Time to read … good week end ! 📖

Gold morning ! 🌟

A lot of radiologists I’ve met genuinely like routine exams: simple, predictable, low-risk. I never did. Routine bores me. Whenever I’m doing repetitive work, my immediate thought is: a machine should be doing this.

If there’s one thing AI should replace, it’s not radiologists it’s “copy-paste”, “login”, error messages, spinning wheels, and the endless clicks that distract us from our real work and piss us off !

The season for diagnosing sinusitis on brain MRI has officially begun

In the AI era, one question tells you everything about your future : What makes you different from ChatGPT?

⏳ Just 2 weeks to go ! The @RSNA Spotlight #Course on #Radiology #AI comes to Barcelona, Oct 16–17. 🇪🇸 Two intense days on everything you need to know about AI in radiology, from safe adoption to real-world deployment. ✅ Registration still open, seats limited ! #Radiology #AI #RSNA

The real M&A game in imaging AI starts now : @GEHealthCare acquires @icometrix 🚀 www.gehealthcare.com/about/newsro...

There are two kinds of radiologists: those who don’t know they miss abnormalities, and those who don’t know they don’t know.

Radiologists are backstabbed every day. While you’re reading 1 case, the universe is quietly loading 10 more in your worklist. You never catch up because unlike other doctors, your worklist is disconnected from patient flow; imaging is done in parallel, and you sit at the end of the funnel.

Friday breaking news 💥 “@TempusAI Announces the Acquisition of @paige_ai, with its strong foundation in #pathology and #AI is rooted in its proprietary dataset of almost 7 million digitized pathology slides that are clinically annotated” www.tempus.com/news/tempus-...

Zero-shot audio prompting of GPT-4o & Sonnet 3.5 lets radiologists report 2× faster with 2× fewer errors vs Nuance Dragon !!! Soon, we won’t subscribe to traditional tools or our local RIS, we’ll go straight to OpenAI & friends. Times are changing 🚀 link.springer.com/article/10.1...

Radiologist confession : Ever smelled your dictaphone ? Surprisingly… not bad.

The images can be messy, but the report should be crisp and concise. That’s what separates the artists from the crowd in radiology.

LLMs will flood the world with AI-generated content. Soon, it’ll be hard to tell what’s human, what’s machine, and what’s true. The value of human domain experts will only grow. 🚀

Things have shifted. I see many smart, skeptical people (in the best sense) now using ChatGPT to reason, search &even fact-check, not just write. Some trust its answers at face value. But it’s still a language model that mimics reasoning. Have we forgotten what it really is ?

Reading articles about radiology AI in peer-reviewed journals is like looking at a star, what you see is already in the past. 💫 read more 👇 medium.com/@amine.korch...

We, physicians, tend to think we “own” our specialty. But the truth is: we own nothing. Read more 👇 medium.com/@amine.korch...

5 years ago, AI in radiology was fringe. Today, it dominates conferences. Half of NHS trusts use at least 1 AI tool. 30–40% of EU radiologists do too. And we’re just getting started. It’s 2025, and with all its potential for medicine and radiology… it’s still Day 1.

Happy Swiss National Day 🇨🇭 🎉 Bonne fête nationale, schöne Bundesfeier, e buona giornata Svizzera !

Reading about radiology AI in a peer-reviewed journal is like looking at a star: what you see is the past. By the time the article is in press, the tech and its performance are already outdated. That’s how fast the field is moving. 🚀

Pay radiologists per case, and they’ll fight for volume. Pay them a salary, and suddenly they’re “not qualified enough” to read half the exams. Incentives shape behavior !

The best feedback isn’t just “likes” It’s knowing something you shared sparked a thought in someone’s mind. That’s how we all grow, by thinking together.

What got easier with the rise of vibe coding isn’t building full medical AI products. Those are still hard. What’s new is that non-technical folks can now prototype and test small-scale ideas from home. That’s still a revolution. It’s DIY AI time.

A teleradiology company once offered me 15€ per brain MRI I read. That’s 1€ per year of training to learn how to do it properly. Can’t argue with the math. 🧮

“Radiologists used to be seen as doctors,” one physician told me. “Now, some are treated like Uber drivers for diagnostics—paid by the case, with zero security.” Radiology’s Reckoning: How AI, Private Equity, And Prenovo Scans Are Unraveling It @forbes.com www.forbes.com/sites/sindhy...

🚨 Registration open! Join us in Barcelona 🇪🇸, Oct 16–17 for the #RSNA SpotlightCourse on #Radiology #AI. I'll be speaking alongside world-class faculty in a hands-on, practical program on how to use, evaluate & deploy AI in real life. Seats limited. Link 👇 @rsnasky.bsky.social #Radiology #AI

0.1–0.5% of “healthy” adults live with a silent but treatable cancer. That’s ~15–20M people worldwide, ~3M in Europe, ~1M in the US. Radiology and liquid biopsy can help detect more of them. But prevention matters most: reduce lifestyle and environmental risks.

We are not alone folks ! 👇 “Will AI Replace Traders?” @bloomberg.com Podcast

🚨 Registration open! Join us in Barcelona 🇪🇸, Oct 16–17 for the #RSNA SpotlightCourse on #Radiology #AI. I'll be speaking alongside world-class faculty in a hands-on, practical program on how to use, evaluate & deploy AI in real life. Seats limited. Link 👇 @rsnasky.bsky.social #Radiology #AI

Duke team just released MRI-CORE: A Foundation Model for Magnetic Resonance Imaging, trained on + 6 million slices from + 110’000 MRI volumes across 18 body locations. 🚀 Link : arxiv.org/abs/2506.12186 #radiology #AI #MRI

Day On ! Have a Gold Morning ☀️

People complain LLMs give different answers to the same question. But when everyone uses ChatGPT, sameness is the real risk. Unpredictability preserves authenticity.

In radiology, the same lesion gets a different size depending on who’s reading. Honestly, just rounding our measurements might double our kappa scores, and make us look more credible

Have a great Sunday ! 🍗

« Radiologists who failed to find an abnormality are viewed as more culpable when they used an AI system that detected the abnormality. » Randomized Study of the Impact of AI on Perceived Legal Liability for Radiologists | NEJM AI‬⁩ @ai.nejm.org #AI #Radiology

Prompting is the new creativity. The future belongs to those who ask better.

It is happening … #radiology #medical #AI

Medicine is a great career, once you’ve made some money elsewhere.

This week, something unusual happened. Because holidays and many colleagues being away, I had one of those particularly heavy workdays with +100 patient cases. Usually, on such days, I finish completely exhausted. But this time, it felt different, and here’s why. 👇 medium.com/@amine.korch...

Must Read for any institution asked to or considering the adoption of AI in clinical practice ! 👇 #AI in Action: A Road Map From the #Radiology AI Council for Effective Model Evaluation and Deployment - Journal of the American College of Radiology www.jacr.org/article/S154...

Radiology among top 3 specialties with highest intent to leave current job … 😱 But to do what ? radiologybusiness.com/topics/healt...

Gemini 2.5: Pushing the frontier of publishing with 3,000+ authors !! At this rate, the next model will be trained just to manage the author list.

Shorting Bitcoin is like shorting AI. It’s a bet against the future.

The paradox of 2025 : Those who publish the most about AI in radiology are often the ones using it the least. Early adoption required curiosity, risk-taking, and a tolerance for imperfection; traits academia rarely rewards.

Kept seeing ON running shoes everywhere, so I bought the stock. It did well. Now I finally bought the shoes too. 👟📈

Tech builds, but doctors know. If physicians are paid more in tech to build, than what they earn in practice, incentives align. Useful product will be built, adoption will accelerate, and clinical impact will compound. Tech × medicine modernizes healthcare.

Randomized study on @NEJM_AI found that : « #Radiologists who failed to find an abnormality are viewed as more culpable when they used an #AI system that detected the abnormality. » Always review AI outputs carefully before signing off. #Radiology #AI ai.nejm.org/doi/full/10....

Behind the scenes, AI is solving healthcare, one model at a time 🚀

If doctors agree more than 90% of the time, the task belongs to a machine, not a human. Any diagnostic test (particularly imaging) with high inter-reader agreement will be automated by AI. Consensus is a signal to code.