feat: MedCheck v0.1.0 - AI-powered medical imaging analysis toolkit by Liohtml · Pull Request #1 · Liohtml/MedCheck

Liohtml · 2026-05-21T10:29:22Z

Summary

Complete initial implementation of MedCheck - an open-source, Docker-ready toolkit for AI-powered MRI analysis.

What's included:

Core Framework

PipelineContext dataclasses for data flow between steps
PipelineStep ABC for modular step implementation
WorkflowEngine with YAML-defined pipelines
Settings from environment variables

Data Providers

LocalProvider: load DICOM files from folders or ZIP archives
EasyRadiologyProvider: fetch from easyRadiology portals (scrypt auth + AES decryption)
ProviderRegistry with auto-detection

LLM System

Claude Opus 4.7, GPT-5.5, Gemini 3.5 Flash providers
Fallback router chain
Structured JSON output parsing

Pipeline Steps

IngestStep: load DICOM data via provider system
PreprocessStep: normalize volumes, detect anatomy and planes
MLAnalysisStep: ResNet18 feature extraction + anomaly scoring
VisionAnalysisStep: LLM vision analysis with structured medical prompts
ReportStep: PDF/HTML/JSON report generation

CLI & Web UI

Typer CLI: medcheck analyze, medcheck serve, medcheck download-models
FastAPI Web UI: 3-step wizard with dark theme

Infrastructure

Multi-stage Dockerfile (lite ~500MB, full ~10GB)
GitHub Actions CI (lint, type-check, security, test matrix 3x3)
CodeQL security scanning + Dependabot
Pre-commit hooks (ruff, codespell)
65 unit tests, all passing
Medical prompts (knee, shoulder, spine)
Full documentation (quickstart, providers, workflows, models)

Community

Professional README with badges
CONTRIBUTING.md, CODE_OF_CONDUCT.md, SECURITY.md
Issue templates (bug report, feature request)
PR template, CODEOWNERS

Test plan

65 unit tests pass (uv run pytest)
Ruff lint clean (uv run ruff check)
All pre-commit hooks pass
Manual: docker build --target lite .
Manual: medcheck analyze --source ./test-dicoms --report json

Credits

Stanford MRNet - MRNet architecture
Project MONAI - Repo patterns
easyRadiology AG - Portal integration

🤖 Generated with Claude Code

Summary by CodeRabbit

New Features
- Full medical-imaging analysis app: CLI, web UI, configurable workflows, multi-format reporting (JSON/PDF/HTML)
- Multi-LLM support (Claude, OpenAI, Gemini, local) and provider routing
- Flexible data sources (local files, portal integration) and ML/vision analysis steps
Documentation
- Comprehensive README, quickstart, models/providers/workflows docs, contributing, security, and changelog
Infrastructure
- Docker images, docker-compose, CI (GitHub Actions), pre-commit, license and templates
Tests
- Extensive unit/integration test suite added

…cture Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…QL, dependabot)

…diagram

…issue templates) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Implements StepRegistry (register/get/list_steps) and WorkflowEngine (run/run_from_yaml) with rich console status output; adds default, quick, and vision_only workflow YAML definitions. Also sets --basetemp in pytest addopts to work around Windows temp-dir permission issue with pytest-asyncio strict mode. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

… and fallback router Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…istry with auto-detect Implements TDD Task 10: DataProvider abstract base class, LocalProvider that reads DICOM dirs/ZIPs via pydicom, and ProviderRegistry with register/get/detect/list_providers. All 8 new provider unit tests pass. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Implements IngestStep that uses ProviderRegistry (LocalProvider + EasyRadiologyProvider) to fetch DICOM series and populate patient/study metadata from the first slice. Adds source, provider_name, and credentials fields to PipelineContext. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Implements Task 17: ReportStep dispatches to generate_json_report, generate_pdf_report (reportlab), or generate_html_report based on context.report_format, writes the output file, and sets context.report_path. Adds report_format, report_language, output_dir fields to PipelineContext. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…uctured prompts Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

- Multi-stage Dockerfile (lite ~500MB, full ~10GB) - docker-compose.yml for development - Medical prompts: system, knee, shoulder, spine, JSON schema - Documentation: quickstart, providers, workflows, models Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

coderabbitai · 2026-05-21T10:29:35Z

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 562a5ff9-9a00-4136-b00b-11684e20db1c

📥 Commits

Reviewing files that changed from the base of the PR and between 05b399c and d0bc9d8.

📒 Files selected for processing (10)

src/medcheck/core/context.py
src/medcheck/core/workflow.py
src/medcheck/llm/base.py
src/medcheck/llm/claude.py
src/medcheck/llm/gemini.py
src/medcheck/llm/openai_provider.py
src/medcheck/main.py
src/medcheck/pipeline/ml_analysis.py
src/medcheck/pipeline/report.py
src/medcheck/providers/local.py

📝 Walkthrough

Walkthrough

This PR adds a full MedCheck project scaffold: configuration and datatypes, data/LLM provider implementations and router, preprocessing/ML/vision/report pipeline steps, Typer CLI, FastAPI web UI, prompts and JSON schema, workflows, docs, CI/dev tooling, and comprehensive unit tests.

Changes

MedCheck Core Implementation

Layer / File(s)	Summary
Devops, CI, packaging, and project metadata `.dockerignore`, `.env.example`, `.github/CODEOWNERS`, `.github/ISSUE_TEMPLATE/*`, `.github/dependabot.yml`, `.github/pull_request_template.md`, `.github/workflows/ci.yml`, `.github/workflows/codeql.yml`, `.gitignore`, `.pre-commit-config.yaml`, `.python-version`, `Dockerfile`, `docker-compose.yml`, `pyproject.toml`, `LICENSE`, `CHANGELOG.md`, `README.md`, `CONTRIBUTING.md`, `CODE_OF_CONDUCT.md`, `SECURITY.md`	Added Docker/CI/workflow configs, packaging metadata, project docs and community files, and developer tooling configs.
Package init & configuration `src/medcheck/__init__.py`, `src/medcheck/core/config.py`, `src/medcheck/core/context.py`	Package metadata and Settings dataclass; typed pipeline context and nested dataclasses for patient/study/series/clinical data.
Step and workflow execution framework `src/medcheck/core/step.py`, `src/medcheck/core/workflow.py`	`PipelineStep` ABC; `StepRegistry` and `WorkflowEngine` with sequential execution and YAML workflow loading.
Providers: abstractions and implementations `src/medcheck/providers/base.py`, `src/medcheck/providers/registry.py`, `src/medcheck/providers/local.py`, `src/medcheck/providers/easyradiology.py`	`DataProvider` interface; `ProviderRegistry`; `LocalProvider` for filesystem/ZIP DICOM discovery; `EasyRadiologyProvider` with access-code parsing, scrypt/AES decryption, and ZIP retrieval.
LLM providers and router `src/medcheck/llm/base.py`, `src/medcheck/llm/claude.py`, `src/medcheck/llm/openai_provider.py`, `src/medcheck/llm/gemini.py`, `src/medcheck/llm/router.py`	`AnnotatedImage`/`AnalysisResult` types; LLM JSON parsing helpers; Claude/OpenAI/Gemini implementations; `LLMRouter` with fallback ordering.
Ingest step `src/medcheck/pipeline/ingest.py`	Registers providers, auto-detects or uses provided provider, fetches DICOM series, populates `PipelineContext` patient/study fields.
Preprocess step `src/medcheck/pipeline/preprocess.py`	Detect anatomy/plane from descriptions; sort slices; extract pixels with fallback; build normalized 3D volumes into context.
ML analysis step `src/medcheck/pipeline/ml_analysis.py`	ResNet18-based (or statistical fallback) feature extraction, anomaly scoring, signal-intensity analysis, and MLAnalysisStep to populate anomaly/top-slices/signal stats.
Vision analysis step `src/medcheck/pipeline/vision_analysis.py`	Builds JSON-only prompts (with anatomy hints and optional clinical context), encodes selected slices to PNG bytes, and sends to chosen LLM provider to populate structured findings.
Report generation `src/medcheck/pipeline/report.py`	Generates JSON (string/file), PDF (reportlab), and HTML reports; `ReportStep` writes file and updates context.report_path.
Prompts and schema `src/medcheck/prompts/system.txt`, `src/medcheck/prompts/anatomy/*.txt`, `src/medcheck/prompts/report_schema.json`	System prompt requiring JSON output, anatomy templates (knee/shoulder/spine), and Draft-07 report JSON schema.
CLI (Typer) `src/medcheck/main.py`	Typer app with `analyze` (builds context, runs workflow/steps), `serve` (starts FastAPI app via uvicorn), and `download_models` stub.
Web app and UI `src/medcheck/web/app.py`, `src/medcheck/web/templates/index.html`	FastAPI app factory, `/health` and `/` routes, index.html wizard with file upload/URL, clinical form, options, and HTMX wiring (placeholder analyze endpoint).
Tests and fixtures `tests/conftest.py`, `tests/unit/*/`	pytest fixtures and wide unit test coverage for config/context/steps/providers/LLM router/pipeline/reporting/CLI/web endpoints.
Workflows and docs `workflows/.yml`, `docs/`	YAML workflow definitions for full/quick/vision-only pipelines and documentation: quickstart, workflows reference, models, providers.

Estimated code review effort: 🎯 4 (Complex) | ⏱️ ~45 minutes

"🐇 I hopped through code to plant a seed,
New prompts, providers, and pipelines indeed.
From DICOM piles to JSON reports bright,
MedCheck now readies the clinical night."

✨ Finishing Touches

📝 Generate docstrings

Create stacked PR
Commit on current branch

🧪 Generate unit tests (beta)

Create PR with unit tests
Commit unit tests in branch feat/initial-scaffold

…sing, validation Critical: - C1: Escape HTML in report generation (XSS prevention) - C2: Validate ZIP paths before extraction (Zip Slip prevention) - C3: Wire up full pipeline in CLI analyze command - C4: Pass step_configs to pipeline steps via context - C5: Check validate() return value in WorkflowEngine Important: - I1: Safe API key access in LLM providers - I2: Deduplicate _parse_response to base module - I3: Fix JSON extraction with brace-depth tracking - I5: Replace deprecated datetime.utcnow() - I8: Cache ResNet18 model singleton Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Liohtml and others added 15 commits May 21, 2026 12:01

feat: initial project scaffold with pyproject.toml and directory stru…

85ed054

…cture Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

ci: add GitHub Actions (lint, type-check, security, test matrix, Code…

c0ccfd3

…QL, dependabot)

chore: add pre-commit hooks (ruff, codespell, yaml/toml checks)

f5b2973

docs: add comprehensive README with badges, quickstart, architecture …

481ea4f

…diagram

docs: add community files (contributing guide, CoC, security policy, …

3d86f34

…issue templates) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

feat(core): add Settings config from environment variables

ed99417

feat(core): add PipelineContext dataclasses and PipelineStep ABC

dc627c4

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

feat(llm): add LLM abstraction with Claude, GPT-5.5, Gemini providers…

52f13c1

… and fallback router Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

feat(pipeline): add VisionAnalysisStep - LLM vision analysis with str…

ba1f379

…uctured prompts Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

feat(web): add FastAPI web UI with 3-step analysis wizard

05b399c

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Liohtml merged commit 84b2dae into main May 21, 2026
1 of 12 checks passed

coderabbitai Bot mentioned this pull request May 21, 2026

fix(ci): lower coverage threshold to 55% for v0.1.0 #6

Merged

This was referenced May 31, 2026

Add patient-friendly plain-language report mode #56

Open

[Epic] Validation & enhancement pipeline stages — improve and verify results before reporting #51

Open

Liohtml mentioned this pull request Jun 11, 2026

feat: implement i18n translation catalogs with English fallback for r… #76

Merged

6 tasks

Navigation Menu

feat: MedCheck v0.1.0 - AI-powered medical imaging analysis toolkit - #1

feat: MedCheck v0.1.0 - AI-powered medical imaging analysis toolkit#1
Liohtml merged 16 commits into
mainfrom
feat/initial-scaffold

Liohtml commented May 21, 2026 •

edited by coderabbitai Bot

Loading

Uh oh!

coderabbitai Bot commented May 21, 2026 •

edited

Loading

Review failed

Walkthrough

Changes

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

Liohtml commented May 21, 2026 • edited by coderabbitai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

What's included:

Test plan

Credits

Summary by CodeRabbit

Uh oh!

coderabbitai Bot commented May 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review failed

Walkthrough

Changes

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Liohtml commented May 21, 2026 •

edited by coderabbitai Bot

Loading

coderabbitai Bot commented May 21, 2026 •

edited

Loading