David Ouyang, MD @David_Ouyang Twitter profile

Pinned Tweet

David Ouyang, MD

2 months

Congrats to @SarnoffCardio fellow and @StanfordMedRes @amey_vrudhula for publishing in @CircAHA ! Using deep learning, we view classify and quantify MR severity in #echofirst . Currently the only such project with open code and model weights. Paper: Code:

Amey Vrudhula

@amey_vrudhula

8 months

Excited to announce EchoNet-MR, our #AI model for high throughput #echofirst MR screening.

1

6

29

3

7

67

Last Seen Profiles

@VixMyers

@sultanheybars

@fat_w9

@hebe226

@AscotSportsBar

@BinorRaja

@badideaarix

@daisuki_ai_ai

@BhawnaSharma_05

@pencitaemanedut

@fayemac_x

@bolisypaletass

@tawu_k

@georgedws

@sezkez_

@qh_iil

@SRavaioni

@cv_girlsbball

@UndeadzOfficial

@KKlSMET

@monoaminoxidasa

@Louisemac

@Sexual_Arabic18

@MA_NEEGUS

@stefangsaenger

@olliep71

@Cebo_Rodriguez

@PavithrenM96669

@Rico330420

@kuveytturkpys

@omar76211546

@lanceclerc

@ModaUK_Ltd

@datajournalism

@JoArcher7

@DikySitepu

David Ouyang, MD

@David_Ouyang

1 year

We are excited to announce that our blinded randomized clinical trial of AI LVEF (EchoNet-Dynamic) has been published in @nature ! @bryandhe @james_y_zou @StanfordDBDS @CedarsSinaiMed 1/n

33

134

462

David Ouyang, MD

@David_Ouyang

3 years

Excited to report some new findings! Additional analysis ongoing.

37

2

432

David Ouyang, MD

@David_Ouyang

5 years

1/7 We’d like to share something interesting for machine learning in cardiology. We trained a neural network, EchoNet-Dynamic, to interpret #echofirst videos to trace the left ventricle and estimate ejection fraction. @bryandhe @euanashley @james_y_zou The results are striking:

20

154

392

David Ouyang, MD

@David_Ouyang

1 year

📢Excited to announce EchoCLIP, or vision-language foundation #AI model for #echofirst interpretation. Using over 1,000,000 cardiac ultrasound videos paired with expert cardiologist interpretations, we enable #echofirst semantic search and zero-shot predictions. EchoCLIP can

17

88

341

David Ouyang, MD

@David_Ouyang

2 years

🚨🚨🚨🚨🚨🚨 We are REALLY excited to present the results of EchoNet-RCT, the first BLINDED RANDOMIZED clinical trial of #AI technology in cardiology at #ESCCongress 2022 in TWO days. Late Breaking Clinical Trial Hot Line Session 3 Highlights of trial design include: 1/🧵

8

68

314

David Ouyang, MD

@David_Ouyang

3 years

In anticipation of #ASE21 this weekend, we'd like to share this AI on #EchoFirst . This frame-to-frame measurement of PLAX was fully automated by EchoNet-LVH. More examples and preprint soon! @ASE360 @CedarsSinai

13

70

310

David Ouyang, MD

@David_Ouyang

5 months

Excited to share our #echofirst foundation model is now published in @NatureMedicine ! Trained on 1M @CedarsSinai A4C echo-report pairs, EchoCLIP has strong performance on wide range of cardiac diagnostic benchmarks, including LVEF in @StanfordMed videos (MAE 7.1%).

Eric Topol

@EricTopol

5 months

What happens when you use >1 million echo video clips with expert interpretation to train an #AI ? Nice advance by @David_Ouyang and colleagues @CedarsSinai @CedarsSinaiMed @NatureMedicine

3

61

218

21

61

279

David Ouyang, MD

@David_Ouyang

3 years

1/8 Using deep learning, we design and implement an AI pipeline for rare disease screening. Published today in @JAMACardio . Led by Grant Duffy, @Paul_P_Cheng and @yuanneal , co-supervised by Dr. Susan Cheng.

8

88

271

David Ouyang, MD

@David_Ouyang

5 years

It’s alive! And no human traced this. Teaser for what’s coming soon!

23

71

270

David Ouyang, MD

@David_Ouyang

5 years

We are excited to announce that EchoNet-Dynamic, our video-based deep learning model for interpreting echocardiogram videos have been published in @nature . @bryandhe @euanashley @james_y_zou @Stanford 1/8

21

65

246

David Ouyang, MD

@David_Ouyang

3 years

Since our paper has been published, I’ve received many questions about Supp Table 3, a critical experiment demonstrating the strength of our pipeline, and why I believe MANY published AI models in medicine are WRONG or at least won’t work well in practice. 1/n A thread🧵:

David Ouyang, MD

@David_Ouyang

3 years

@euanashley @Ron_Witteles @JigPatelMDPhD 5/8 As a screening test, we prioritize specificity over sensitivity by training our model with negative examples from other forms of LVH. This results in high specificity / not ‘overfitting’ on just identifying hypertrophy. Lower reported AUCs, but better performance in practice.

1

4

15

3

43

150

David Ouyang, MD

@David_Ouyang

7 months

Excited to present work out today at @NEJM_AI with @amey_vrudhula @jwestonhughes @yuanneal . We show that question formulation is important medical AI. In developing AI-ECG models, we show that choosing a REGRESSION task outperforms a CLASSIFICATION task for for low EF.

9

29

144

David Ouyang, MD

@David_Ouyang

1 year

Tremendous anchoring effect with AI, even the best radiologists were wrong >50% when AI gave the wrong suggestion. Very much suggests blinding and randomization necessary to truly evaluate these systems.

Laura Heacock, MD

@heacockmd

1 year

The other @radiology_rsna paper is on one of my favorite topics in AI: automation bias. Do we inherently rely too much on these smarter tools? Dratsch and Chen et al evaluated automation bias in mammography.

4

13

32

2

27

120

David Ouyang, MD

@David_Ouyang

2 years

We're excited to announce the release of EchoNet-Peds, an AI model for #EchoFirst assessment of LVEF for pediatric populations. Lead by @bryandhe and Charitha Reddy @StanfordChild s https://www.onlinejase.com/article/S0894-7317(23)00068-8/fulltext 1/n

2

25

116

David Ouyang, MD

@David_Ouyang

1 year

Healthcare systems, your vendors are eating your lunch and taking your data. At this point, Epic should pay hospitals for using its EHR. I’m sure its technically legal but there’s no chance EpicGPT isn’t finetuning on patients who don’t realize their data is used this way.

Pearl Freier

@PearlF

1 year

Privately held healthcare software company Epic has said that 78% of all patients in US hospitals/&related practices use their medical records software. They announced GPT-4 partnership w/ Microsoft at #HIMSS23 & plan to integrate generative AI into Epic's

20

67

181

7

30

108

David Ouyang, MD

@David_Ouyang

5 years

We're excited to announce our work predicting phenotypes from echocardiogram images with deep learning was published today! @GhorbaniAmirata @abidlabs @bryandhe @jonc101x @HeartBobH @euanashley @james_y_zou

9

30

101

David Ouyang, MD

@David_Ouyang

2 years

These predatory journals are getting more and more aggressive…

16

7

103

David Ouyang, MD

@David_Ouyang

8 months

The precision (and imprecision) of medical testing is one of my favorite teaching points in the CCU. Is there a clinical change when the LVEF goes from 52% to 48%? What about when the RVSP goes from 30mmHg to 35mmHg? MVO2 67% to 62%? In this issue of @JournalASEcho , with

1

11

94

David Ouyang, MD

@David_Ouyang

2 years

We are excited to present our prospective blinded randomized clinical trial of AI vs sonographer initial interpretation of LVEF in the #LBCT HOTLINES Session 3 of #ESCcongress 2022 on August 27. #WeNeedRCTs

2

11

91

David Ouyang, MD

@David_Ouyang

5 years

At @NeurIPSConf #ML4H , we present our new, soon to be available, public EchoNet-Dynamic dataset of 10k deidentified and labeled echocardiograms. @bryandhe @euanashley @james_y_zou . Made possible by @mattlungrenMD @StanfordAIMI

3

20

89

David Ouyang, MD

@David_Ouyang

4 years

The time has passed by incredibly quickly, but this is my last week of @StanfordCVFellows ! It’s been an incredible journey mentored by @joshuawknowles @EldrinL @HeartBobH and I am forever grateful. 1/n

14

3

89

David Ouyang, MD

@David_Ouyang

2 years

Grateful for this tremendous honor and thanks to support from @james_y_zou @euanashley @StanfordDeptMed Echo Lab, as well as tremendous insight and infrastructure from @StanfordAIMI @curtlanglotz @mattlungrenMD 1/

J. Brian Byrd

@jbrianbyrd

2 years

David Ouyang, M.D. ( @David_Ouyang ) of @StanfordMed wins the Early Career Research Symbiont Award this year! David wins for his amazing contribution, EchoNet-Dynamic, the largest publicly released dataset of echocardiogram videos. Check it out! 2/

2

7

22

1

6

81

David Ouyang, MD

@David_Ouyang

10 months

We are thrilled to be funded by the @MooreFound and @KPDOR 's AIM-HI initiative to study the deployment of AI to screen for cardiac amyloidosis working with an excellent team of physician scientists across four diverse sites. @Paul_P_Cheng @tygluckman @FarazAhmadMD @SmidtHeart

6

9

76

David Ouyang, MD

@David_Ouyang

3 years

Had a great time presenting at Grand Rounds for the Department of Cardiology! Really optimistic for the future of Cardiovascular AI at @CedarsSinai , going from development to deployment. Exciting news to come later this month! #echofirst

1

78

David Ouyang, MD

@David_Ouyang

9 months

Excited to contribute to a great group lead by @zakkohane to showcase impactful work in clinical AI! AI will most certainly change healthcare - the onus is on clinicians and researchers to understand how and why! @AndrewLBeam @arjunmanrai @DrXiaoLiu @lhpeng @charlottehaug

Cara Martinez, MASPR

@Cara_PR_Comms

9 months

Congrats to @David_Ouyang , one of @SmidtHeart 's leading experts in #AI , for being named a deputy editor of @NEJM AI—a newly established, peer-reviewed journal from the publishers of the highly respected, New England Journal of Medicine. @CedarsSinai

1

3

21

9

13

76

David Ouyang, MD

@David_Ouyang

2 months

Congrats to @SarnoffCardio Victoria Yuan and team for using deep learning to characterize genetic contributors for subtypes of left ventricular hypertrophy. #EHJDigital #UKBB

2

10

73

David Ouyang, MD

@David_Ouyang

10 months

My response to the @nih_nhlbi 's Request for Information around Generative AI.

2

16

71

David Ouyang, MD

@David_Ouyang

1 year

Tremendous work by star UCLA medical student Victoria Yuan and @BioEngUCLA PhD student @milos_ai to use deep learning to characterize apical and septal hypertrophy, isolate specific genetic variants, and impact on clinical outcomes. 1/n

2

12

78

David Ouyang, MD

@David_Ouyang

1 year

Tremendous work lead by @yuanneal showing AI-ECG models can be developed and deployed in diverse populations including @DeptVetAffairs veterans. @SFVAMC @CedarsSinaiMed 1/n

3

14

70

David Ouyang, MD

@David_Ouyang

1 year

We are excited to present our work on ECG detection of chronic kidney disease (CKD) published in @CommsMedicine ! @LauriHolmstrom @kayzayai @jwestonhughes @james_y_zou @SumeetSChugh 1/n

2

15

67

David Ouyang, MD

@David_Ouyang

4 years

We are excited to present EchoNet-Labs, our newest video-based deep learning model to predict biomarkers from echocardiogram videos. @jwestonhughes @james_y_zou @CedarsSinai @Stanford 1/n

Deep Learning Prediction of Biomarkers from Echocardiogram Videos

Laboratory blood testing is routinely used to assay biomarkers to provide information on physiologic state beyond what clinicians can evaluate from interpreting medical imaging. We hypothesized that...

www.medrxiv.org

4

17

66

David Ouyang, MD

@David_Ouyang

6 months

We are excited to be attending #ACC24 , proud of the trainees and collaborators in their excellent work and presentations. Please stop by to say hi and check out our research!

2

5

66

David Ouyang, MD

@David_Ouyang

2 years

Excited to hang out with old and new friends and talk at my alma mater! @UCSF @JamesETooley @jpirruccello @GeoffTison @arnaout_rima @yuanneal @JChorba @ucsf_echo @UCSFCardiology

Leila Haghighat, MD MPhil

@LeilaHaghighat

2 years

🗓️10/26: Eager to hear @UCSFCardiology grand rounds with @David_Ouyang on "Development to Deployment of Cardiovascular AI." Streaming tomorrow live at 12 PM PST:

0

17

2

6

64

David Ouyang, MD

@David_Ouyang

3 years

To be explicit, the academic cardiologist of tomorrow needs to know how to program. Domain knowledge is rapidly learnable, while knowing dplyr, regex, sql, and/or pytorch is a strong differentiator.

Svati Shah

@SvatiShah

3 years

And shout out from Dr Braunwald to genetics @CSHeartResearch @APLandstrom @pnatarajanmd @amitvkhera @CarolynHoMD @neallakdawala @PersonalizedMed

0

17

33

3

8

63

David Ouyang, MD

@David_Ouyang

11 months

Check out Late Breaking Science 8 at #AHA2023 AI at the bedside @AHAScience @HeartBobH @leftbundle @rohan_khera @VTaqMD @mvperez92

0

9

62

David Ouyang, MD

@David_Ouyang

2 years

@nabeelqu @DynamicWebPaige Check out @mkaic_ ’s work!

Kai Christensen

@kayzayai

2 years

I'm starting to get the hang of these #StableDiffusion latent space tours! Spherical coordinates instead of euclidean, cubic spline interp instead of linear, finding the right density of waypoints... Anyway here's a Ghibli one I made with all my newfound knowledge:

2

14

140

0

3

61

David Ouyang, MD

@David_Ouyang

2 years

1/n We are excited to present for the first time at #HRS2022 , PreOpNet, the first ECG-based deep learning algorithm for pre-operative risk stratification. @CMAlbertEP @CedarsSinai

Zahra Azizi, MD, MSc

@ZahraaziziMD

2 years

#HRS2022 Join us at Practice Improvement session chaired by Dr @EP_mom1 , where Dr Turakhia @leftbundle and Dr @David_Ouyang are going to discuss “Electrocardiographic Deep Learning For Predicting Post-procedural Mortality” #EPeeps @StanfordCDH @StanfordDeptMed

1

2

4

3

8

59

David Ouyang, MD

@David_Ouyang

4 months

Second first author research publication by star @SarnoffCardio fellow and soon to be @StanfordMedRes @amey_vrudhula ! @LilySternMD @Paul_P_Cheng @pieroricchiuto @Ron_Witteles @JigPatelMDPhD

1

10

55

David Ouyang, MD

@David_Ouyang

2 years

Can you tell which annotations were made with #AI and which were made by clinician? Tune in on 8/27 (14 days!) for our BLINDED, RANDOMIZED, PROSPECTIVE clinical trial at @escardio #ESCCongress 2022.

David Ouyang, MD

@David_Ouyang

2 years

We are excited to present our prospective blinded randomized clinical trial of AI vs sonographer initial interpretation of LVEF in the #LBCT HOTLINES Session 3 of #ESCcongress 2022 on August 27. #WeNeedRCTs

2

11

91

3

18

56

David Ouyang, MD

@David_Ouyang

4 months

Liver disease is often asymptomatic, leading to under diagnosis and under treatment. As @ASE360 and @CVPR starts, we are excited to present EchoNet-Liver: Opportunistic Screening of Chronic Liver Disease with AI Enhanced Echocardiography @CedarsSinaiMed @SmidtHeart

2

12

58

David Ouyang, MD

@David_Ouyang

8 months

Sharing code and data is science. Not sharing code nor data is marketing. The choice is between building a foundation for others to build on or create a closed system to be monetized.

Demi Adedinsewo, MD

@DemiladeMD

8 months

Fantastic talks to close out #AICARDIOLOGY2024 🔑 takeaways 📌 #AI implementation is challenging 📌 We need to thoroughly evaluate risk 📌 Sharing clinical data/code can be a high price, think data privacy 📌 Healthcare professionals will benefit from some AI education

0

1

8

4

9

55

David Ouyang, MD

@David_Ouyang

2 years

These echos do not exist. #stablediffusion #CLIP #echofirst #echoCLIP

1

5

55

David Ouyang, MD

@David_Ouyang

6 months

Great time at #ACC24 ! Great job by the entire team! @SmidtHeart @CedarsSinaiMed

David Ouyang, MD

@David_Ouyang

6 months

We are excited to be attending #ACC24 , proud of the trainees and collaborators in their excellent work and presentations. Please stop by to say hi and check out our research!

2

5

66

3

9

54

David Ouyang, MD

@David_Ouyang

2 years

There's tremendous work on predicting all sorts of labels from medical imaging (age, sex, race, outcomes, diagnoses). It begs the question - which tasks are reasonable/unreasonable, and how do we think about confounding? In this recent paper, we study this question. 1/n

1

9

51

David Ouyang, MD

@David_Ouyang

6 months

Thank you and it was great sharing the stage and working together! @ppsengupta1 @YanamalaNaveena @AChoiHeart @PlayfordDavid

Naveena Yanamala

@YanamalaNaveena

6 months

A Fantastic talk by @David_Ouyang reflecting on what it takes to implement AI tools in Clinical Practice…

0

15

1

6

54

David Ouyang, MD

@David_Ouyang

5 months

Great work by @Ray29750200 @CedarsSinaiMed for proof of concept smartwatch ECG to detect hyperkalemia in ESRD patients! External validation at @stanfordmed by @jwestonhughes @ajrogers_md @mvperez92 @james_y_zou and CGMH in Taiwan

Ray Chiu

@Ray29750200

5 months

Excited to share our new work #KardioNet , an AI-enabled smartwatch ECG algorithm for serum potassium monitoring. Trained on ~300k ECG and potassium pairs at @CedarsSinai , Kardio-Net is designed to enhance potassium monitoring via smartwatch for ESRD patients. #digitalhealth

1

5

27

1

12

53

David Ouyang, MD

@David_Ouyang

9 months

Members of the research group enjoying Hawaii to start 2024 at #PSB24 ! Presenting projects ranging from #echofirst to MRI, genetics to precision phenotyping.

5

3

52

David Ouyang, MD

@David_Ouyang

6 months

Great job @yuanneal , Nathan, @grantmduffy and the rest of the team! @SmidtHeart @CedarsSinaiMed @SFVAMC

npj Digital Medicine

@npjDigitalMed

6 months

Deep learning of ECGs can help identify patients with hidden #atrialfibrillation better than predictions based on clinical risk factors & echo measurements. This may offer additional opportunities to guide patient screening & support early intervention.

0

8

25

2

8

52

David Ouyang, MD

@David_Ouyang

3 months

Prospective RCT of AI-Enhanced ECG for STEMI Activation @NEJM_AI Excellent trial building upon and complementing recent work in @NatureMedicine Excellent editorial by @RobertAvramMD and @wfearonmd

2

10

49

David Ouyang, MD

@David_Ouyang

2 years

Excited to extend our work in cardiac ultrasound to POCUS and the emergency department. Strong performance in both assessing video quality and function as evaluated by expert ER physicians @bryandhe @james_y_zou @StanfordEMED @CedarsSinaiMed

6

8

51

David Ouyang, MD

@David_Ouyang

9 months

We reverse engineered the data structure from @PhilipsHealth ultrasound for 3D voxel data. At #PSB24 , @grantmduffy presented our 3D echo toolkit! This allows for synthetic 2D data generation, as well evaluation of the impact of foreshortening on LVEF assessment.

6

13

47

David Ouyang, MD

@David_Ouyang

2 years

To be honest, my heuristic is if someone is claiming an AUC near 1, I have a strong prior that the model will NOT work. That also goes for vendors that claim a variance of 0 on test-retest performance.

Eric Topol

@EricTopol

2 years

A medical #AI deep neural network algorithm had an AUC performance of 0.994 (near perfect) and outperformed doctors. Then it showed "unexpected and potentially harmful algorithmic behavior" @LancetDigitalH

7

73

204

4

6

50

David Ouyang, MD

@David_Ouyang

1 year

As a physician, I don't want my assessments and plans to be used to train models for billion dollar private companies. Epic has a track record of being a closed sandbox hostile to 3rd party, not flexible to individual institution needs, and monopolistic practices.

#JinJiyanAzadi

@smellyrajma

1 year

Frankly as a patient I would not want my medical records to be so openly accessible. We need a physical Ubikey for healthcare to prevent massive learning on our medical records

0

1

2

4

8

48

David Ouyang, MD

@David_Ouyang

2 years

We're excited to work with @StanfordAIMI to release the EchoNet-LVH data, 12k PLAX #echofirst videos with detailed annotations and measurements! @Paul_P_Cheng @bryandhe @james_y_zou @salernomdphd

Stanford AIMI

@StanfordAIMI

2 years

📣New AIMI Dataset! EchoNet-LVH includes 12k labeled echocardiogram videos and expert annotations to provide a baseline to study cardiac chamber size & wall thickness. Huge thanks to @David_Ouyang and @CedarsSinai @StanfordAIMI collaborators and staff!

0

10

32

1

6

47

David Ouyang, MD

@David_Ouyang

3 years

Thanks for the opportunity to share our work and really excited for the future of AI in cardiovascular imaging! @EvanMuse @ppsengupta1 @ACCinTouch

Evan D. Muse

@EvanMuse

3 years

Cardiovascular imaging under a sea change with novel AI/ML technologies @David_Ouyang #PrecisionMedicine @GiorgioQuer

0

5

16

2

4

45

David Ouyang, MD

@David_Ouyang

6 years

It takes a village! Thanks again to all my mentors! @euanashley @FaRodriguezMD @HeartBobH @Ron_Witteles @jonc101x @joshuawknowles @StanCVFellows @StanfordDeptMed @StanfordMed

American College of Cardiology

@ACCinTouch

6 years

For more than 35 years, the ACC/ @Merck Research Fellowship Awards have supported #ACCFIT seeking to launch careers in #CVD research. Congrats to #ACC19 recipients, Drs. David Ouyang and Rishi Wadhera. #TheFaceOfCardiology

0

3

31

3

6

43

David Ouyang, MD

@David_Ouyang

4 years

I'm incredibly excited for the field + upcoming work from our multi-disciplinary/institution work @CedarsSinai @StanfordDeptMed ! I continue to benefit from tremendous mentorship @euanashley @james_y_zou @joshuawknowles @EldrinL @HeartBobH and lessons learned from @StanCVFellows !

Robert Harrington

@HeartBobH

4 years

Great to see former @StanfordDeptMed resident and @StanCVFellows @David_Ouyang kicking off this session. He has done very nice work in #echo imaging worth reviewing. Kudos David. #AHA20 @AHAScience @joshuawknowles @CHaeffele_MD @EldrinL @Ron_Witteles

1

4

24

2

3

40

David Ouyang, MD

@David_Ouyang

6 months

Strong work lead by @moorejh and @gracielagon in bringing AI education to @CedarsSinaiMed !

Jason H. Moore, PhD

@moorejh

6 months

We are pleased to announce the launch of our new PhD Program in Health Artificial Intelligence at Cedars-Sinai Medical Center under the leadership of @gracielagon . Apps due in the Fall of 2024 #artificialintelligence #healthAI #bioinformatics #datascience

9

49

182

0

7

40

David Ouyang, MD

@David_Ouyang

2 years

Really want to highlight this delightful post-presentation conversation with @JamesDThomasMD1 as we discuss the tremendous opportunity, technical challenges (interoperability in medical software), and landscape of #echofirst #AI . @CedarsSinai @escardio

ESC TV at #ESCCongress 2022 - Meet the Trialist - EchoNet-RCT

Meet the Trialist - EchoNet-RCTHosted by James Thomas, USAWith David Ouyang, USAAn ESC TV programme recorded at ESC Congress 2022The video content has been m...

www.youtube.com

4

8

37

David Ouyang, MD

@David_Ouyang

1 year

Tremendous work by @us2_ai and I'm a huge fan of their software and the space (AI automation of echo), but there are a few challenges with this type of analysis I want to point out and emphasize for the critical appraisal of AI (as well as reading papers in this space). 1/n

Ritu Thamman MD

@iamritu

1 year

@us2_ai algorithm quantifies AS severity with no human input beyond image acquisition,matches expert’s measurements on AV Vmax (r = 0.97) MPG (r = 0.94) AVA (r = 0.88) across normals & all grades of AS severity LVOT lowest agreement b/w AI & human(r=.76)

0

5

22

1

8

39

David Ouyang, MD

@David_Ouyang

2 years

In work lead by @yuanneal , we show #EchoFirst can predict CAC score, CAC=0 as well as high CAC. This AI-CAC similarly risk stratifies mortality. #powerofzero

Dr.Patricia Pellikka

@pattypellikka

2 years

Novel #AI model detects CAD from PLAX or A4C #Echofirst view. Congrats @David_Ouyang @JournalASEcho Prediction of Coronary Artery Calcium Using Deep Learning of Echocardiograms

0

14

39

4

8

39

David Ouyang, MD

@David_Ouyang

7 months

Also congrats to @amey_vrudhula to match at @StanfordMedRes @StanfordMed ! He is quite the catch for the program! Will either have 3 or 4 published original research papers from the @SarnoffCardio fellowship by the time residency starts! @Ron_Witteles @AJRogers_ICU @Paul_P_Cheng

Victoria Parikh, MD

@vnparikh

7 months

So proud of you @amey_vrudhula , @SarnoffCardio fellow extraordinaire! And huge thanks to the one and only @David_Ouyang for being an amazing mentor and scientist. What a cool project! @chaz_hong @CardioOncology @jpirruccello @euanashley @SeanM_Wu

2

3

27

1

3

38

David Ouyang, MD

@David_Ouyang

10 months

Check out this great work lead by @milos_ai being presented at @SymposiumML4H ! Using UKBB, we can generate synthetic cardiac MRI image using GANs. Sampling within the latent space between systole and diastole can generate videos. And these generated videos demonstrate

Milos Vukadinovic

@milos_ai

10 months

Create synthetic cardiac MRIs with stunning realism using GANcMRI #AI !🫀 Excited to share that our latest work “cMRI video generation and physiologic guidance using latent space prompting” has been accepted @SymposiumML4H @NeurIPSConf 2023 Proceeding Track. 1/🧵

5

16

79

0

9

36

David Ouyang, MD

@David_Ouyang

3 years

Exciting to see randomized trials for AI technologies in cardiology! Positive endpoint and showing utility in improving patient care! Excited for more precision phenotyping in ECG and echo. Tremendous foresight by @MayoClinic ! @EricTopol

Peter Noseworthy, MD

@noseworthypeter

3 years

Artificial intelligence–enabled electrocardiograms for identification of patients with low ejection fraction: a pragmatic, randomized clinical trial | Nature Medicine

8

77

283

0

5

35

David Ouyang, MD

@David_Ouyang

6 years

1/6 Just gave my ACC/Merck Research Talk at #ACC19 . I'd like to summarize my key points online. First, special thanks to @euanashley , James Zou, David Liang, @FaRodriguezMD @HeartBobH @jonc101x @StanfordCVFellowship @Ron_Witteles for their mentorship. #ACCFIT #CVImaging #HAI

2

6

35

David Ouyang, MD

@David_Ouyang

1 year

In addition to our work with TRANSTHORACIC echocardiography, we're exciting to announce our work on AI of TRANSESOPHAGEAL #echofirst in collaboration with @KSteffner @stanfordanes @DrBowdish @SmidtHeart and @james_y_zou @StanfordDBDS 1/n

3

8

36

David Ouyang, MD

@David_Ouyang

3 years

To promote further innovation in the field, in addition to our results, we release a LARGE (12k videos) dataset and code!

GitHub - echonet/lvh: EchoNet-LVH is a deep learning model that quantifies ventricular hypertrophy...

EchoNet-LVH is a deep learning model that quantifies ventricular hypertrophy and predicts etiologies of increased wall thickness and LVH (amyloidosis, HCM, etc). - GitHub - echonet/lvh: EchoNet-LV...

github.com

JAMA Cardiology

@JAMACardio

3 years

In this study, a deep learning workflow was able to automate wall thickness evaluation while facilitating identification of hypertrophic cardiomyopathy and cardiac amyloidosis.

0

1

16

3

7

36

David Ouyang, MD

@David_Ouyang

4 months

I will be at @CVPR #CVPR2024 presenting our work on EchoCLIP, foundation models for echocardiography, multimodal cardiology models, and clinical deployment. With new funding, we are recruiting at all levels for our team! Come talk with us!

2

7

36

David Ouyang, MD

@David_Ouyang

4 years

Next steps: I’m excited to join @CedarsSinai in the Department of Cardiology under @CMAlbertEP . I’ll be in the #echofirst lab and continue working on AI in cardiology! I’m looking to work with talented technical and medical people! Research focus here: 5/n

2

0

36

David Ouyang, MD

@David_Ouyang

1 year

@kayzayai @milos_ai @yuanneal @CedarsSinaiMed @SmidtHeart @LorenzoRighett7 Preprint at: Model and weights at: 6/Fin

9

7

35

David Ouyang, MD

@David_Ouyang

1 year

Thanks for the opportunity to present at @MICCAI_Society CauseMIC: Causality in Medical Image Computing. @GlockerBen @kbatmang If you're interested in doing a postdoc and wanted unfettered, large scale access to medical (cardiovascular) imaging data, please contact me!

1

4

35

David Ouyang, MD

@David_Ouyang

7 months

Excited to share this perspective with @pattypellikka on @ASE360 @JournalASEcho 's approach to AI in #echofirst . Rigorous science includes open code, open datasets, and external validation. If anyone feels bottlenecked by the ability to share data because of PHI concerns, I'm

0

9

34

David Ouyang, MD

@David_Ouyang

3 years

Excited to invite @RoxanaDaneshjou to @CedarsSinai and hear how she is pushing the boundaries of dermatology with improving image acquisition and reducing bias in datasets and AI models! @james_y_zou @StanfordMed

Roxana Daneshjou MD/PhD

@RoxanaDaneshjou

3 years

How cool is this -- Cedars-Sinai has a division of AI in Medicine ( @David_Ouyang is one of their rising stars). Excited to be talking to their group this afternoon about building foundations for AI in Dermatology, with fairness as a core tenet.

0

22

0

5

33

David Ouyang, MD

@David_Ouyang

4 years

Thanks so much to @SiemensHealth @Siemens for the opportunity to review prior work and future opportunities of #AI in Echocardiography and #AI in Cardiology! Tremendous potential integrating across entire stack of acquisition -> processing -> interpretation! @CedarsSinai @ASE360

1

3

33

David Ouyang, MD

@David_Ouyang

5 years

An honor to present at #ASE19 ! The future of echo and cardiovascular imaging is bright with opportunities to better diagnose and treat patients! @euanashley @HeartBobH @StanfordAIMI @ASE360

Tiffany Chen

@tiffchenMD

5 years

Phenomenal YIA presentation by @David_Ouyang on deep learning of #echofirst to predict systemic phenotypes #ASE2019 @ASE360 @euanashley

0

5

12

3

4

32

David Ouyang, MD

@David_Ouyang

3 years

8/8 Finally, no research is complete without a prospective clinical trial. With @LilySternMD @jigPatelMDPhD , we have started screening all echos done at @Cedars for Cardiac Amyloidosis. The first 3 patients flagged by the algorithm were seen by this week!

5

32

David Ouyang, MD

@David_Ouyang

4 years

#AHA2020 Monday 8:30AM PST / 10:30 AM CST Come check out the LIVE session: 🤖 Hype or Hope? Artificial Intelligence and Machine Learning in Imaging 🚀 Hype 🚂 vs hopeful🧑‍🔬discuss how AI fits into CV Imaging! Ft @mariellesc1 @alexbrattmd @asset25 @damini_dey Playfield and Rubin!

2

8

33

David Ouyang, MD

@David_Ouyang

3 months

Congrats to the @InVision_AI team! Clear identification of etiologies of HF, particularly HFpEF, is critical for early access and targeted treatment!

InVision Medical Technology Corporation

@InVision_AI

3 months

We are excited to announce a new CPT code for AI-enhanced echo for detection of Cardiac Amyloidosis! Yesterday, the AMA has issued new CPT code 0932T, a new Cat 3 CPT code to track InVision Precision Cardiac Amyloid usage starting 1/1/2025.

0

2

10

0

5

32

David Ouyang, MD

@David_Ouyang

8 months

@rSanti97 This is money spent on actually treating sick patients who might otherwise die. This is a low margin service. 17% of the entire federal budget is spent on debt service.

2

4

31

David Ouyang, MD

@David_Ouyang

2 years

New @ASE360 #EchoGuideApp ! Great resource at fingertips for reviewing normal/abnormal measurements in #EchoFirst !

0

6

30

David Ouyang, MD

@David_Ouyang

5 years

@bryandhe @euanashley @james_y_zou @PyTorch 6/7 And thanks to our friends @StanfordAIMI @curtlanglotz @mattlungrenMD , we will be releasing our dataset before the #ML4H workshop at @NeurIPSConf . 10k deidentified A4c echo videos used to train EchoNet-Dynamic as a resource for ML.

3

6

30

David Ouyang, MD

@David_Ouyang

10 months

In parallel, congrats to @bryandhe and the @InVision_AI team for receiving @US_FDA breakthrough designation for #amyloidosis screening. We are excited about technologies that can shorten the journey to diagnosis for rare diseases. @Amyloidosis_ARC @ASE360

InVision Medical Technology Corporation

@InVision_AI

10 months

We are excited to announce InVision has secured FDA breakthrough device designation for #ai enabled #amyloidosis detection. The AI algorithm integrates information from multiple #echofirst views and is the only peer reviewed algorithm of its kind.

0

3

25

2

4

29

David Ouyang, MD

@David_Ouyang

3 years

Congrats @jwestonhughes ! A comprehensive deep learning enabled assessment of the relationship between cardiac imaging and biomarkers published at @EBioMedicine ! @james_y_zou @StanfordDBDS @CedarsSinaiMed

David Ouyang, MD

@David_Ouyang

4 years

We are excited to present EchoNet-Labs, our newest video-based deep learning model to predict biomarkers from echocardiogram videos. @jwestonhughes @james_y_zou @CedarsSinai @Stanford 1/n

4

17

66

0

7

28

David Ouyang, MD

@David_Ouyang

10 months

I was gonna summarize the results of our recently published work at @LancetDigitalH , but I think @carlospedrotti did an excellent job! Check out his thread. @CMAlbertEP @SmidtHeart @CedarsSinaiMed

Dr. Carlos Pedrotti

@carlospedrotti

10 months

Exciting insights! Discover how deep learning and ECGs can revolutionize predicting post-procedural mortality in this recently published paper from @LancetDigitalH by @David_Ouyang . Here are the key findings. #healthcare #cardiology #ecg #digitalhealth

2

12

26

2

4

30

David Ouyang, MD

@David_Ouyang

1 year

Fantastic conference @SAILhealth - some of the best, most in-depth in-person AI conversations I’ve had since 2020! @rohan_khera @moorejh @arjunmanrai @JamesADiao @DrXiaoLiu @DanielYangMD @FarazAhmadMD @mvperez92

Rohan Khera

@rohan_khera

1 year

Nice to finally meet and catch up with friends & meet new colleagues at @SAILhealth - an amazing venue for multiple stakeholders in healthcare AI. @kdpsinghlab @David_Ouyang @FarazAhmadMD @pranavrajpurkar Great conversations at this #NoSuit Symposium. Thanks @zakkohane !

0

4

50

1

6

30

David Ouyang, MD

@David_Ouyang

5 years

@bryandhe @euanashley @james_y_zou 5/7 EchoNet-Dynamic was implemented in @pytorch and the code is available at:

GitHub - douyang/EchoNetDynamic: EchoNet-Dynamic is a deep learning model for assessing cardiac...

EchoNet-Dynamic is a deep learning model for assessing cardiac function in echocardiogram videos. - GitHub - douyang/EchoNetDynamic: EchoNet-Dynamic is a deep learning model for assessing cardiac ...

github.com

1

5

28

David Ouyang, MD

@David_Ouyang

1 year

Appreciate @ScienceMagazine highlighting our work in how AI detects race in medical imaging. Individuals are inconsistent in how they report their own race - to extent that AUC of 0.99 is unlikely given label noise. We argue unaccounted confounders drive most of the "signal".

Science Magazine

@ScienceMagazine

1 year

#AI -predicted race variables from medical images pose risks and opportunities for studying health disparities, argues a new #SciencePerspective .

4

7

29

1

5

29

David Ouyang, MD

@David_Ouyang

10 months

The article states my opinion correctly but I want to clarify - currently RCTs are not required to clear AI SaMD products, and while that is ok, payers feel the bar is not high enough and requiring more evidence to pay for medical AI products. I would argue that’s actually a

Bo Wang

@BoWang87

10 months

Fascinating development! I share @David_Ouyang ’s view that FDA approval is a basic (low-bar) step, and we definitely need more extensive perspective validation. The field of AI in cardiology is ripe with opportunities, given the numerous low-hanging fruits. Looking forward to

2

22

4

3

28

David Ouyang, MD

@David_Ouyang

5 years

@bryandhe @euanashley @james_y_zou @PyTorch @StanfordAIMI @curtlanglotz @mattlungrenMD @NeurIPSConf @HeartBobH @GhorbaniAmirata @StanfordMed @paheidenreich @ASE360 @StanfordHealth @StanCVFellows @StanfordDeptMed 8/7 (Oops!) Our @medrxivpreprint is at

Interpretable AI for beat-to-beat cardiac function assessment

Accurate assessment of cardiac function is crucial for diagnosing cardiovascular disease[1][1], screening for cardiotoxicity[2][2],[3][3], and deciding clinical management in patients with critical...

www.medrxiv.org

4

7

28

David Ouyang, MD

@David_Ouyang

9 months

Congrats @abidlabs and the @Gradio team! I was very excited to be a guinea pig for the product and the ability to give real-time feedback on model performance is still a piece that's lacking in many areas of medical AI. The most exciting work happens at the intersection of

Abubakar Abid

@abidlabs

9 months

Embraced by Hugging Face: the Inside Story of Our Startup’s Acquisition In late 2021, our team of five engineers, scattered around the globe, signed the papers to shut down our startup, Gradio. For many founders, this would have been a moment of sadness or even bitter

48

120

722

0

3

27

David Ouyang, MD

@David_Ouyang

4 years

We're excited to work with @abidlabs to make the EchoNet-Dynamic segmentation model available to physicians and non-programmers on @gradioML !

1

10

28

David Ouyang, MD

@David_Ouyang

2 years

I look forward to the challenge, but have to say, I definitely have the tougher side of this debate! Maybe we should have this debate again at an AI conference instead of an #echofirst conference next!

James D Thomas, MD

@JamesDThomasMD1

2 years

You might think that @David_Ouyang and I always agree on AI. But today we do battle at #EchoAsia with David claiming that AI will replace the Echocardiographer in the near future. Balderdash! Watch my defense! Here we are in happier times at ESC… @ASE360

1

3

32

4

2

28

David Ouyang, MD

@David_Ouyang

1 year

@_cartick @suchenzang It’s the wrong MRI, a knee MRI has specific protocols. The ‘screening mri’ is a very limited study that doesn’t visualize many things well (like the heart, which requires timing) and preys on the health anxieties of the average person (insurance wouldn’t cover such a study).

1

2

27

David Ouyang, MD

@David_Ouyang

4 years

Very important points! AI needs good data! @mattlungrenMD has been pioneer and tackling hurdles for data sharing. We worked w @StanfordAIMI to release 10k echos @ASE360 @FedericoAsch @AkhilNarangMD @iamritu @AHA_Research @HFpEF @JamesDThomasMD1 @sri_adu

Jordan Strom, MD, MSc, FACC, FASE

@DocStrom

4 years

#AI #MachineLearning Michael Blum: we need very large and well annotated datasets to make algorithms generalizable, but these are hard to come by. This requires security in data sharing technologies. #ASE2020 @ASE360

0

3

13

1

9

27

David Ouyang, MD

@David_Ouyang

5 years

Thanks for the opportunity to describe our work and the opportunities in ML for cardiovascular imaging. @NorthwesternMed has tremendous vision, infrastructure, and talent to be at the forefront! @NMHheartdoc @JamesDThomasMD1 @HFpEF @ramseywehbemd

Josh Cheema, MD, MSCI, MSAI

@JCheemaMD

5 years

⁦ @David_Ouyang ⁩ sharing his work predicting EF from TTE using a deep learning model, and sharing his dataset and algorithms publicly, all as a 3rd year fellow. A pioneer in this field!

0

16

2

0

27

David Ouyang, MD

@David_Ouyang

3 years

Strain is one of the most exciting research measurements in #EchoFirst , and we're excited to announce our deep learning straining (DLS) implementation. 1/n

Alan Kwan

@alanckwan

3 years

Can we adapt a semantic segmentation model to measure myocardial strain? You bet. @David_Ouyang

0

2

7

1

6

26

David Ouyang, MD

@David_Ouyang

1 year

Achieving sufficient data scale is a key bottleneck for medical AI. To our knowledge, this #echofirst model is the largest model yet for echocardiography, with 10x more data than our prior work and (1000x more data than current commercial models). Great job @kayzayai @milos_ai

1

3

27

David Ouyang, MD

@David_Ouyang

3 years

4/n We are so used to thinking of ML as ‘black-box’ that we have forgotten machine learning training is designing a case-control study. The choice of NEGATIVE controls is just as important as POSITIVE cases. Standard epi study design questions still apply.

1

8

26

David Ouyang, MD

@David_Ouyang

2 years

I finally found the paper I'm looking for - this paper is as close as it gets for test-retest performance of LVEF on cardiac MRI!

1

2

25