README.md 초안

README.md

@@ -1,3 +1,96 @@
----
-license: cc-by-nc-4.0
----
+
+# MISAW-Seg: Pixel-level Surgical Tool Segmentation in Microsurgical Anastomosis
+
+## Dataset Overview
+
+MISAW-Seg is a surgical segmentation dataset that extends the original [MISAW dataset](https://www.synapse.org/Synapse:syn21776936/files/) by introducing semantic segmentation annotations for microsurgical tools involved in artificial vessel anastomosis tasks. The original dataset included kinematic data, workflow annotations, and stereo video recordings, but lacked pixel-wise annotations for surgical tools. 
+
+This dataset provides new segmentation masks created using the Roboflow platform, enabling semantic segmentation research in microsurgical environments.
+
+## Data Collection
+
+The MISAW-Seg dataset extends the original MISAW dataset, which consists of microsurgical training sessions involving artificial vessel anastomosis. The data was collected by the Department of Mechanical Engineering at the University of Tokyo using a master-slave robotic platform (Mitsuishi et al., 2013). Both kinematic data and stereo video data were synchronously acquired at 30 Hz.
+
+Workflow annotations were created by the MediCis team at the LTSI Laboratory, University of Rennes, using the “Surgery Workflow Toolbox” provided by IRT b<>com (Garraud et al., 2014). These annotations include labels for surgical phases, steps, and left/right hand activities based on action, target, and tool.
+
+While the original dataset included kinematic, video, and workflow data, it did not include any pixel-level annotations for surgical instruments. In this work, we manually annotated surgical tools using the Roboflow platform and created semantic segmentation masks for each frame, enabling tool segmentation tasks in microsurgical environments.
+
+## Data Details
+
+Data in MISAW-Seg is organized into a directory tree for better accessibility. Each directory stores raw image data, segmentation masks, and annotation files.
+
+### Directory Structure
+
+MISAW-Seg/
+├── images/ # Raw surgical frames (460x540 px)
+│ ├── frame_0001.png
+│ ├── frame_0002.png
+│ └── ...
+├── masks_png/ # Pixel-wise semantic segmentation masks (PNG format)
+│ ├── frame_0001.png
+│ ├── frame_0002.png
+│ └── ...
+├── annotations/
+│ └── misaw_seg.json # COCO-style annotation file
+└── README.md
+
+### Image Format
+
+- Extracted from stereo-microscope videos (original resolution: 960×540 px)
+- The center 40 pixels were removed to correct for stereo alignment, resulting in 920×540 px
+- One side (left or right) was cropped to generate 460×540 px frames
+- Frame rate: 30 fps
+
+### Annotation Formats
+
+1. **COCO Format** (`annotations/misaw_seg.json`)
+
+   The COCO-style annotation file contains polygon-based segmentations, bounding boxes, and category IDs for each object. Additional metadata such as surgical phase and step can also be optionally included in the `image` field.
+
+2. **PNG Format** (`masks_png/`)
+
+   Pixel-level semantic segmentation masks are provided in PNG format. Each mask shares the same filename as the corresponding image (e.g., `frame_0001.png`) and stores class IDs as pixel values.
+
+### Segmentation Classes
+
+The dataset contains the following classes:
+
+| Class Number   | Class Name              | RGB Color        |
+|----------------|-------------------------|------------------|
+| 0              | Left artificial vessel  | (0, 255, 0)      |
+| 1              | Left needle holder      | (255, 255, 0)    |
+| 2              | Needle                  | (0, 255, 255)    |
+| 3              | Right artificial vessel | (0, 0, 255)      |
+| 4              | Right needle holder     | (255, 0, 0)      |
+| 5              | Wire                    | (255, 0, 255)    |
+
+## Annotation Method
+
+Segmentation masks were manually created using the Roboflow platform. Annotators performed frame-level labeling of surgical instruments and maintained consistency across temporal frames. Annotated data was reviewed to ensure high quality and anatomical accuracy in the segmentation boundaries.
+
+![Annotation Process](./annotation_screenshot.png)
+
+The annotation process included the following steps:
+
+- Video frames were extracted from stereo-microscope recordings at 30 fps.
+- Annotators manually drew polygon masks for each visible surgical instrument using Roboflow's editor.
+- The annotation tool allowed for fine-grained control over vertex placement and class assignment.
+- Each frame was labeled with one or more objects belonging to six predefined surgical tool categories.
+- Annotated masks were reviewed for spatial consistency and object completeness.
+- Final masks were exported in both PNG (pixel-wise class map) and COCO JSON (polygon-based) formats.
+
+Although the annotators were not medical professionals, they followed a consistent labeling guide and used tool appearance and continuity across frames to ensure annotation quality.
+
+## License
+
+---
+license: cc-by-nc-4.0
+---
+
+## Citation
+
+If you use this dataset in your research, please cite:
+
+```bibtex
+@misc{}
+```