Emese Sükei

Excited to present our work on 3D retinal OCT modeling with xLSTM at #ISBI2025 today! xLSTM captures spatial-temporal structure across OCT volumes—pushing beyond slice-based models for better retinal disease analysis. Come find me at the #poster session!

🚀 Published in @SciReports: Our multimodal #ContrastiveLearning study uses 2D & 3D retinal images to enhance prediction accuracy in #Ophthalmology. It also shows that lower-cost fundus imaging can still deliver clinical insights. Learn more! 👀🔍 🔗rdcu.be/dZa2N

COMBINATORIAL OPTIMIZATION with DIFFUSION MODELS. presented at #ICML2024 this week!

A list explaining key ML concepts and papers: github.com/dair-ai/ML-Pap… v/@omarsar0

📢 Exciting News! 📚 Our article on AI-driven patient functionality monitoring using #digitalphenotyping is finally out! 🌐Discover how we harness multimodal data for accurate evaluations📊 Read on ScienceDirect: sciencedirect.com/science/articl…

After very rich three-days program at #MIDL2023 there's more to come: #MONAI workshop on #DeepLearning for #MedicalImaging 📍 Online and in-person 📅 Thursday 13 July, 9:00-13:30 (Nashville time, UTC-5) 🌐monaimidlmeetup.splashthat.com

And on the last day of #MIDL2023 we can hear a series of talks about #segmentation and #ComputerAidedDiagnosis. Tune in 👇 youtu.be/6T5TCBJaYl4

Up now: keynote by Dr. Yesha on integrating autonomous AI into the clinical workflow.

The 2nd day of #MIDl2023 continues strong with inspiring oral sessions on #selfsupervisedlearning and #semisupervisedlearning, followed by a keynote by Dr. Yesha. Follow along 👇 youtu.be/S1z7zbSDmmI

Excited to chat about our work on #multimodal #contrastivelearning in retinal imaging. Come by!

#reproducibility is a commonly encountered issue in #DeepLearning research. Here's a great #checklist proposed by Attila Simko et al that addresses this: midl.io/reproducibilit…

1st day of #MIDL2023 kicking off with strong with #oralsessions on #segmentation, #unsupervisedlearning and #graphlearning, and of course #postersessions.

Another bias: If you're a non-native English writer, GPT detectors will frequently classify your writing as #AI-generated cell.com/patterns/fullt… @james_y_zou @liang_weixin @Stanford

Excited to be travelling to Nashville for #MIDL2023 to showcase our work on #multimodal #ContrastiveLearning for #retinalimaging. Come by for a chat during the poster sessions tomorrow! Short paper 👇 openreview.net/forum?id=newla…

Statistical modeling from first principle. But with memes and GIFs. Let's start.

1. 💡The #MSCA play a crucial role in supporting the EU’s steps towards #genderequality in #research and #innovation. ♀️ On the international #GirlsinICT Day, we celebrate the talent of our women #researchers & creators. Read more: europa.eu/!QT6gfp See thread👇

"And then we just cancel this infinity with that one and we get the answer"

Can AI tell us when it can predict better than humans and when humans are better? In our AISTATS23 (oral) with @MITIBMLab paper we build AI models that 1) complement humans and 2) can defer to humans when necessary arxiv.org/abs/2301.06197

Geometric Learning concepts (Equivariance, Invariances) carried over to interpretability-methods: if a neural net is invariant to a certain transformation of the input, also the feature-importance should be invariant. Overly complicated notation, imho arxiv.org/abs/2304.06715