exercise for a few reasons.

   Making something cheap is challenging. It takes little deviation from idealized theory if all of your instruments are built of nothing but the highest quality materials machined to tight tolerances. If you're pushing the limits of physics then that might be necessary. But if you're not then it's likely cheaper and faster and easier to compromise precision and performance where it's not needed. Figuring where that is true requires attention to details of the operation of the instrument and choices made outside of the usual ones for biomedical research. Experimental science often has a hangup around an approach being the best or most precise or highest resolution or whatever. This is often leads to thinking that the best instrument is the only acceptable option. Instead one should start from what is required to answer the scientific question at hand. Any performance beyond that is usually superfluous.

   Making something cheap makes it more accessable. Scientific equipment is expensive. Making it cheaper means more folks can possibly take advantage of this technology. The work presented here is within the domain of a well-supervised group of high school students. Or if someone wants to wipe the floor with their classmates at a science fair this might do it.

   Making it cheap makes it more fun. The bar for trying something that costs $20 is a lot lower than if it costs $200 or $2000. The pressure for that experiment to yield something useful is lower, too. Making it cheap leaves more room for creativity and individuality. As in here, one can play with the laser cutter and 3D printer in a new way, explore some optics ideas, or just have their own little machine they built because it's amusing to build things.

Parts + Price

• Laser - $36, eBay 80 mW 450 nm diode laser + controller

• Excitation filter - $21, eBay 450/40 BP, 15 mm diameter

• Cylindrical lens - $8 for 5, eBay 5 mm diameter x 11 mm length

• Excitation objective lens - $36, eBay Canon EF 35