vasc ai

We’re thrilled to be highlighted! At @vasc_ai, our mission is to enhance vascular care through innovation. Our tool is free to use, and we’re dedicated to creating more cutting-edge solutions that empower clinicians, researchers, and educators. The journey has just begun! 🚀

@mattsmeds @VESurgery This is one way to get people to go the talks 😂

Have to be honest… it’s a good meeting. @VESurgery #VESS25

@yuejianing Would have to agree with everyone else here, open surgical repair with vein, rifampin soaked graft or cryopreserved vein are all viable options.

Bilateral infectious iliac aneurysms, 28 years old, male. Distal branches are small, unhealthy and occluded. What should be done? #infectedAAA #AortaEd

5/ Error Analysis: General LLMs Are Prone to Critical Mistakes We classified errors into two categories: 🔴 Logical Errors: The LLM finds relevant information but misapplies it. 🔴 Information Errors: The LLM misses key clinical details or fabricates data. 📊 Breakdown of Incorrect Answers: ❌ ChatGPT-3.5: 34.5% logical errors, 65.5% information errors ❌ ChatGPT-4: 23.0% logical errors, 76.5% information errors ❌ ChatGPT-4o: 25.5% logical errors, 74.7% information errors ✅ VASC.AI: 100% of errors were logical—zero information errors This confirms that VASC.AI eliminates factual inaccuracies, but complex clinical reasoning still requires refinement.

4/ Why VASC.AI Excels: Retrieval-Augmented Generation (RAG) Unlike general LLMs, VASC.AI integrates a structured knowledge retrieval system, allowing it to: 📚 Search >250,000 clinical abstracts, guidelines, and landmark trials 📚 Utilize up-to-date Society for Vascular Surgery & ESVS guidelines 📚 Ground responses in evidence, reducing “hallucinations” This RAG approach ensures accuracy and reliability in vascular surgery recommendations.

🚨 THREAD: Why VASC.AI Outperforms General LLMs in Vascular Surgery: AI has immense potential in medical education, but general-purpose large language models (LLMs) are not optimized for specialized fields like vascular surgery. Our study demonstrates why VASC.AI (chat.vasc.ai), a retrieval-augmented generation (RAG) model, surpasses general LLMs. 🧵👇

@mattsmeds For a patient with symptomatic venous insufficiency and GSV reflux below the knee who has not responded to compression therapy, endovenous ablation is a suitable next step.

Symptomatic venous insufficiency (sore/tired/aching legs) and GSV reflux only below knee. Failed compression. What do you do?

@sama 11. Though superintelligence seems like science fiction now, we're confident it will soon become reality; acting with care while maximizing benefits is vital given our work's profound implications—OpenAI cannot operate as an ordinary company.

@sama 10. Looking ahead to 2025, we anticipate AI agents joining the workforce significantly impacting company outputs while exploring superintelligence potentials—envisioned as accelerating scientific discovery beyond current capabilities.+

I highly recommend reading Sam Altman’s (@sama) latest blog post on OpenAI’s journey toward AGI. It’s a thoughtful reflection on the progress, challenges, and responsibilities of developing transformative technology, blending optimism with introspection. #AI #AGI #VascularSurgery #Innovation