aat-roi-review-tool

KO | EN

gitlite — search

#python #react #ai #golang #go #linux #docker #nodejs #kubernetes #awesome-list #pytorch #mcp #ios #openai

aat-roi-review-tool

(★ 23)

Web-based ROI and lane annotation review tool for AAT image datasets.

aat-roi-review-tool 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# AAT IEF Gel ROI/Lane Review Tool

Local research tool for reviewing automatic ROI and vertical lane segmentation on Alpha-1 Antitrypsin (AAT) isoelectric focusing (IEF) gel images.

This repository contains the code, configuration, and documentation needed to run the review tool. It does not include real gel images by default.

## What This Tool Does

- scans gel images placed in `dataset/Originial/`
- assigns stable `IMG_0001` style image IDs
- generates vertical projection metadata
- runs the v6 horizontal-ROI-aware lane boundary pipeline
- serves a local FastAPI and static Web review interface
- lets users adjust shared ROI and lane boundaries
- supports lane-level category annotation
- exports reviewed geometry and training CSV files

This is a human-in-the-loop research tool. It is not a clinical diagnostic device.

## Repository Layout

```text
Web/                  FastAPI backend and static review UI
scripts/              Dataset inventory, projection, segmentation, and overlay scripts
src/                  Shared image loading helpers
configs/              Segmentation pipeline configuration files
docs/                 User and data-policy documentation
dataset/Originial/    Place local gel images here; real images are ignored by Git
dataset/metadata/     Inventory and review metadata CSV files
outputs/              Generated automatic segmentation outputs; ignored by Git
```

## Requirements

- Windows 10/11 or equivalent Python environment
- Python 3.12 recommended
- A browser

Python dependencies are listed in `requirements.txt`.

## Quick Start

From the repository root:

```powershell
start_web.bat
```

Then open:

```text
http://127.0.0.1:8000
```

Manual setup is also supported:

```powershell
python -m venv .venv
.\.venv\Scripts\python.exe -m pip install -r requirements.txt
.\.venv\Scripts\python.exe -m uvicorn app:app --app-dir Web --host 127.0.0.1 --port 8000
```

## Screenshots

The screenshots below show the internal research review workflow with real example gel images from the private project environment. Raw source images are still excluded from Git.

![Main review interface](https://raw.githubusercontent.com/BrilliantWhite/aat-roi-review-tool/main/docs/assets/screenshots/01-main-review-interface.png)

Main review interface with the image list, automatic ROI, and lane boundary overlays.

![Update dataset action](https://raw.githubusercontent.com/BrilliantWhite/aat-roi-review-tool/main/docs/assets/screenshots/02-update-dataset-action.png)

The `Update dataset` action rescans local images and reruns the automatic v6 segmentation pipeline.

![ROI and lane review](https://raw.githubusercontent.com/BrilliantWhite/aat-roi-review-tool/main/docs/assets/screenshots/03-roi-and-lane-review.png)

ROI and lane overlays are reviewed visually before being accepted as training or restore data.

![Lane annotation modal](https://raw.githubusercontent.com/BrilliantWhite/aat-roi-review-tool/main/docs/assets/screenshots/04-lane-annotation-modal.png)

Clicking a lane opens the annotation modal for category entry and lane-level inspection.

![Export review results](https://raw.githubusercontent.com/BrilliantWhite/aat-roi-review-tool/main/docs/assets/screenshots/05-export-review-results.png)

Reviewed geometry and lane annotations can be saved and exported from the action panel.

## Adding Images

1. Put supported image files into:

```text
   dataset/Originial/
   ```

2. Supported formats:

```text
   .png
   .jpg
   .jpeg
   .bmp
   ```

3. Open the Web tool and click `Update dataset`.

The update action rebuilds the inventory, creates conservative quality placeholders for new images, regenerates projection metadata, reruns v6 segmentation, and refreshes the Web image list.

## Review Workflow

1. Select an image from the left list.
2. Inspect automatic ROI and lane boundaries.
3. Drag ROI or lane boundaries if needed.
4. Click a lane to open the annotation modal.
5. Enter a lane category when known.
6. Save reviewed geometry.
7. Export restore or training CSV files when needed.

Reviewed outputs are written under:

```text
Web/review_exports/
```

These files are ignored by Git by default because they may contain research data.

## CSV Import and Export

- `review_restore_export.csv` is the reproducible segmentation file. Use `Export restore CSV` to create it, and `Import restore CSV` to restore reviewed ROI/lane geometry later.
- `lane_annotations_review.csv` is the lane-level annotation sidecar. Use `Import annotation CSV` when you only want to load category labels as a temporary edit session.
- `training_lanes_export.csv` is the downstream training table. It can also be imported as annotation input, but it is not the preferred restore file because its purpose is model training rather than full review-state recovery.
- `Restore exported CSV` restores the existing `Web/review_exports/review_restore_export.csv` back into the reviewed ROI/lane/annotation CSV files. Use this after resetting images if you need to recover the last exported restore snapshot.

## Command-Line Pipeline

The Web `Update dataset` button runs this sequence:

```text
scripts/build_image_inventory.py
scripts/generate_vertical_projections.py
scripts/detect_lane_roi_boundaries_v6.py
scripts/create_lane_roi_boundary_overlays_v6.py
```

The same scripts can be run manually from the repository root with the project virtual environment.

## Tests

Run:

```powershell
.\.venv\Scripts\python.exe -m unittest test_build_image_inventory.py test_dataset_refresh.py
```

## Data and Ethics

Real clinical or research gel images should not be committed to a public repository unless the responsible institution has approved sharing. See `docs/data_policy.md`.

## Current Maturity

The tool is suitable for internal research review and controlled collaboration. It still needs broader validation, stronger packaging, and clearer clinical governance before being treated as a mature external product. See `docs/tool_maturity_assessment.md`.

## Known Limitations

- Automatic v6 segmentation still requires human visual review.
- ROI placement can fail on images with strong headers, labels, page marks, or unusual crops.
- Lane category labels are user-supplied and not inferred clinically.
- The project does not currently provide a validated phenotype classifier.

## License

Choose and add a project license before public release.

aat-roi-review-tool

aat-roi-review-tool 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

aat-roi-review-tool

aat-roi-review-tool 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README