PlaTiF: Tibial Plateau Fracture Dataset

This dataset was developed to address the lack of publicly available, expert-annotated radiographic data for tibial plateau fractures. It includes anteroposterior (AP) knee radiographs and coronal CT slices from real-world clinical cases collected at Shariati Hospital, affiliated with Tehran University of Medical Sciences. Each fracture is categorized according to the Schatzker classification system and includes expert-reviewed tibial bone segmentation masks. The dataset is intended to support research in AI-driven fracture detection, classification, and preoperative surgical planning, and was inspired by the growing need for open-access orthopedic imaging resources to advance clinical decision support tools.

🧠 Patient Dataset Structured for Tibial Plateau Fracture:

📋 Patient Demographics and Clinical Metadata:
An accompanying Excel file titled Tibial Plateau Fracture Metadata.xlsx is included in the dataset. 

📂 Dataset Path

.\Patient Data_Part 1

.\Patient Data_Part 2

.\Patient Data_Part 3

.\Patient Data_Part 4

Each .mat file corresponds to a unique patient and contains all data related to their X-ray and segmentation masks.

📌 FILE NAMING CONVENTION:

▶ Format:          Patient_ID_XXX.mat

▶ Example:         Patient_ID_001.mat, Patient_ID_204.mat

▶ Description:     XXX is a 3-digit unique patient ID derived from CT/X-ray folder structure.

📦 FILE CONTENT STRUCTURE:

Each .mat file contains a variable named Patient_ID_XXX, which is a struct with the following format:

📁 Patient_ID_XXX

├── 🧾 im0

│                  ├── 🖼️ OriginalImage   → Original X-ray image  
│                  ├── ⚫ BW                     → Binary mask of segmented tibial plateau  
│                  ├── 🖼️ maskedImage    → X-ray masked with segmentation  
│                  └── 🏷️ label                  → Class label (1–7) for Schatzker fracture type  

├── 🧾 im1

│              └── ...

└── 📐 Coronal_CT (optional) → Associated CT image if available

✔ imX fields (im0, im1, im2, ...) represent multiple views/images for each patient.

🧠 USAGE NOTES:

  • This dataset is designed for AI-driven Schatzker fracture classification  
  • Fields are consistent across all patients  
  • Images and masks are spatially aligned  
  • Use the label field in im0 for patient classification or dataset grouping

🧾 SUMMARY:

• 🔢 Patients:                One .mat per patient  
• 🖼️ Image views:         Multiple (im0, im1, ...)  
• 🧩 Segmentation:      Binary mask per image  
• 🏷️ Label Classes:        1 to 6 (Schatzker types)+7 (No fractures)  
• 📦 Optional CT:          Coronal CT image per patient  
• 💾 Format:                  MATLAB .mat (v7 or higher)