New research from @ucsfci2.bsky.social fine-tunes foundation segmentation models across five musculoskeletal anatomies, with 83% of scores reaching a Dice coefficient of 0.90 or above. Check it out ➡️ bit.ly/4fENWRM

A text-only AI alert can prompt more scrutiny than an annotated thumbnail. In a new review, @ucsfci2.bsky.social's @alitejanimd.bsky.social, @shadowdoc.bsky.social, @drdremdphd.bsky.social, @monganmd.bsky.social & collaborators examine how presentation shapes automation bias. bit.ly/4fgrbT1

Generating quantitative biomarkers from MRI typically requires experts to manually delineate structures across hundreds of slices. @ucsfci2.bsky.social researchers published a new approach in npj Digital Medicine that automates that process across five anatomies. bit.ly/4fENWRM

3 DICOM formats carry AI results to radiologists, from static overlays to editable structured reports. @ucsfci2.bsky.social's @alitejanimd.bsky.social, @shadowdoc.bsky.social, @drdremdphd.bsky.social, @monganmd.bsky.social & collaborators break down what each format allows. bit.ly/4fgrbT1

Imaging volume is growing. Radiologist shortages are worsening. Bunkerhill Health shared with @ucsfci2.bsky.social what they've learned building foundation models for chest X-ray & body CT. intelligentimaging.ucsf.edu/news/what-la...

Scaling radiology #AI is harder than it looks. At the recent @ucsfci2.bsky.social SRG Pillar Meeting, Bunkerhill Health's Abhijeet Shenoi & George Hu broke down what works, what doesn't, & what comes next for radiology foundation models. bit.ly/4xY1S0z

Congratulations to @ucsfci2.bsky.social's Dr. Alexandra Gersing on receiving a 2026 RSNA Research & Education Foundation Research Seed grant for AI-enabled MR muscle biomarker research exploring the role of muscle in low back pain.

New research in Radiology: Imaging Cancer highlights how AI as a second reader improved mammography sensitivity by 8.4%, catching more interval and future cancers. @ucsfci2.bsky.social's Dr. Maggie Chung breaks down the findings: doi.org/10.1148/ryca...

In a new #AJR video features @ucsfci2.bsky.social's #ClinicalFellow Ali Tejani & #NYUGrossman's Dr. Marc Triola discuss ethical #AI integration in medical education, AI literacy & building curricula that can keep pace with rapid change. doi.org/10.2214/AJR....

What predicts poor cardiac MRI quality before imaging begins? @ucsfci2.bsky.social's Dr. Jae Ho Sohn & collaborators found two independent factors: cognitive/communication impairment & respiratory compromise. onlinelibrary.wiley.com/doi/10.1002/...

Current meningioma measurement relies on 2D slices & manual diameter checks. @ucsfci2.bsky.social's Pierre Nedelec & Dr. @drdremdphd.bsky.social developed a volumetric #AI tool that tracks tumor changes longitudinally, with confidence intervals built in. bit.ly/4f32JWC

Poor cardiac MRI quality means repeat scans & delayed decisions. @ucsfci2.bsky.social's Dr. Jae Ho Sohn & collaborators used an LLM to identify pre-imaging factors linked to poor image quality: cognitive/communication impairment & respiratory compromise. bit.ly/4aiI9hJ

Meningiomas grow slowly, making subtle changes easy to miss. @ucsfci2.bsky.social's Pierre Nedelec & Dr. @drdremdphd.bsky.social built a "one-click" AI tool for 3D volumetric tumor tracking across time. Radiologists can now see growth trends immediately after MRI export. bit.ly/4f32JWC

Bunkerhill Health is building foundation models for radiology, & two of the people behind that work are joining us to share what they've learned. Join us for the upcoming SRG Pillar Meeting to hear more from Abhijeet Shenoi & George Hu. bit.ly/4tM79EZ

If you're curious about what it actually takes to build & deploy foundation models in radiology, join us for the upcoming SRG Pillar meeting. Abhijeet Shenoi & George Hu from Bunkerhill Health will talk through that process on June 8. bit.ly/4tM79EZ

Faster, smarter MRSI reconstruction ➡️ @ucsfci2.bsky.social's Dr. Janine Lupo presents a multi-domain #DeepLearning approach with metabolite-aware loss this afternoon at #ISMRM2026. Find it in the Deep Learning Meets k-Space poster session. @ismrm.bsky.social bit.ly/3QGOfBP

Starting off this morning at #ISMRM2026 with @ucsfci2.bsky.social's David Mikhael is presenting on network states linked to the therapeutic efficacy of deep brain stimulation for OCD. @ismrm.bsky.social radiology.ucsf.edu/research/ismrm

No public dataset previously existed for researchers building computational models around primary CNS lymphoma. #RadRes Michael Romano introduced the first at April's SRG Pillar Meeting.

UCSF Residents Present Research on Multilingual Patient-Friendly Radiology Reports and a First-of-Its-Kind CNS Lymphoma Dataset

UCSF radiology residents Yannan Yu, MD, and Michael Romano, MD, PhD, presented during the recent UCSF ci2 SRG Pillar Meeting. Yu presented her research on using large language models (LLMs) to…

bit.ly

This morning at #ISMRM2026, former @UCSF_Ci2 #postdoc Rupsa Bhattacharjee (@Rupsa_B) presents on the basic principles of quantitative MSK MRI. radiology.ucsf.edu/research/ismrm @ISMRM

Fetal ventriculomegaly is common & usually benign, but some cases progress to hydrocephalus. A new #DeepLearning method from @ucsfci2.bsky.social predicts the need for postnatal intervention with 86% sensitivity & 100% specificity.

New Deep Learning Approach Predicts Need for Postnatal Intervention in Fetal Ventriculomegaly

A team of researchers from the University of California, San Francisco's Department of Radiology and Biomedical Imaging and Center for Intelligent Imaging (ci2) developed a deep learning method for…

bit.ly

96.4% of patients surveyed said including simplified radiology reports alongside originals would be beneficial. #RadRes Yannan Yu presented her research on multilingual, patient-friendly reports at the recent @ucsfci2.bsky.social SRG Pillar Meeting.

UCSF Residents Present Research on Multilingual Patient-Friendly Radiology Reports and a First-of-Its-Kind CNS Lymphoma Dataset

UCSF radiology residents Yannan Yu, MD, and Michael Romano, MD, PhD, presented during the recent UCSF ci2 SRG Pillar Meeting. Yu presented her research on using large language models (LLMs) to…

bit.ly

#RadRes Yannan Yu & Michael Romano presented research at the recent @ucsfci2.bsky.social's recent SRG Pillar Meeting on multilingual patient-friendly radiology reports & a first-of-its-kind CNS lymphoma dataset. Read more: bit.ly/4el9geX

Which fetuses with ventriculomegaly will need surgical intervention after birth? A new @ucsfci2.bsky.social study used #DeepLearning to answer that question with an AUC of 0.97. bit.ly/4dI5n3r

New Blog Post ➡️ "New Deep Learning Approach Predicts Need for Postnatal Intervention in Fetal Ventriculomegaly."

New Deep Learning Approach Predicts Need for Postnatal Intervention in Fetal Ventriculomegaly

A team of researchers from the University of California, San Francisco's Department of Radiology and Biomedical Imaging and Center for Intelligent Imaging (ci2) developed a deep learning method for…

intelligentimaging.ucsf.edu

Radiology reports should work for all patients regardless of language. During our next SRG Pillar meeting, #RadRes Yannan Yu presents on multilingual patient-friendly reports using LLMs. #RadRes Michael Romano on the UCSF PCNSL imaging & genomics resource.

SRG Pillar Meeting with Yannan Yu and Michael Romano | intelligent imaging

Meeting Details Zoom Details: Meeting ID: 922 7067 8528 Password: 845445

bit.ly

Can #AI help predict which fetuses with ventriculomegaly will need postnatal intervention? Research from @ucsfci2.bsky.social's Dr. @drdremdphd.bsky.social & colleages introduces a #DeepLearning method for fetal brain MRI that shows strong results. www.ajnr.org/content/47/3...

Join us April 13 for the next @ucsfci2.bsky.social Pillar meeting. Two talks from @ucsfimaging.bsky.social #RadRes: Yannan Yu on using #LLMs for multilingual patient-friendly radiology reports & Michael Romano on the UCSF PCNSL imaging & genomics resource. bit.ly/3Qig4Qo

Members of @ucsfci2.bsky.social were featured in #RadiologyToday for their work on the UCSF-GE HealthCare Care Innovation Hub, including Drs. Sharmila Majumdar, Andreas Rauschecker, Peder Larson & Duan Xu. www.radiologytoday.net/archive/rt_M...

Join us April 13 for the 2026 Frontiers in Computational Precision Health Conference. @ucsfci2.bsky.social's Drs. Maggie Chung & Adam Yala will be among the panelists tackling the hard questions about health AI.

At @ucsfci2.bsky.social, we're moving beyond just looking at the pixels on a scan. By using medical-grade #LLMs, like UCSF Versa, we can instantly analyze a patient's clinical history to tailor cardiac MRI & CT protocols & interpretation in real-time. #NationalHeartMonth