Lightnews — Scholar-powered news

Brian Northan @truenorth-ia.bsky.social · Aug 16

I was searching for a Python utility for Fourier Ring/Shell Correlation. All I got in a blue sky search were 4 small posts from @christletx.bsky.social ... all I got from chatgpt was installation instructions for non-existent libraries and github repos... anyone know a good tool for this?

3 1 3

Reposted by Brian Northan

Crick BioImage Analysis Symposium @cbias.bsky.social · Aug 11

📢 After #CBIAS2025, join our hands-on 2-day napari training by @crick.ac.uk @ucl.ac.uk @liverpooluni.bsky.social
From conda to plugins to Jupyter, practicals on real data with expert guidance.
For Python/ImageJ coders. Apply by 1 Sept – limited to 30! crick.ac.uk/whats-on/cbias-napari-workshop-2025

CBIAS napari workshop 2025

crick.ac.uk

17 24

Reposted by Brian Northan

Beth Cimini 🔬💻📊 @bethcimini.bsky.social · Aug 5

Episode 11 - In which @erinweisbart.bsky.social and I talk about just how much we love ilastik (@ilastik-team.bsky.social) and how it's great for so many problems! Also, that you should use a small brush.

2 5 17

Reposted by Brian Northan

Sebi @sebi06.bsky.social · Jul 30

The 1st working version of my little @napari.org plugin (napari-czitools) is almost ready.

It contains sample data, allows to read CZI metadata and can read CZI images (or parts of it) by using the sliders to crop what should be read.

Still quite some things to improve ...

1 3 9

Reposted by Brian Northan

Ben Gallusser @bengallusser.bsky.social · Jul 30

Cell tracking is never perfect, and it's important to understand the types of errors your solution contains.

Here is my stab at this: Divisualisations in @napari.org. github.com/bentaculum/d...

You spin tracks out upwards from the playing video. Green edges are correct, FP in magenta, FN in cyan.

1 17 52

Brian Northan @truenorth-ia.bsky.social · Jul 30

I just published a new video demonstrating how to use Pixi and Napari-Easy-Augment-Batch-DL for repeatable and reproducible deep learning segmentation. In this example, I use Pixi to set up a Python environment then launch Napari to explore and compare Cellpose3 and Micro-SAM. youtu.be/3ObAnSTBri8

Pixi and Napari-Easy-augment-Batch-DL

YouTube video by Brian Northan

youtu.be

7 14

Brian Northan @truenorth-ia.bsky.social · Jul 21

Great to see @alexhego.bsky.social share instructions on a DL pixel‑classification problem we worked on at image.sc. If you’re working on a challenging 3D DL semantic‑segmentation problem and need more options than most tools provide, check out this repository. github.com/AlexHego/DL-...

GitHub - AlexHego/DL-pixel-classification

Contribute to AlexHego/DL-pixel-classification development by creating an account on GitHub.

github.com

2 10 21

Reposted by Brian Northan

Fabrizio Musacchio @fabmusacchio.bsky.social · Jul 13

Btw, and in the spirit of #OpenScience, all my #Napari #bioimage analysis #tutorials are freely available — from basics to #segmentation, #tracking, and #clustering. Feel free to explore and share ✌️

🌍 www.fabriziomusacchio.com/teaching/tea...

12 31

Brian Northan @truenorth-ia.bsky.social · Jul 15

Giraffe spots would be something to measure that would be corrupted by the noise too. So would want to keep them and after restoration it would be nice to be able to count the correct number of spots, shape of spots and (importantly ) intensity of spots...

1

Brian Northan @truenorth-ia.bsky.social · Jul 12

Even with the really good microsam/cellposesam generalist models, annotating specialist labels and retraining is always better.

1

Brian Northan @truenorth-ia.bsky.social · Jul 12

Love to hear more about testing. I'm always skeptical when I hear "new method B" outperforms previous method "A" because it often turns out "A" was not run optimally. Seems reasonable in an unsupervised or general mode this new method could be better than generalist Stardist/Cellpose but ...

1 2

Brian Northan @truenorth-ia.bsky.social · Jun 17

You will be missed. I really appreciated you getting the North East Bioimaging meeting organized the last couple years.

1

Reposted by Brian Northan

BIAUnit @biaunit.bsky.social · Jun 17

🚀 Manual mask annotation slowing you down?
💡 SAMJ is a new ImageJ/Fiji plugin using Segment Anything (SAM) for fast, interactive segmentation — no GPU needed, one-click install, real-time results.
🔬 Perfect for large biomedical images.
arxiv.org/abs/2506.02783
#AI #Bioimaging #ImageJ #OpenSource

1 13 55

Reposted by Brian Northan

Constantin Pape @cppape.bsky.social · Jun 16

We released version 1.6 of micro_sam:
- Improvements for automatic tracking.
- A new experimental mode for object classification.
- **New versions of the LM and EM models**
The models fix artifacts in automatic segmentation, see old vs. new prediction and better 3D segmentation results due to it.

3 6 20

Reposted by Brian Northan

FocalPlane @focalplane.bsky.social · Jun 9

New ‘How to’ post on FocalPlane 🔬💻

Lea Kabjesz shares a step-by-step guide to get you started annotating 2D and 3D images in Cellpose, including video tutorials.
focalplane.biologists.com/2025/06/05/a...
#bioimageanalysis #cellpose #imaging #HowTo

Annotating Images in Cellpose - FocalPlane

Annotating Images in Cellpose - How to

focalplane.biologists.com

1 13 33

Reposted by Brian Northan

Dave Barry @david-j-barry.bsky.social · May 30

Need help with bioimage analysis? Contact the UK Euro-BioImaging Node to see how we can help!

www.eurobioimaging-access.eu/nodes/uk-node

12 19

Reposted by Brian Northan

AI4Life @ai4life.bsky.social · May 27

Want to host your own AI models without relying on external infrastructure?

Enter BioEngine
ai4life.eurobioimaging.eu/announcing-b...

Your go-to for flexible, self-hosted bioimage analysis. Open source, scalable, and built for you!

5 11

Brian Northan @truenorth-ia.bsky.social · May 27

Where did you order it? Seems like the retail price is suppose to be about $2000, however online they peaked at like $5000 a month ago, but seem to be down to $3000 now.... still way above suggested retail price.

1 1

Brian Northan @truenorth-ia.bsky.social · May 27

I've started a thread on the imagesc forum about how ambiguous labels might affect deep learning models. If you've encountered mislabeled data, strange training artifacts, or examples where the former caused the latter — I'd love to hear your insights. forum.image.sc/t/are-we-see...

Are we seeing downstream effects of ambiguous labelling on new generalist models?

Below is a rather long post about some issues I’ve been seeing in generalist models ‘finding’ and ‘not finding’ objects, and my hypothesis as to why. Love to hear if others are seeing similar. The i...

forum.image.sc

1 5

Brian Northan @truenorth-ia.bsky.social · May 20

2) Machine Learning - low labeling work/low training time and resources. Sparse-label deep learning low labeling work/high training time and resources. Dense deep learning high labeling work/high training time and resources. So which one to use based on the trade offs?

1 2

Brian Northan @truenorth-ia.bsky.social · May 20

1) Hi @haesleinhuepf.bsky.social on this topic are you aware of any good white papers or blogs on different approaches to semantic segmentation specifically Machines Learning vs Sparse-label deep learning vs Dense-label deep learning? So basically....

1 2

Reposted by Brian Northan

Marco Dalla Vecchia @marcodv.bsky.social · May 16

We've been trying nnInteractive (github.com/MIC-DKFZ/nap...) to quickly create labeled data for later training an Omnipose model. I whatever black magic you did there, keep at it! It's such a life-changer!

1 4 11

Reposted by Brian Northan

napari @napari.org · May 16

Reminder: if you go to the GloBIAS conference we can help you contribute to napari at the hackathon! A bunch of the team will be there! (Great thanks to the GloBIAS team for accommodating us!!!)

More details on the conference website.

GloBIAS @globias.bsky.social · May 15

Deadlines extended - you can still register for the GloBIAS conference! We have extended most deadlines. Please check our webpage: tinyurl.com/mr3kx5a5! Looking forward to welcoming you to Kobe, for training, developing and discussing a broad range of BIA topics!

2 7

Reposted by Brian Northan

Beth Cimini 🔬💻📊 @bethcimini.bsky.social · May 15

Episode 5 - sure, most cells have one nucleus, but what happens when they DON'T? @erinweisbart.bsky.social and I try to get to the bottom of what on earth you can do when your cells and nuclei just don't match up on the latest episode of "Ask Erin/Dear Beth"

6 11

Brian Northan @truenorth-ia.bsky.social · May 16

This is a nice overview of normalization techniques which are very important for deep learning training. It is pretty common that differences in the training and testing image intensity ranges lead to unpredictable errors so learning to troubleshoot normalization issues is very useful.

Mara Lampert @marabuuuu.bsky.social · May 16

Image normalization: trivial enough to automate, impactful enough to quietly sabotage your results.
Here’s a Jupyter notebook I wrote on how, why & when normalization matters, feedback 💛ly welcome.
👇

haesleinhuepf.github.io/BioImageAnal...

Image Normalization — Bio-image Analysis Notebooks

haesleinhuepf.github.io

1 2 12