@jytinevez
Research engineer and core facility manager at the Institut Pasteur https://research.pasteur.fr/en/team/image-analysis-hub/ Everything #TrackMate #Mastodon-sc (the tool to track lineages in microscopy videos, not the social networks etc). Views are my own
🤖 How can #AI unlock the full potential of biological data?
🔬 Our @florianjug.bsky.social develops advanced AI methods to help life scientists extract deeper insights into health and disease within the MIND innovation ecosystem.
Watch the interview 👉 www.youtube.com/watch?v=ucZl...
Appose processing hackathon with @jytinevez.bsky.social, @ctrue.bsky.social, @gaellel.bsky.social, @strigaud.bsky.social, Carlos and Thomas . Simple POC with Java calling python through Appose to flip blocks. Promising!
While we sit around to find out how bad exactly it all is, a bit of cheer: the best baguette of Paris competition was won by a Sri Lankan who appears to have started as a dish washer in restaurants and eventually trained as a Boulanger.
www.paris.fr/pages/la-mei...
Final version @nature.com of our paper describing unconventional multicellular development in a choanoflagellate inhabiting an extreme environment. A ton of new data since the first @biorxivpreprint.bsky.social preprint (which we've kept updating).
A brief 🧵 (carried over from the old place)
Thanks @pasteur.fr , @pasteuredu.bsky.social , @france-bioimaging.bsky.social and @prairie-psai.bsky.social for support!
If you need bioimage analysis for your research, and want to learn to get things done with it, apply!
See you in beautiful Paris this Spring to do bioimage analysis together!
The course is organized in practical tutorials, that will be practiced on your laptop, and mainly based on academic open-source scientific software: Fiji, Icy, QuPath, Ilastik
Here is the program from last year, knowing that the one from this year will be similar
In this edition, organized with the @biaunit.bsky.social , we will focus on teaching the basics of bioimage analysis and machine learning for image analysis in Life Science.
The course is focused on beginners, with little or no experience in image analysis, but who needs it for their research
If you want to skip and register immediately, the link is this one:
www.pasteur.fr/en/education...
One week left to register to the Bioimage Analysis with AI course in Pasteur!
This June, 1 - 5.
I give a bit more details in the thread below
👇
⚠️ 1 week left for Pasteur Course "Advanced Bioimage Analysis With AI"! 👩🔬💻
Intensive course (June 1-5, Paris): theory + hands-on, AI focus in bioimaging.
📍 Institut Pasteur
🗣️ English
⏰ Deadline: March 1
🔬 Beginner basics, batch scripts, interactive sessions
👉 Apply: https://bit.ly/4c8rFKZ
Another banger from Talley
Join our team! Postdoctoral position to study the malaria parasite Plasmodium falciparum 🔬🦟 research.pasteur.fr/en/job/postd...
New Member Alert: In January, @awehenkel.bsky.social , along with the Image Analysis Hub, under the direction of @jytinevez.bsky.social at @pasteur.fr joined SBGrid. Welcome!!
With F-BIAS services included!!
Request our services there!
Delighted to share our latest preprint "Filopodia-mediated trans-endocytosis"
www.biorxiv.org/content/10.6...
Work driven by many people including Hanna, Marcela, Sujan, Marie-Catherine, Anna and Monika. Fantastic collaboration with Johanna Englund, Emilia Peuhu, and Eija Jokitalo
Thanks @pasteur.fr and @france-bioimaging.bsky.social !!
We hope this will be useful to the wide scientific community.
We have developed this approach first for microbiology because these are the projects we received in the core, but of course, nothing prevents you from applying it to tracking any kind of objects.
Happy tracking!!
The preprint contains dataset, analysis notebooks, and detailed instructions how to perform the optimization, the ground-truth creation, and how to make sense of the results.
All of the work we did to make this happen is now available in TrackMate v8. You simply need to download a fresh Fiji LATEST (important) to get it.
(fiji.sc)
In the second one, bacteria swim, their shape matters not, but the track feature we get from them do.
We illustrate the approach on two use cases: 1/ Growth defect in E. coli and 2/ Assessing how mutations affect H. pylori motility. Both cover widely different situations: For the first one, bacteria grow and divide. Their shape and timing of cell division is important for the downstream analysis.
This is the approach we introduce in this preprint. It contains a methodology to **generate tracking protocols**, that are automatically optimized and yield a known accuracy estimate.
For instance, if the optimization gave you a CCA of 95%, and that you find by comparing two conditions that the cell cycle length changes by 30%, you can be confident that these 30% are not and artifact from tracking errors. (You still need to cross validate, tough)
The metric value at optimum is especially useful to assess the validity of the downstream analysis.
The first part is useful of course because you get an objective optimum for your analysis pipeline in an automated manner. No need for never-ending trial and error, without a clear stopping criterion.
At the end you get the tracking configuration that maximizes the metric, and the value of this metric.
Then the TrackMate-Helper simply tests all tracking configurations in a grid search. For each algorithm and parameters of the search, it computes the selected metric and iterate.
This is typically an overnighter. Or two.
2/ Provides a tracking ground truth over a small movie except.
3/ Configures a parameter sweep.
