DIAGENS Shakes Up Medical Imaging with AI Prowess
When it comes to revolutionizing how the healthcare sector tackles imaging, DIAGENS (HKEX: 2526) is firing on all cylinders. Their latest marvel, iMed-6.0, is here, packing a punch with its new suite of features that leave others eating dust. Ever seen a company turn the tables like this? Well, strap in because DIAGENS just took the lead in the MedBench multimodal API evaluation with their iMed-6.0 Thoth and Aegis coming first and second overall. They've even surpassed what was thought as the peak until now on the MMMU Test. Yeah, it's a big deal.
Breaking Down Tech Barriers
Imagine this: a CT model designed to look at 509 imaging targets. Not just peeking, but actually dissecting, understanding, and delivering insights on exactly what's going on in those CT scans—whether it's the chest or the abdomen. Who would have thought collaboration between human innovation and all that fancy AI stuff could handle that much complexity, right?
"We want clinicians and hospitals to build AI around the clinical challenges they know best," said Dr. Ning Song, CEO of DIAGENS.
And boy, have they delivered. DIAGENS is pushing out specialty AI models like it's just another day at the office. The clincher? Zero-code tools, offering every doc out there the opportunity to play AI developer without touching a line of code.
Raising the Stakes in Medical Benchmarks
You heard it: top dogs in the MedBench evaluation. With Thoth scoring a solid 64.4 and its sibling, Aegis, not far behind at 61.7, DIAGENS isn't just part of the playing field. They're reshaping it. These scores aren't just numbers; they're a testament to DIAGENS making a qualitative leap in medical AI.
Why These Results Matter
Scoring on medical visual perception, text extraction, and semantic understanding threw the spotlight on iMed-6.0's power—88.8 and 84.9, respectively. For those keeping score, they’ve blown past previous ceilings by over 15%. That's not just outperforming; that's redefining capability.
One Model, 509 Diagnostic Lenses
Let’s lay it out: a single CT model grasping 509 diseases and imaging findings. Pretty spicy for something operating on non-contrast and contrast-enhanced CT examinations. The key numbers here? An internal validation with a macro-averaged AUC of 0.855 across 40,000 pairs, hitting 0.930 for 249 key targets. That's a substantial dig into precision without compromising scope.
- 263 chest targets
- 391 abdominal targets (with overlap noted once)
These results shout potential louder than any marketing campaign. Talk about a tool that offers both breadth and depth, eliminating the need for piecemeal AI solutions.
Democratizing AI Like Never Before
DIAGENS isn’t just handing over some fancy toolset and calling it a day. Through their iMedLoop platform, they're democratically opening doors for anything from data prep to deploying tailored models at scale, all via a zero-code interface. As of mid-2026, DIAGENS was in cahoots with 99 hospitals, working on 158 projects. Ain't just talk—they're walking the walk, setting an enviable pace in healthcare AI.
Final thoughts? DIAGENS is not messing around. They're setting standards, not just raising them. While healthcare folks might be welcoming this en masse, the broader investor landscape might want to take notice too. As fancy AI these days isn't just fluff—it means business, real business.
