メニュー
Ground CXR findings using MI2 zero-shot classification. Returns per-finding probabilities bucketed into positive/possible for the agent to use when writing the report.
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/microsoft/healthcareai-examples --skill ground-zeroshotコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
Ground CXR findings using the trained MI2 adapter. Returns per-finding positive/possible/negative calls for the agent to use when writing the report.
Entry point for the MedImageInsight plugin, covering the getting-started walkthrough, how the skills work, what MI2 is, and troubleshooting.
Read a chest X-ray end-to-end: ground findings with MedImageInsight, then write a structured radiologist report (Indication, Technique, Findings, Impression). Invoke for any "write/draft a CXR report" request.
SKILL.md を表示中
SOC 職業分類に基づく
| name | ground-zeroshot |
| description | Ground CXR findings using MI2 zero-shot classification. Returns per-finding probabilities bucketed into positive/possible for the agent to use when writing the report. |
| argument-hint | - image_path: required file path or http(s) URL to the CXR image (PNG or JPEG) - extra_labels: optional list of additional label strings to score alongside the default set |
Calls the MI2 zero-shot classifier and returns bucketed per-finding probabilities. Does not write a report — the agent produces the report using the write-report skill after grounding.
mcp_medimageinsig_zeroshot_classify call per image.Resolve the image. If image_path is an http(s) URL, pass it through to the tool unchanged — do NOT download it, curl it, or turn it into a local file; the MCP server fetches URLs itself. Otherwise it is a local path: the MCP tool runs in a different working directory, so if it is not absolute, resolve it via realpath; if it can't be resolved, ask the user.
Assemble the label list. Read the canonical label list from below and use those strings verbatim. Append any caller-provided extra_labels.
Call the classifier. Call mcp_medimageinsig_zeroshot_classify with file_path = absolute path and labels = assembled list. One call per image.
Bucket the scores. Use these cutoffs:
Return the findings. Return the positive and possible labels with their raw scores. Do not write a report. The agent decides what to do next.
Pass these strings as the labels argument to zeroshot_classify.
Format is "<modality> <body_part> <view> <condition>" — MI2 is sensitive to
this structure, do not reword.
x-ray chest anteroposterior Atelectasis
x-ray chest anteroposterior Cardiomegaly
x-ray chest anteroposterior Consolidation
x-ray chest anteroposterior Edema
x-ray chest anteroposterior Effusion
x-ray chest anteroposterior Emphysema
x-ray chest anteroposterior Fibrosis
x-ray chest anteroposterior Hernia
x-ray chest anteroposterior Infiltration
x-ray chest anteroposterior Mass
x-ray chest anteroposterior Nodule
x-ray chest anteroposterior Pleural_Thickening
x-ray chest anteroposterior Pneumonia
x-ray chest anteroposterior Pneumothorax