Good news! CPH's @amalaa.bsky.social and colleagues @roxanadaneshjou.bsky.social and Tom Harvitgsen have received a grant from the NIH Director's office for work on participatory mutimodal AI and editable foundation models to improve diagnosis and tx. @stanforddeptmed.bsky.social
What happens in SAIL 2025 stays in SAIL 2025 -- except for these anonymized hot takes! 🔥 Jotted down 17 de-identified quotes on AI and medicine from medical executives, journal editors, and academics in off-the-record discussions in Puerto Rico
irenechen.net/sail2025/
Real world data (RWD) analyses need diverse expertise in medicine/stats/causality. As a result, RWD causal analysis is slow and prone to error. AI tools can help.
Perspective in NEJM AI w/Maya Petersen, @amalaa.bsky.social, Chris Holmes, Mark van der Laan
ai.nejm.org/stoken/defau...
Fourth lecture of CPH 290: Julia Adler-Milstein provided an overview of the Division of Clinical Informatics and Transformation at UCSF and their approach for deploying and evaluating AI models within the health system.
youtu.be/lBMPJ8R3Zd4?...
In Poster Session 4 (4:30 pm - 7:30 pm PST), Lars van der Laan will present our work on combining model calibration and prediction intervals by integrating Venn-Abers into conformal prediction.
If you're attending #NeurIPS2024 be sure to stop by our lab's posters today!
In Poster Session 3 (11 AM - 2 PM PST), Frances Dean will present our work on learning physics-based models of cardiac hemodynamics using non-invasive data to simulate disease and treatment outcomes.
In our third CPH 290 lecture, we hosted the CTO and co-founder of OpenEvidence, one of the most widely used AI-based medical search platforms. He shared how his team uses LLMs + other tools to enhance the visualization, presentation and synthesis of clinical evidence.
youtu.be/8-ui5UsD5OU?...
I'm recruiting PhD students for Fall 2025! CS PhD Deadline: Dec. 15th.
I work on safe/reliable ML and causal inference, motivated by healthcare applications.
Beyond myself, Johns Hopkins has a rich community of folks doing similar work. Come join us!
Congratulations 🎉🎉
Probably should be making pies, but it’s much more fun to blog about calibration.
This one gets into more practical issues w/recently proposed solutions like multicalibration, discussing results from a recent empirical evaluation
statmodeling.stat.columbia.edu/2024/11/26/p...
In the second lecture of CPH 290, @jameszou.bsky.social gave an overview of the evaluation practices and regulatory landscape of medical AI. We posted the lecture on Youtube:
youtu.be/f8wr_Md6xnk?...
Would love to be added. Thanks!
It's tough to gain visibility as a young researcher, and it's job market season! Are you a theoretical computer science PhD/postdoc on the job market?
I don't have a crazy juge audience but I'll try to help a bit: fill this form, and I'll tweet your pitch and info!
docs.google.com/forms/d/e/1F...
Would love to be included. Thanks!
Most clinical AI models are only evaluated for accuracy in retrospective datasets w/o real-world evaluation of clinical utility. In the the first lecture of our CPH 290 class, Julian Hong (UCSF) discussed a UCSF RCT of AI-driven risk prediction in radiotherapy acute care.
youtu.be/GAf4yc61gv8?...
There have been a few studies on a larger RCT that looked more deeply into automation neglect/bias in human-AI interactions in radiology (though no LLMs were involved in the treatment arm): www.nature.com/articles/s41... , www.nber.org/papers/w31422
Hello BlueSky world! 🦋 We're recruiting PhD students at UC Berkeley and UCSF!
If you're interested in machine learning for healthcare, statistics/causal inference, or medical vision-language models, we'd love to hear from you: docs.google.com/forms/d/e/1F...
