Doug Shepherd

Thanks! Your benchtop mesoSPIM paper was very helpful!! Confirmed a lot of what we were seeing and gave us the kick to finish this off.

Additionally, @adamezracohen.bsky.social group's paper that used this beast of an objective for widefield was very helpful when thinking about how to optimize the light path to retain as much performance as possible.

There are a lot of nice papers dealing with field curvature in LSFM, including work that from @nvladimus.bsky.social, @adamkglaser.bsky.social, @aakinwo.bsky.social, and others whose bluesky handles I don't know.

Schematic on how to correct field curvature along light-sheet propagation axis.

We're still exploring how to improve LSFM performance with "big" detection air objectives, where field curvature often limits achievable FOV. Here, Steven introduces curvedASLM, a method to dynamically correct field curvature (along one axis) in axially scanned setups www.biorxiv.org/content/10.6...

Special thanks to @moffittlab.bsky.social for a lot of help and advice along the way! More to come soon on speeding up the experimental side of large scale 3D spatial omics.

GPU-accelerated, self-optimizing processing for 3D multiplexed iterative RNA-FISH experiments Imaging-based spatial transcriptomic approaches rely on iterative labeling and imaging of carefully prepared samples, followed by solving a computational inverse problem to determine the location and ...

If you've seen anyone from the lab talk about RNA-FISH over the past years, they've shown data somehow related to this preprint. After a wandering road writing it up, happy to put out some thoughts on scalable 3D imaging-based spatial transcriptomics data processing.
www.biorxiv.org/content/10.1...

Pseudo-colored fluorescence microscopy image of expanded paramecium and tetrahymena cells.

We are really enjoying having Nick Galati visit us on sabbatical for some fun protist work, including our first forays into expansion microscopy!

Max projection xy&xz views of expanded paramecium and tetrahymena cells. Scale bar is ~5.5 um accounting for expansion. Epifluorescence + deconvolution.

We tried your FEP chamber setup and the results are great with an Olympus 60X/NA 1.0 water dipping and U-ExM ciliate samples. Thanks for putting it out there!

Microscope built on rails

🔬We are now accepting applications for the 2026 Analytical and Quantitative Light Microscopy course at MBL. Come learn to be an excellent microscopist!
Questions? Ask here or DM me.
@mblscience.bsky.social @aqlm.bsky.social
#microscopy #fluorescence

Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences: Vol 383, No 2304

Our special issue of Phil Trans Roy Soc A on "Biological Fluid Dynamics" Emerging Directions" is officially out now! Originating from discussions at one of our SIG meetings, this special issue has 16 research articles and reviews on all things bioactive! royalsocietypublishing.org/toc/rsta/202...

Ultrafast tomograms in a flash Fourier synthesis optical diffraction tomography encodes hundreds of views per exposure, enabling kilohertz-rate, label-free three-dimensional imaging of complex biological and soft-matter dynamics pr...

Accompanied by a nice perspective piece from our colleague Lei Tian, whose own work on high-information content computational imaging was a strong inspiration for our own.
www.science.org/doi/10.1126/...

Fourier synthesis optical diffraction tomography for kilohertz rate volumetric imaging Highly multiplexed diffraction tomography enables kilohertz volume imaging for dynamic soft matter and biological systems.

@ptbrown1729.bsky.social and the rest of the team's work on high-speed optical diffraction tomography plus structured illumination microscopy is finally out! www.science.org/doi/10.1126/...

GitHub - QI2lab/opm-processing-v2: Code for fast deconvolution, deskewing, registration, and fusion of OPM/diSPIM/LLSM data. Code for fast deconvolution, deskewing, registration, and fusion of OPM/diSPIM/LLSM data. - QI2lab/opm-processing-v2

Tensorstore is much faster for read/write than vanilla zarr. Tools are here: github.com/QI2lab/opm-p...

Currently we only support our custom pymmcore-plus OPM format and ASI’s OPM format. Happy to work with you to support CZI LLSM files if you open an issue and provide some example data.

We have a bunch of fast skewed data handling tools (flatfield, decon, deskew, stitch) in Python, but they are not meant for real-time lazy viewing - we do that in our own custom acq. software. We use Tensorstore to write with compression at reasonably fast speeds.

We use a number of FLIR and Basler cameras with the high end Sony CMOS chips. For short exposures, they can work well. Another thing to look out for is that the hot pixels fluctuate on top of the true dead pixels. We correct a static map and then do a second correction that looks for outliers.

Our experience is the dark current (no cooling) and dead pixels (quality control) are higher than your standard sCMOS. We've mapped the pixel by pixel gain for a few different industrial CMOS chips and often find some non-linearities.

Adaptive optics oblique plane microscopy of 100 um thick mouse brain slice with immunofluorescence.

Spending some time at #lsfm25 cleaning up code. It now has distributed GPU gradient-consensus RL decon (@jamesdmanton.bsky.social @andrewgyork.bsky.social) our fast orthogonal deskew, and stitching/fusion to ome-ngff v0.5. Example AO-OPM data on uncleared mouse brain slice github.com/QI2lab/opm-p...

Thanks! We use a Nikon 100x/1.35 silicone oil for O1. The objectives in the setup are still the same as our previous Snouty-OPM papers.

As usual, all the code is developed in the open at our github. Control: github.com/QI2lab/opm-v2 and processing: github.com/QI2lab/opm-p....

The control code is fully hardware triggered and pretty performant already. We are actively working on speeding parts of the processing (fusion to ome-zarr).

Immunofluorescence image of a mouse brain slice, with nuclei (magenta) and microglia (orange) shown.

Zoomed in Immunofluorescence image from the full mouse brain slice, with nuclei (magenta) and microglia (orange) shown.

Steven has our Snouty-OPM running again after our lab move. He added projection mode, sensorless AO, and a new control GUI based on @talley.codes + crew's pymmcore-plus/pymmcore-gui. Here are some IF-labeled microglia (orange) in the middle of an uncleared 60um mouse brain slice. No deconvolution!

Out of nowhere, our institute just froze all indirect accounts - no matter when or where the indirects were generated. With completely unclear direction on how this works going forward.

No words…

California condor perched on a rock with the Grand Canyon in the background.

In a brief respite from current events, I got to see a California Condor for the first time while running in the Grand Canyon this weekend! What an amazing animal.

Unfortunately, public lands and those who protect them are also targets of this admin. Keeping fighting…

Thank you!

Super cool! Would be fun to play with at some point.

This is yet another reason (in addition to MANY others) to call your representatives and senators. Likely more is going to be required, but it's a starting point.

CHCs served roughly 31 million people in 2023, 90% of which were at or below 200% the federal poverty level. That includes roughly 9.1 million pediatric patients.

Despite the court orders and the executive branch's claim that Medicaid is not effected, their Medicaid funding portals remain frozen.

Community Health Center Patients, Financing, and Services | KFF This brief reports on community health center patients, services, experiences, and financing in 2023 and analyzes changes from 2019 (pre-pandemic) through 2023 using data from the Uniform Data System ...

My partner told me today many community health centers (she works at one) may need to declare bankruptcy if the Federal funding freeze is allowed to continue. If you aren't familiar with CHCs, here is a good overview: www.kff.org/medicaid/iss...

Left to right: pseudo-widefield, FISTA-SIM, Wiener filter SIM, and Bayesian SIM reconstructions of mitochondria labeled with PkMito Red in a live HeLa cell. Top to bottom: decreasing signal, starting at an average of 762 photons per pixel and dropping to an average of 16 photons per pixel.

A nice treat in this craziness. Ayush’s work on SIM reconstructions using physical noise modeling is out: www.nature.com/articles/s44... The low signal (avg of 16 photons/pixel) reconstructions, achieved without deep learning, are quite impressive in my opinion. We’re working on a faster GPU version

Yeah, lead spheres and torsion bar suspended by 20 micron diameter tungsten wire. It's a really nice setup, but tricky for them to get right because of how long it takes to settle down before you can take data.

Fixed the Cavendish experiment for our second semester advanced lab, fun with gravity!