# Appendix300: Surgical video and patient metadata of 330 laparoscopic appendectomy cases from five institutions

## 📝 Project Description

**Appendix300** is a curated, multi-institutional dataset for Surgical Data Science (SDS) applications comprising laparoscopic video recordings of appendectomies and comprehensive clinical metadata.

### ✨ Key Features

- **Multicenter Data:** Data from five distinct surgical centers.
- **Integrated Metadata:** Linked case data comprising intraoperative video, preoperative clinical symptoms, preoperative laboratory parameters, intraoperative appendicitis grading, and postoperative histopathology.
- **Expert Annotations:** Temporal labels for appendix visibility and consensus-based severity grading.

## 📂 Dataset Structure

The dataset is organized to link video files directly to clinical spreadsheets for easy ingestion into machine learning pipelines.

```
Appendix300/
├── Full_Videos/
│   ├── Center1/
│   │   ├── Center1_001/
│   │   │   └── Center1_001.mp4     # Re-encoded, de-identified laparoscopic video
│   │   └── ...
│   ├── Center2/
│   ├── Center3/
│   ├── Center4/
│   └── Center5/
├── Images/
│   ├── Center1/
│   │   ├── Center1_001/
│   │   │   └── frames.zip          # Up to 200 equidistant sampled frames
│   │   └── ...
│   ├── Center2/
│   ├── Center3/
│   ├── Center4/
│   ├── Center5/
│   └── unpack.py
├── Video_Snippets/
│   ├── Center1/
│   │   ├── Center1_001.mp4         # 100-second clip centered on dissection start
│   │   └── ...
│   ├── Center2/
│   ├── Center3/
│   ├── Center4/
│   └── Center5/
├── overall_merged.csv              # Combined metadata index across all centers
└── readme_Appendix300.md
```

## 📊 Data Details

### 3.1 Study Population

| Group              | Count | Description                             |
|--------------------|-------|-----------------------------------------|
| **Total Patients** | 330   | Multi-institutional cohort              |
| **Cases**          | 325   | Laparoscopic appendectomies             |
| **Controls**       | 5     | Non-appendectomy laparoscopic surgeries |

### 3.2 Metadata Variables

| Category           | Parameters Included                                                                                                                      |
|--------------------|------------------------------------------------------------------------------------------------------------------------------------------|
| **Preoperative**   | Age, Sex, BMI, preoperative symptoms (pain migration to right lower quadrant, anorexia, Nausea/vomiting, point tenderness in the right lower quadrant, rebound peritonism, body temperature), patient history (previous intraabdominal surgery, active tumor disease, pregnancy, duration of preoperative antibiotic treatment), laboratory parameters (hemoglobin, leukocytes, granulocytes, PMN cells, CRP)                                                                          |
| **Intraoperative** | Timestamps at which the appendix is fully visible prior to invasive preparation, the corresponding video filename (`Video` column), and laparoscopic grade (Gomes et al. scale) |
| **Postoperative**  | Histopathological grade, presence of appendiceal carcinoid                                                                               |

Details on clinical data collection, cleaning, preprocessing, and annotation are available in the Scientific Data data descriptor for this dataset. 

### 3.3 Video Specifications

- **De-identification:** All extracorporeal sequences have been removed.
- **Full Videos:** Re-encoded MP4s organized by center. Most cases have one file per case (`CenterX_NNN.mp4`); cases with multiple source recordings are split into versioned files (`CenterX_NNN_v01.mp4`, `CenterX_NNN_v02.mp4`, …). The `Video` column in `overall_merged.csv` links each metadata row to the specific file containing the annotated timestamp.
- **Snippets:** 100-second windows centered around the start of dissection, stored flat per center (`Video_Snippets/CenterX/CenterX_NNN.mp4`).
- **Images:** Up to 200 equidistant frames per case, stored as `frames.zip` (use `Images/unpack.py` to extract).

## 🚀 Usage & Task Examples

This dataset supports a variety of computational modeling tasks:

- **Inflammation Grading:** Classifying appendicitis severity based on intraoperative (laparoscopic) and postoperative (histopathologic) grades.
- **Generalization Studies:** Evaluating model performance across different hospital environments and hardware.

## 📑 Citation

If you use this dataset, please cite both the data descriptor (outlining details on data collection, cleaning, preprocessing, and annotation), and the original manuscript (benchmarking Swarm Learning for decentralized patient-level prediction tasks). 

## 🔐 Access & Licensing

- **License:** [CC BY 4.0](http://creativecommons.org/licenses/by/4.0/)
- **Contacts:**
  - Fiona R. Kolbinger: [fiona.kolbinger@tu-dresden.de](mailto:fiona.kolbinger@tu-dresden.de)
  - Max Kirchner: [max.kirchner@nct-dresden.de](mailto:max.kirchner@nct-dresden.de)
  - Oliver Lester Saldanha: [oliverlester.saldanha@tu-dresden.de](mailto:oliverlester.saldanha@tu-dresden.de)