Joseph Cho
@joseph_cho1
Followers
118
Following
182
Media
9
Statuses
48
We are excited to announce 🔥EchoAI-Peds 🔥, the first multi-task deep learning model for pediatric #echofirst analysis. It's been a pleasure to lead this alongside @mrudangm14 under the guidance of @hiesingerlab and in close collaboration with @jolleylab. Link below⬇️ [1/n]
1
6
13
Want to join @GoogleDeepMind as a Student Researcher for 6 months starting in January (PhD students)? 🧬 The project will be focused on AI for Science and AI for Cancer! Send me a DM and also apply below👇
69
163
2K
We've been busy at the @HiesingerLab! Many thanks to all co-authors for their efforts and to the @NIH NHLBI for their support! #echofirst #pediatrics
We are excited to announce 🔥EchoAI-Peds 🔥, the first multi-task deep learning model for pediatric #echofirst analysis. It's been a pleasure to lead this alongside @mrudangm14 under the guidance of @hiesingerlab and in close collaboration with @jolleylab. Link below⬇️ [1/n]
0
1
6
Congratulations to @joseph_cho1, @mrudangm14 and team on their pre-print describing the first multi-task ML model for interpreting pediatric echocardiograms! @HeartCare4Kids @CHOP_Research
We are excited to announce 🔥EchoAI-Peds 🔥, the first multi-task deep learning model for pediatric #echofirst analysis. It's been a pleasure to lead this alongside @mrudangm14 under the guidance of @hiesingerlab and in close collaboration with @jolleylab. Link below⬇️ [1/n]
0
1
4
And a big thank you to some of our amazing co-authors: @dhamank24, @MatthewMDuda, @DahlanAdil, @arav_krishnan, @MattLeipzig, @rohanshad, @cyrilzakka, and @curtlanglotz. Finally, this work would not be possible w/o support from the NHLBI. [7/n]
0
0
4
Notably, our model demonstrated robust performance across patient age, patient sex, and studies with varying number of videos. [6/n]
1
0
2
Pediatric labels matter!!! Our model achieved macro-averaged AUROC values of 0.91 and 0.89 on the internal and external test sets, respectively. Moreover, our model significantly outperformed adult echo foundation models trained on substantially larger datasets. [5/n]
2
0
2
We developed our model on 700,000+ videos from 11,000+ studies @StanfordMed. We assessed model efficacy on an internal, held-out dataset. We also assessed model generalizability on a spatially and temporally distinct patient cohort at @ChildrensPhila. [4/n]
1
0
2
Existing pediatric echo models are limited to single tasks and specific views. To address this, we develop a ViT capable of simultaneously detecting 28 CHDs, structural and functional abnormalities, repairs, and interventions directly from complete pediatric echo studies. [3/n]
1
0
2
In a bit of a bittersweet moment, this week will be my last at Hugging Face 🤗 for the time being. I will really miss my time working with some of the most talented and kindest folks in the industry, led by a trio of incomparable leaders. Very excited to announce what’s next!
13
3
113
Our paper on autonomous scientific research is accepted to Findings of #EMNLP2025! 🎉 We introduce Agent Laboratory, a framework that accelerates scientific discovery by teaming human researchers with LLM agents.
16
235
1K
📢🩻 Many new updates for MedGemma, for more details check out: - our technical report ( https://t.co/Pji0cCrIa3) - our blog post ( https://t.co/2DcQWtFid3): - Model Card ( https://t.co/FG1YXVMUqC) - Github ( https://t.co/h9YEUZbQ5x)
lnkd.in
This link will take you to a page that’s not on LinkedIn
Introducing new models for research & development of health applications: MedGemma 27B Multimodal, for complex multimodal & longitudinal EHR interpretation, and MedSigLIP, a lightweight image & text encoder for classification, search, & related tasks. → https://t.co/I318jVmsYD
1
7
43
🚨New Post 🚨 If you're like me and you've been putting off learning about GRPO, now is as good a time as ever, especially with Deepseek-R2 just around the corner!
1
5
32
I’m making the model weights for our foundational CMR vision encoders (see pinned post) available free for academic use. Paper still in peer review, but can’t wait to see what everyone builds with this! Git: https://t.co/k2Qgl3Mbtd Weights: https://t.co/xhlWdhlCfa
#HNY2025
2
7
38
Had a great time talking about finite elements and generative AI in medicine at @BmeSjsu Pathways Seminar last week! Many thanks to my dear friend Prof. @EllaSugerman for the invite and to my @HiesingerLab mates @joseph_cho1 and @cyrilzakka for sharing their great results 🔥
0
3
32
Excited to share our latest research progress (joint work with @DrYangSong ): Consistency models can now scale stably to ImageNet 512x512 with up to 1.5B parameters using a simplified algorithm, and our 2-step samples closely approach the quality of diffusion models. See more
arxiv.org
Consistency models (CMs) are a powerful class of diffusion-based generative models optimized for fast sampling. Most existing CMs are trained using discretized timesteps, which introduce...
Introducing sCMs: our latest consistency models with a simplified formulation, improved training stability, and scalability. sCMs generate samples comparable to leading diffusion models but require only two sampling steps.
22
88
684
Due in part to the large number of runaway takes in medical AI, I'm launching a regular thread to spotlight noteworthy and more leveled research papers in the space and/or related fields with promising applications in healthcare. RTs appreciated. Let's begin!
5
15
95
I finally released my new video on YouTube about Diffusion Models / Score-Based Generative Models. https://t.co/1G97UOZQH2 Literally planned this for a year and put so much work in. I think this approach to diffusion models is so intuitive and highly recommend giving that a
18
87
676
Excited to release HuggingChat 💬 - a native macOS app that brings powerful open-source language models straight to your desktop - with markdown support, web browsing, code syntax highlighting and much more!
56
195
2K