🎓 Happy to announce that I am officially done with my Transitional Year residency! Up Next: Diagnostic Radiology 🥳@UTMBRadiology

#RSNA2025 #Trainees ⭐️Do you want to know how effective the Radiology in training program can be or do you want to have an informal chat about it? 📍Come join me for our poster pres.! Happening now at QI #8 station @VChernyakMD @AlessiaGuarnera @eniscy @radiology_rsna

Glad to wrap up this year’s RSNA by attending my mentor Dr. Turkbey’s talk @radiolobt #RSNA25 #Prostate #AI

Recent UTMB graduate @Celina_NahyunJo and current @UTMBRadiology resident with the one and only @VChernyakMD, our #RadInTraining Editor! @RITEditor #RSNA25

Radiology-in-Training Program overview during the @Radiology_RSNA Editorial Board Meeting by Dr. Chernyak @VChernyakMD Also grateful to receive the Recognition Award with Special Distinction 🙏🏼🎖️ @RITEditor @RadiologyEditor #RSNA25

Curious about the work of the Radiology Trainee Editorial Board? Join @VChernyakMD , the Rad-in-Training Editor, and current board members as they share insights into the application process, discuss their experiences on the team, and answer your questions. 📍Residents Lounge,

Radiology Editorial Board Meeting launched with introductory remarks by the new Editor-in-Chief @SuhnyAbbara #RSNA25 #Radiology @Radiology_RSNA

Glad to see the collective effort of our team and collaborators are being recognized! 🎊

1/10 Can #AI help create those eye-catching visual abstracts we see in radiology journals? 🤖🤔 New @Radiology_RSNA study tests GPT-4o's ability to generate visual abstracts (VAs) vs. human-created ones. @RITEditor @VChernyakMD @Radiology_Editor #Tweetorial #RadInTraining 👇

🚨📈 CT use in the emergency department has nearly doubled in the past decade among Medicare patients. New @Radiology_RSNA study breaks down ED imaging trends from 2013–2023. 🧵on what’s driving the growth ⬇️ #RadInTraining #TWEETORIAL @radiology_rsna @RITEditor @VChernyakMD

#RIAM (Radiology in a Minute)@radiology_rsna Microbubble-based Radiosensitization of Hepatocellular Carcinoma: Evaluation of Safety and Efficacy in a Phase II Randomized Trial Full study link : https://t.co/cZ8BAMNZoe @RITEditor @RadiologyEditor @VChernyakMD #RadInTraining

An amazing human being and one of the most considerate people I know —a true friend, compassionate physician, and dedicated team player! Bonus for his co-residents: Delicious Turkish food and stories from his hikes, camps, and outdoor adventures 🏕️ #Match2026 #InternalMedicine

12/12 That’s a wrap on this #tweetorial, thanks for reading! @PaulYiMD @jjeudymd @woojinrad @FelipeKitamura @AndrewSmith_AI @DrLindaMoy @vishwa_parekh 👉 https://t.co/kItLyI1otY #RadInTraining @VChernyakMD @RITEditor @Radiology_RSNA @RadiologyEditor

11/12 📝Summary: Generative #AI in #radiology is powerful, however, ➡️Regulation, privacy, and bias must be addressed. Radiologists, regulators, and vendors must collaborate to ensure these tools are safe, fair, & truly help the patients.

10/12 Ways to mitigate bias: • Diverse datasets with demographic metadata • Stress-testing with modified prompts • Vendor transparency on bias evaluation • Continuous post-deployment monitoring

9/12 Bias in generative AI: • Different outputs based on patient race/sex in prompts • Reinforcing race-based medical tropes • Stereotyping radiologists in image generation Bias = real-world harm + perpetuated inequities.

8/12 Best practices for privacy: • Local or secure-cloud deployment • Federated learning for multi-site data • Vendor transparency about data use • Guardrails against jailbreaks

7/12 Data privacy risks: • Training data may inadvertently include PHI • Jailbreak attacks can extract sensitive info • Commercial models often require sending data off-site → HIPAA concerns

6/12 Beyond accuracy, safe regulation must consider: · Reproducibility · Robustness to hallucinations · Human-AI interaction (automation bias, workflow impact) ❗️Accuracy ≠ safety.

5/12 ✅B: LLM summarizing a chart → classified as “Non-Device CDS.” But ambiguity remains: if the same tool generates a risk score, it may fall under FDA oversight. This gray zone highlights why regulatory clarity is urgent.

4/12 POLL: Under current FDA guidance, which of the following would NOT be considered a “medical device”?