Microsoft Ignite 2026 hands-on lab. This repository supports a hands-on lab on building radiology AI agents in VS Code with GitHub Copilot, grounded in MedImageParse — a foundation model that segments medical images from simple text prompts. The lab follows one brain MRI case from a bare LLM to a measured, context-aware report, and is organized into three chapters, each in its own folder under chapters/.
Budget about 30 min setup (do this before arriving) + 60 min content across the three chapters.
🚧 Work in progress for Ignite 2026. The lab structure, agents, skills, and MCP server scaffolding are in place, and the source MRI case has been selected. The MedImageParse calls, lesion measurement code, exported PNG slices, and eval results are placeholders — search the repo for
PlaceholderandTODO(mip)to find them.
- Chapter 1 — Simple agent + skill (
chapters/chapter1-simple-agent/): use a general-purpose LLM with a prompt-based skill to draft structured brain MRI reports — and see where it falls short. - Chapter 2 — Using foundation models with agents (
chapters/chapter2-foundation-models/): wrap MedImageParse lesion segmentation, plus a measurement tool that turns masks into millimeters, in an MCP server the agent can call. - Chapter 3 — Patient context (
chapters/chapter3-knowledge-sources/): ground agent outputs in a (synthetic) patient history and a prior report, so the agent reports interval change.
chapters/— one folder per chapter. Each has aREADME.md, astarter/skeleton, and ananswer-key/reference.samples/— brain MRI you can hand to the agent: a glioblastoma follow-up case from MU-Glioma-Post (CC BY 4.0), available as NIfTI volumes, DICOM series, and per-slice PNGs. See samples/README.md.eval-results/— reference scoring numbers for the chapter-to-chapter comparisons (🚧 placeholder — see eval-results/README.md).docs/— troubleshooting, Codespaces, and no-admin setup guides.
The data and patient records in this repo are synthetic or de-identified and are not for clinical use. The brain MRI comes from the MU-Glioma-Post dataset on The Cancer Imaging Archive, used under CC BY 4.0. See samples/nifti/README.md for the required citation. The patient chart wrapped around it is fictional. MedImageParse is provided for research and model development exploration; it is not designed or intended to be deployed in clinical settings as-is.
Install and verify the following before the workshop. Expect about 30 minutes of setup. Do this before arriving — Wi-Fi and time will be tight on the day.
🚧 Placeholder — lab environment. Confirm whether Ignite attendees get a pre-provisioned lab VM or bring their own machine, and trim this section accordingly.
-
VS Code (latest stable): https://code.visualstudio.com/
-
GitHub Copilot and GitHub Copilot Chat extensions, signed in with an account that has Copilot access.
-
Git, and a local clone of this repository opened as your VS Code workspace.
-
uv (https://docs.astral.sh/uv/) — Python package manager. Install user-scope (no admin required):
- macOS / Linux:
curl -LsSf https://astral.sh/uv/install.sh | sh - Windows (PowerShell):
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
Then run
uv syncfrom the repo root. - macOS / Linux:
Can't install locally? See docs/no-admin.md or run the whole lab in a GitHub Codespace.
Chapter 2 calls a shared MedImageParse endpoint. Credentials (MIP_MODEL_ENDPOINT and MIP_MODEL_API_KEY) are distributed at the start of the session. You'll set them up in Chapter 2 Setup; the working path will be mcp/medimageparse-server/.env at workspace root.
The workshop/ directory at repo root holds shared Python utilities for the workshop, including a small script to verify your env is healthy.
uv --version # 0.x.x
uv sync # one-time, ~1–2 min on a good connection
uv run workshop/check_env.py # expected: "Workshop env OK — Python 3.11"If any command above fails, resolve it before arriving.