Leon Chen (@transcranial) header
Leon Chen (@transcranial) profile picture

Co-founder @md_ai_ - AI-powered reporting and annotation for radiology. Software/ML/AI engineer, physician, hobbyist electronic musician.

Seattle, WA
Joined May 2009
Photo by Leon Chen (@transcranial)Photo by Leon Chen (@transcranial)
Leon Chen retweeted
My dear friends, I’d like to show you my latest toy project Introducing Handwired: turn your regular JavaScript code into a neural net! - no training - exact outputs vs the JS program - fully differentiable All runnable in a browser, without GPU
459
1,024
469
11,359
1,471,622
Leon Chen retweeted
Announcing GPT-4, a large multimodal model, with our best-ever results on capabilities and alignment: openai.com/product/gpt-4
2,005
15,927
5,348
59,585
12,483,579
New preprint! Q: How accurate is GPT-3 at predicting Dx and triage advice compared to Harvard docs and to an average person Googling their symptoms? For Dx, GPT-3 is much more accurate than your average person and almost as accurate as the Harvard docs! For triage, not so much.
7
21
2
82
25,182
Leon Chen retweeted
Great work! Glad we could contribute in a small way to this massive effort.
Excited to announce the @hiti_lab's release of the EMory BrEast Imaging Dataset (EMBED) on @AWSOpenData containing 20,000 women and over 500,000 mammographic images. @judywawira @ImonBanerjee6 @EmoryRadiology #breastcancer #AI registry.opendata.aws/emory-…
2
4
1,205
Excited to announce the @hiti_lab's release of the EMory BrEast Imaging Dataset (EMBED) on @AWSOpenData containing 20,000 women and over 500,000 mammographic images. @judywawira @ImonBanerjee6 @EmoryRadiology #breastcancer #AI registry.opendata.aws/emory-…
3
19
4
76
12,902
📣 CRFM announces PubMedGPT, a new 2.7B language model that achieves a new SOTA on the US medical licensing exam. The recipe is simple: a standard Transformer trained from scratch on PubMed (from The Pile) using @mosaicml on the MosaicML Cloud, then fine-tuned for the QA task.
40
315
54
1,468
426,735
Fascinating: glioblastomas form scale-free networks, with small populations of activating periodic tumor cells in hubs. Target the network hub cells; bring down the network.
This Post is from an account that no longer exists. Learn more
1
1
3
1,137
[new post] "4 questions on AI" aashay.substack.com/p/4-ques…
4
4
33
Pretty impressive ability to interpret clinical shorthand.
2
1
6
Wherein ChatGPT very confidently hallucinates non-existent academic references. If we could hook up PubMed to ChatGPT and have it produce actual references, this could be incredibly useful.
Asked ChatGPT an interesting onc clinical question. To my surprise received a definitive answer but when I asked for the reference: it manufactured a fake JAMA Oncology citation and Phase 3 trial. 😂
3
🥂🥂🥂IT’S OUT!🥂🥂🥂 Pitfalls of image analysis metrics 2.0: arxiv.org/abs/2104.05642 More than 50 illustrations compiled by image analysis experts from over 60 institutions. What a team effort! Initiated by #HelmholtzImaging, @ProjectMONAI and @MICCAI_Society and… (1/n)
7
106
14
269
Very excited to have 2 new papers in press today in Lancet Digital Health, alongside an editorial from the journal highlighting our work. I am immensely proud of the work we have done here and honestly think this is the most important work I have been involved in to date 🥳 1/7
10
64
10
324
Leon Chen retweeted
You don't want your pneumonia models running on bone x-rays. Understanding high-level image content is important for building datasets, but also a crucial first step in successful medical AI deployment. Check out the @kaggle competition:
NEW InClass @kaggle competition launched! Help a public Brazilian university deploy its #AI models to carry out AI validation by sorting body parts on X-rays. @unifesp @ddiunifesp @UNIFESP_Rad #dasa kaggle.com/competitions/unif…
3
7
#ADPKD affects >12M people worldwide 🌎 "Deployed Kidney Segmentation for PCKD MRI" | @WeillCornell | @rogosin | @WCMRadiology was just published at @Radiology_AI @RSNA! • Paper: lnkd.in/eaciYs4c • Code & Models: lnkd.in/e2kSK8X9 #pytorch #python #ai #mri
3
11
4
34
Oh no, neural networks can feel pain too?!
Impressed by @CohereAI large model text generation in a medical domain without any finetuning: Seed in blue. Generated text in bold.
6
Excited to announce the Emory BrEast imaging Dataset (EMBED) developed by the @hiti_lab, representing 116,000 women including 50,000 African Americans, 3.5M images, and 40,000 ROIs tied to imaging descriptors and pathologic outcomes bit.ly/3rHJsB7
4
29
5
105
We’ve made public what we believe is the best and most extensively evaluated AI system for breast MRI. It's as good as radiologists, works across all demographics/cancer subtypes--and may reduce patient biopsies too! Pre-print: medrxiv.org/content/10.1101/…, and a short thread below.
12
60
7
219
Introducing #AlphaCode: a system that can compete at average human level in competitive coding competitions like @codeforces. An exciting leap in AI problem-solving capabilities, combining many advances in machine learning! Read more: dpmd.ai/Alpha-Code 1/
159
2,129
942
7,503