SaigaLab - AI Measurement Copilot for DICOM (Medical Images)

Hey Product Hunt 👋 I'm Kuda. Ten years ago, Geoffrey Hinton - a godfather of AI - said: "People should stop training radiologists now. In five years, deep learning will do better." Today? AI hasn't replaced radiologists. It hasn't even kept up with demand - scan volumes keep climbing. And most of the AI tools that do exist either solve one tiny task or cost a fortune. A crazy amount of radiology research just never reaches the people who need it - especially in developing countries. I did my Master's in Medical Imaging, worked at medical-device and software companies in Germany, published a few cited radiology-AI papers - and a few years ago, I tried to start a radiology-AI startup and failed. Since then I've freelanced on 12+ AI products, but this problem never let me go. So instead of waiting for the perfect product, I stitched together the best open-source tech that already exists - a DICOM viewer, a chat framework, an LLM, and segmentation models like TotalSegmentator, VoxTell - into one thing: 🧠 SaigaLab - an AI Measurement Copilot for medical images. One thing I want to be clear about: it measures, it doesn't diagnose. It quantifies, you stay in control. You open a scan and just chat with it. It can already do real measurements: 🧠 Volumes (brain, ventricles) 📐 Sizes, distances & angles 📊 Evans Index on brain MRI, or practically anything, just prompt It's pretty area-agnostic, but my neuroradiologist friend has experimented on it mostly with brain MRI and CT. Two honest notes: It's an early prototype and not optimized for speed yet - give it a second to think. It is not a medical device - strictly research & education. And because the underlying models are non-commercially licensed, access is 100% free right now. I'd genuinely love your feedback: What would make this actually useful to you? What measurement should it learn next? If people find it valuable, I'll go all-in on it. 👉 Try it: saigalab.com Thanks for taking a look 🙏 I'll be hanging out in the comments all day, ask me anything. Kuda