🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

🥳 Ziyun Liang making us proud on main stage of #MICCAI24! Introducing her work on Unsupervised Anomaly Segmentation: IterMask2 - a method towards automatic segmentation of "any" brain lesion in MRI paper/code: arxiv.org/abs/2406.02422 Engineering Science, Oxford Oxford BioMedIA, Engineering and Physical Sciences Research Council scholar

A huge thanks to everyone who attended my oral talk and poster at #MICCAI2024 on unsupervised anomaly segmentation on ‘any’ brain lesion. I truly enjoyed the interesting discussions I had with all of you. paper: arxiv.org/abs/2406.02422 University of Oxford Kostas Kamnitsas Engineering Science, Oxford

On out-of-distribution detection. New public data and insights:

🤖⚕️Hermione Warr presents UNSURE workshop 2025 #MICCAI24 a new framework for > auditing quality of generated radiology reports with respect to diagnostic information < Idea: Enforce consistency between semantics estimated from scan & semantics in text report Come have a chat!

Investigating style-transfer and the effect of different preprocessing in MRI, which can amplify/introduce distribution shift if not done appropriately.

🚀 Excited to introduce our work: F³OCUS: Federated Finetuning of Vision-Language Models with Optimal Layer Selection! 🌟 Client-specific layer importance via NTK eigenvalues. Diverse layer selection across clients with meta-heuristic optimization. 🔗 arxiv.org/abs/2411.11912

"How did you make these cool slides?!?!" 🤩 - common audience question each time Harry presented his PhD work. So instead of just giving tips to our lab, he made a short blogpost + video to share widely. Good slides enable more effective explanation of content. Quick tips👇

Useful: Federated-learning works at Neurips24 👇

Towards foundational models for Brain Lesion Segmentation with Multi-Modal MRI, Vol2 from our lab: We show feasible to train 1 model with Federated Learning on clients with different brain lesions and MRI modalities, with performance similar to centralised. Oral WACV

🚀 Attending #AAAI2025 in Philadelphia? Come visit posters #390 & #991 from University of Oxford Noble Lab today! 📍 12:30-2:30 PM | #ComputerVision & #AI for #Social #Impact Sections 🩺 Presenting two first-author papers on #Multimodal #FederatedLearning for #HealthcareAI! (1/n)

Radiomics uses data from routine radiological images to noninvasively profile tumors, predict cancer behavior, and improve treatment planning for liver cancers. ja.ma/3QD8oVW Areeb Kostas Kamnitsas Oxford Hepatopancreatobiliary Surgery Department

Are you at Arizona for #WACV2025 and interested in #FederatedLearning for Multi-Modal MRI? Find Felix Wagner at the poster session this evening!

Everything new is also old. This from my 1984 PhD thesis: "AI is an experimental science, yet the complexity of its programs and problem domains often makes the interpretation of results very difficult. Programs often contain so many components and parameters that limitations on

Great work from Jiayuan, to be presented at ICCV. Learns prefered "style" of segmentation of user 👇

If one argues again that research on AI reliability or regulation is BS and we should just "accelerate" to next product, remind them one of the best teams tried to finetune their AI and it turned Hitler in production unexpectedly. Then imagine this happening to a healthcare AI.