Pinned Tweet
A Radiologist's Basic Intro to AI and ML. After reading this short book, you will be prepared to read radiology journal articles that showcase AI and ML applications. #ML #AI #radiology #teaching @amazon amazon.com/dp/1092768319/ref…
2
10
FDA Breakthrough Device for two independent companies, Mosaic Clinical Technologies and Aidoc: they analyze a chest radiograph, generate a draft report, and hand it to a radiologist for review, edit and sign-off. linkedin.com/feed/update/urn…
1
26
We have heard the beating of the AI drum for years.
But the pitch has been the same since 2016, and most of us have learned to tune it out.
It is worth it to accept that #AI has made a significant change. #VLM #radiology linkedin.com/feed/update/urn…
77
A Radiologist's Introduction to Foundation Models: The Best Time in History to Be a Radiologist, now available on Amazon. To celebrate the launch, the #Kindle edition will be available free on Amazon through June 18. No code needed.
Link on Amazon: a.co/d/0cn7zUuM
1
125
Editing AI reports can take longer than dictating from scratch.
Dictating a new study follows a natural sequence: scan, perceive, formulate, articulate. Editing an #AI draft reverses it. #vlm #radiology linkedin.com/feed/update/urn…
1
36
If AI can generate the report, what is a radiologist's signature worth? The market is already deciding. Article 14. #vlm #radiology linkedin.com/feed/update/urn…
38
For thirty years, radiology drifted away from consultation. AI may be the forcing function that reverses it. Article 13 of 14 #VLM #radiology linkedin.com/feed/update/urn…
27
Vendors can show us the AUC, the sensitivity, the validation cohort. What we don't know yet is whether the reports are leading to the right clinical decisions. We need to get to outcomes. Article 12 #VLM #radiology linkedin.com/feed/update/urn…
27
New #foundation models are performing opportunistic screening across chest, brain, and abdomen simultaneously, finding things the scan was not ordered for, on studies ordered for entirely different indications. Article 11 of 15 #VLM #radiology linkedin.com/feed/update/urn…
29
Article 10 looks at #TEFCA, the federal framework connecting imaging data nationwide, what it does and doesn't say about AI training, and what Europe built as a comparison. Article 10 of 15. #VLM #radiology linkedin.com/feed/update/urn…
32
Who is liable when #radiology #AI gets it wrong? The legal answer hasn't changed, but the regulatory and contractual architecture defining that answer is shifting fast, and the practices building governance infrastructure now will not be scrambling. #vlm
linkedin.com/feed/update/urn…
1
1
40
Most radiology practices deploying AI have a solid pre-validation process.
Article 8 of my series on AI and #VLMs in #radiology addresses the failure modes that show up after deployment: drift, bias and scanner batch effects.
linkedin.com/feed/update/urn…
18
#Residents have asked me a question a few times this week:
Where do I start learning about #AI in #radiology given all the new models and without getting too technical?
I’ve been building a practical answer to that. #vlm linkedin.com/posts/tyvachon_…
35
Agentic AI in Radiology: From Co-Pilot to Autopilot Article 7 of 14 - AI generates a radiology report, you review it, you decide what happens next.
An #agent is different.
linkedin.com/pulse/agentic-a…
32
Three things will separate the future radiology departments that will thrive from those that won't. Move upstream. Capture the gains. Own a piece of the infrastructure. Article 5 of my ongoing series. linkedin.com/pulse/new-radio…
15
Data from Northwestern Medicine shows generative AI driving up to 40% reading efficiency gains across 11 hospitals in a clinical deployment. The ceiling for CT may be higher. Here is Article 4 of my series on #Radiology in the Age of AI and #VLMs. linkedin.com/pulse/what-woul…
29
Two peer-reviewed frameworks now exist that are more useful than just AUC. Article 3 of 15. linkedin.com/pulse/94-sensit…
9
VLM are drafting reports at 11 hospitals in clinical practice. Article 2 of my series breaks down how Vision-Language Models work, what they can and can't do today, and why the interface we're using to review their output may be the wrong design entirely. linkedin.com/pulse/from-algo…
19
The era of single-task radiology AI is over. A new class of model doesn't just detect findings, it reasons across the entire study, the report, and the patient's history simultaneously. Article 1 of 14 in my new series: Radiology in the Age of AI and VLMs. linkedin.com/pulse/end-algor…
23
JAMA Health Forum: How did quality measures change at federally qualified health centers (FQHCs) in 2020 to 2021 compared with before the COVID-19 pandemic?
"10 of 12 FQHC quality-of-care performance measures declined substantiall…lnkd.in/guKTDxMH lnkd.in/gT2p2GvB
44
Nice way to hear additional author comments on the BMJ Thorax podcast particularly when it comes to lung cancer screening.
Dr. Sam Janes and team address 'small but growing pulmonary nodules' and when to refer in their recent pub…lnkd.in/gwPaQJgh lnkd.in/gDh-k792
54

