Transillumination for Widefield Fluorescence Microscopy - Escaping the Tyranny of Dichroic Mirrors

Hardware
1

Hi uForum! Longtime lurker, firsttime poster, etc. etc.

I’ll preface this by saying this is more of a thought and educational exercise for me, at least until I convince someone to throw some grant money my way :slight_smile:

In my pursuit of automated, high-throughput, high-plex (>5 colors) widefield fluorescent imaging of slides, I continue to run into a rate limiting step of the long switching times (250-1000ms) for dichroic filter turrets or linear sliders. The rest of the equipment on the TTL chain (excitation, filter wheels, xy(z) stage, camera) sit waiting for that cha-chunk of the dichroic changer.

Removing two dichroic changes per field saves about 1 second, multiplied by however many fields are being acquired per sample, so time savings scale geometrically with magnification.

My question is this: cost being no object, are there any fundamental issues precluding the creation of useful transmitted fluorescent scopes?

There are two key assumptions that make this hypothetical instrument temporally superior to epiilluminated scopes:

  1. Each field/tile must be acquired with all colors before the next field/tile is imaged.
  2. The instrument/user is already using a fast ‘emission cleanup’ filter wheel in conjunction with the multipass dichroic. Meaning each color requires a 30-50ms wheel change.

I understand that phosphor-coated LED light sources are susceptible to secondary emission via emitted fluorescence of the sample propagating back into the excitation source. I’ve only seen this with ‘white light’ LEDs and am unsure if it occurs in monochromatic light sources using light pipes. This should be solvable with ND filters in the excitation path or using non-excitable light sources such as a lamp.

Additional benefits would include the availability of all emitted light. I dream of using two filter wheels (one populated with shortpass filters, the other with longpass) for flexible emission selection, useful for broad wavelength collection with dim fluors, or for separating short and long stokes-shift fluors on the same excitation wavelength from each other e.g. 488 and PerCP.

Happy to hear how silly of a thought this is, or if it’s already out there somewhere and I missed it.
Happy Holidays!
Riley

2

Hey @Riley_Hannan Welcome!

Not to dismiss the fundamental question (i think it’s always worth asking/revisiting these sorts of questions)… but I guess my first question would be: why not use a multi pass dichroic? Many folks who are after speed in multicolor imaging will use the setup you described (ttl triggered everything with ex/em wheels), and then plop a quad pass filter into the dichroic wheel and not turn it at all. Is that precluded here for some reason?

As for why not transmitted fluorescence, my assumption is that it puts too much burden on the emission filter: that is, emission filters have a limited capacity to block unwanted light (the OD of the filter) in addition to simply selecting the band of light you do want. So, given the relatively minuscule amount of emission photons of a fluorophore relative to the ridiculous amount of photons needed for excitation, I suspect your signal to noise/background ratio would be pretty terrible (without adding perhaps a lot of emission filters in sequence?)

1 Like
3

Thanks so much for the reply!

Multipass dichroics are great for up to five colors, and I should have specified all the dichroics in my example above would be multipass. A common use case for me is two quad-pass dichroics (DAPI/FITC/TRITC/CY5 and CFP/YFP/mCherry/Cy7) but pushing past the plex of the dichroic requires a mirror change, and thus the interminable wait. Being able to use the 6-8 discrete excitation colors common in solid state illumination systems without a dichroic change is the dream.

Regarding the blocking of excitation light, I think that’s a vital aspect I’ve completely missed. My tacit assumption would be that high quality filters with OD6 in the blocked bands would be sufficient. I realize I’ve never considered how many orders of magnitude separate the intensity of exciter light to emitted light. I imagine it’s many more than 6. I’ll have to do some maths and see what OD would be required. I’ll pipe back up when I do.

Happy holidays!

1 Like
4

I’ll mention that in light sheet, we can actually get away with this sort of thing… since the beam of excitation isn’t going directly into the collection lens. So, perhaps another angle you could take here; depending on your resolution needs, would be to use a higher na on the excitation than on the emission, and illuminate at steep angles you’re not collecting (ie, darkfield fluorescence)

1 Like
5

I think seeing the optical path of a light sheet for the first time made me go down this rabbit hole in the first place!

I had originally taken a brief look at darkfield condensers but found it nearly impossible to get illumination field of view data for various models. I’ll do some more thorough searching. It would certainly make the emission light train more manageable! I’m worried I will end up reinventing a light sheet with my sample being a bunch of stacked coverslips, ha :woozy_face: