https://rabimba.github.io/tags/webgpu/WebGPUHugo Blox Builder (https://hugoblox.com)en-usMon, 18 May 2026 00:00:00 +0000https://rabimba.github.io/media/icon_hu_6d534efae7b021f2.pnghttps://rabimba.github.io/tags/webgpu/https://rabimba.github.io/projects/drmri-ai/Mon, 18 May 2026 00:00:00 +0000https://rabimba.github.io/projects/drmri-ai/<p>Dr.MRI.AI is a web-based DICOM viewer built on a simple observation: the hard part of medical AI isn&rsquo;t the model, it&rsquo;s knowing what to send it. A typical MRI study contains 200+ slices, but only a handful carry the signal. Dr.MRI.AI plans the slice selection before any multimodal analysis runs.</p> <p><strong>How it works — Plan → Review → Synthesize:</strong></p> <ul> <li><strong>Plan</strong>: Gemma 4 produces a structured plan (which series, slice range, window/level) from your clinical question in plain English. You can accept, edit, or reject it.</li> <li><strong>Review</strong>: The multimodal model analyzes only the focused subset — 200+ slices reduced to ~16 frames, ~99% fewer image tokens.</li> <li><strong>Synthesize</strong>: A final text-only pass produces a structured report with clickable slice references; export findings plus reviewed evidence as one bundle.</li> </ul> <p><strong>Privacy-first architecture:</strong></p> <ul> <li>DICOM files are processed entirely in the browser — no data is uploaded to any server</li> <li>Fully browser-local inference via <code>onnx-community/gemma-4-E2B-it-ONNX</code> through Transformers.js + WebGPU</li> <li>Provider-agnostic: Gemini API, OpenAI-compatible endpoints, Ollama (MedGemma planning + Gemma 4 review), or fully on-device Gemma 4 Browser</li> </ul> <p>Built with React 19 + TypeScript + Vite and Cornerstone3D for medical image rendering. Featured in the Google Developer Experts May–Jun 2026 AI community roundup. Educational/research tool — not certified as a medical device.</p>