carsten marr retweeted
Introducing HistoGPT: A Foundation Model for AI-Powered Pathology Reports.
💡A new AI tool with the potential to transform dermatopathology by automating process of generating high-quality pathology reports from histology images.
🔗t1p.de/83o0a
Great collaboration and great model design from @ValentinKoch_ !
This Post is from an account that no longer exists. Learn more
carsten marr retweeted
Proud of the first chapter, hopefully of many, of our collaboration with @JungmannLab published in @CellCellPress where SUM-PAINT was developed for imaging neurons. Kudos to Eduard Unterauer @jekriste @mahipal_ganji @l_masu @MarrCarsten @FelipeOpaz_o. sciencedirect.com/science/ar…
carsten marr retweeted
Difference in Piezo1-dependent Ca2+-influx, so what? To show "what", we teamed up with @MarrCarsten's lab, adopting their custom-developed RBC morphology analysis algorithm. Ca2+-influx causes K+-efflux and loss of water from RBC - cells shrink.
8/n
Amazing collaboration indeed with generous funding from @VolkswagenSt
Happy to share the outcome of a long-standing collaboration with @DrukkerMicha and @MarrCarsten on mechanisms of species-specific differentiation speed biorxiv.org/content/10.1101/… @mpimoph @HelmholtzMunich @LACDR A thread
carsten marr retweeted
#PositionPaper 🇩🇪
Experts offers solutions to navigate the complexities of using #patient #data in #medical #research, addressing privacy concerns while maximizing #AI's potential.
👉Read more about their recommendations t1p.de/yw7a6
#HelmholtzMunich @MarrCarsten
Amazing work from @ValentinKoch_ and @sophiajwagner !!! Thank you both!
This Post is from an account that no longer exists. Learn more
... how could I forget to tag @sophiajwagner (Hope you enjoy your research visit @AI4Pathology !) and Robert from @laion_ai (also in US right now?) ? Best regards from Munich!
I am thrilled to announce HistoGPT, a vision language foundation model that generates highly accurate pathology reports from gigapixel whole slide images.
medRxiv: medrxiv.org/content/10.1101/…
GitHub: github.com/marrlab/HistoGPT
Hugging Face: huggingface.co/marr-peng-lab…
carsten marr retweeted
I am thrilled to announce HistoGPT, a vision language foundation model that generates highly accurate pathology reports from gigapixel whole slide images.
medRxiv: medrxiv.org/content/10.1101/…
GitHub: github.com/marrlab/HistoGPT
Hugging Face: huggingface.co/marr-peng-lab…
carsten marr retweeted
🎉 Stoked that @ErnstRoell's first Ph.D. work "Differentiable Euler Characteristic Transforms for Shape Classification" was accepted @iclr_conf #ICLR2024
👉A new, super fast topological layer (based on Euler Characteristic Transform)
#geometry #topology #MachineLearning
🧵1/n
carsten marr retweeted
🎉@iclr_conf: Simplicial Representation Learning w/ Neural k-Forms
#Geometry #MachineLearning #ICLR2024
📜openreview.net/forum?id=Djw0…
TL;DR: Learnable differential forms = ꜰᴇᴀᴛᴜʀᴇ ᴍᴀᴘꜱ for combinatorial complexes (like graphs), enabling classification + more!
🧵1/n
carsten marr retweeted
My new lab at the Institute of AI for Health at @HelmholtzMunich is looking for postdocs! Great projects in AI for discovery of antimicrobial peptides and Computational oncology. RT or apply! jobs.helmholtz-muenchen.de/j…
Great projects, great environment, great supervisor ;-). Pls share / apply!
To future PhD students! My lab offers two super exciting positions to work on AI-driven antimicrobial peptide discovery. Please share and apply. Can't wait to work on this topic with you☺️! Both positions are funded by @ERC_Research. Check the ad here: jobs.helmholtz-muenchen.de/j…
carsten marr retweeted
Interested in #Foundation #Models in #Health / #Medicine?
-> Consider stopping by: April 10, @HelmholtzMunich
3 #keynotes, an exciting scientific program, and networking opportunities await. Registration / abstract submission: events.hifis.net/event/1221/
@fabian_theis @MarrCarsten
carsten marr retweeted
Takeaways:
- finetuning can be achieved with low resources (we used 1 GPU only + batch size 32 was sufficient for training ViT-g)
- even for small datasets finetuning strong CV foundation models might be better than histo-pre-trained feature extractors
3/
carsten marr retweeted
⚡️Want to save compute when training your feature extractor for histopathology?🔬💻#CPath
Study on 1(!) GPU for 2 CRC tasks led by our fantastic master student @roth_benedikt13, co-supervised with @ValentinKoch_
pre-print: arxiv.org/abs/2401.04720
code: github.com/beneroth13/dinov2
1/
carsten marr retweeted
Check out the new promo video of my PhD program 📹👇🏼
Congratulations #EpiCrossBorders! 🇩🇪🏴
@epigeneticsHMGU @HelmholtzMunich @EdinburghUni
Learn more about EpiCrossBorders ! Video featuring our #Epigentics focused international PhD program -joint btw. @HelmholtzMunich and @EdinburghUni - is now online .
👁️ and share:
youtube.com/watch?v=SKSgswYx…
carsten marr retweeted
📢 Our (@KazeminiaSalome, @MarrCarsten) paper on topologically regularized multiple instance learning (MIL) for rare anemia disorder classification is on arXiv!
arxiv.org/abs/2307.14025
1/8
Congrats Scott and @dominik_waibel for this feat!!! A truly inspiring interdisciplinary work with @Meier_Lab between @PioneerCampus and @CompHealthMuc at @HelmholtzMunich
This Post is from an account that no longer exists. Learn more
carsten marr retweeted
Exciting research by @JRohrscheidt and @Pseudomanifold tackles challenges in #datascience caused by violations of the "manifold hypothesis." 👀 See how the novel approach will empower researchers for better decisions in analysing intricate datasets here 👉 bit.ly/gfdrt


























