WalnutPCCTReconCodes

This repository provides MATLAB code for loading, correcting, reconstructing, and performing spectral analysis on projection data from the Walnut Photon-Counting CT (PCCT) Dataset, acquired using a custom micro-cone-beam PCCT system. The dataset includes multi-energy raw projections of 15 walnut samples.

📖 This repository accompanies our scientific publication:
Zhou, E., Li, W., Xu, W. et al. A cone-beam photon-counting CT dataset for spectral image reconstruction and deep learning. Sci Data 12, 1955 (2025). https://doi.org/10.1038/s41597-025-06246-4

---

📸 Data Examples

1. Visualizing Projection Data (ImageJ)

Raw projection files (.raw) can be opened using ImageJ:

• File → Import → Raw
• Width: 2063
• Height: 505
• Data type: 16-bit unsigned
• Byte order: Little-endian

This allows direct inspection of photon-counting projections.

---

2. Reconstruction and Spectral Imaging Examples

Below are example results demonstrating the reconstruction workflow and spectral outputs.

🔹 Reconstruction Workflow and Results

• Left: Reconstruction pipeline overview
• Right: Example reconstructed CT images

---

🔹 Material Decomposition and Virtual Monoenergetic Imaging

• Left: Material decomposition results (shell and pulp)
• Right: Virtual monoenergetic images at different energies

---

🔧 System Requirements

Due to the large volume of high-resolution projection data and memory-intensive reconstruction tasks, the following system configuration is recommended:

• MATLAB R2024a or later
• 64 GB RAM or more
• GPU with CUDA support and ≥8 GB memory (e.g., NVIDIA RTX 2080 or above)
• Windows 64-bit OS (precompiled MEX files provided for this platform)

---

📦 Dependencies

This codebase relies on the TIGRE Toolbox, an open-source GPU-accelerated CT reconstruction library supporting FDK and iterative algorithms.

🔁 Quick Installation

To automatically install and configure TIGRE, simply run the requirements.m file:

requirements

---

🧩 Repository Structure

WalnutPCCTReconCodes/
├── requirements.m                  # One-click setup for TIGRE dependency
├── WalnutDataRecon.m              # Main script for projection correction and CT reconstruction
├── WalnutSpectralRecon.m          # Main script for material decomposition and VMI
├── /mexfiles/                     # Precompiled MEX files for Windows 64-bit
├── /functions/                        # Supporting functions (correction, recon, etc.)
└── /pictures/                         # sample figures

---

🚀 Core Functionalities

1. 🌀 Projection-Domain Correction & Reconstruction

Run the WalnutDataRecon.m to reconstruct high, low, and total energy bin images with optional artifact correction: You can configure:

• Reconstruction algorithm (FDK, SART, MLEM, etc.)
• Angular sampling (full/sparse views)
• Energy bin (Total / High / Low)
• Whether to apply:
  • Non-uniformity correction (STEPC)
  • Ring artifact removal
  • 3D-TV denoising

2. 🧪 Image-Domain Spectral Reconstruction

For material decomposition and virtual monoenergetic imaging (VMI), run the WalnutSpectralRecon.m Set monoenergetic image energies via:

recon_para.WalnutVMI_E = 10:10:80;

Reconstructed results include:

• Walnut shell and pulp decomposition
• Energy-dependent VMI volumes

---

📊 Example Use Cases

• Deep learning model training for material decomposition or sparse-view reconstruction
• Detector calibration studies (e.g., bad pixel correction, ring artifact correction)
• Virtual monoenergetic imaging (VMI) synthesis for contrast analysis
• Spectral CT algorithm benchmarking using real PCCT data

---

📎 Citation

If you use this dataset or code, please cite:

Zhou, E., Li, W., Xu, W. et al. A cone-beam photon-counting CT dataset for spectral image reconstruction and deep learning. Sci Data 12, 1955 (2025). https://doi.org/10.1038/s41597-025-06246-4

---

📮 Contact

For questions or feedback, please open an issue or contact email: zhou.en.ze@qq.com.

---

📑 License

This project is licensed under the MIT License. Note: The included TIGRE framework is licensed under BSD.