MedVision Reconstructor

MedVision Reconstructor is a Python application designed for reconstructing medical images via Radon Transform, Filtered Back-Projection (FBP), and Simultaneous Algebraic Reconstruction Technique (SART). The project includes a user-friendly interface built with PySide6 that lets you load medical images, select reconstruction parameters (angle range, filters), and visualize both the sinogram and reconstructed images.

---

Features

1. Interactive GUI (PySide6)

   • Intuitive interface for loading images and setting reconstruction parameters.
   • Real-time visualization of sinograms and reconstruction outputs.

2. Multiple Reconstruction Methods

   • Filtered Back-Projection (FBP) with multiple filters:
     • ramp, shepp-logan, cosine, hamming, hann
   • SART (Simultaneous Algebraic Reconstruction Technique)

3. Radon Transform & Sinogram Generation

   • Leverages scikit-image for computing radon transforms.

4. Animation of Incremental Reconstruction

   • Enables an angle-by-angle view of how reconstruction quality evolves.

5. Test Script

   • A standalone script (test.py) that demonstrates sinogram creation and reconstruction with FBP or SART.

---

Project Structure

Below is the directory structure for MedVision-Reconstructor:

MedVision-Reconstructor/
 ├── images/                         # Folder containing sample medical images
 ├── medical_env/                    # (Optional) Virtual environment folder
 ├── src/
 │    ├── main.py                    # Main GUI application (PySide6)
 │    ├── MI.py                      # Reconstruction class (Radon, FBP, SART)
 │    └── test.py                    # Test script for demonstration
 ├── .gitignore
 ├── LICENSE
 ├── README.md
 └── requirements.txt

---

Installation

1. Clone the Repository

   git clone https://github.com/elifpulukcu/MedVision-Reconstructor.git
   cd MedVision-Reconstructor

2. (Optional) Create & Activate a Virtual Environment

   python -m venv medical_env
   source medical_env/bin/activate      # On Unix / Mac
   # or
   medical_env\Scripts\activate         # On Windows

3. Install Dependencies

   pip install --upgrade pip
   pip install -r requirements.txt

   If no requirements.txt exists, install the main libraries manually:

   pip install PySide6 matplotlib scikit-image numpy

---

Usage

1. Launching the GUI

Run the main application to open the GUI:

cd src
python main.py

Key Interface Elements:

1. Load Image
   • Select .png, .jpg, .jpeg, or .bmp images from your filesystem.
2. Maximum Angle
   • Enter the maximum angle (e.g., 180) for radon transform sampling.
3. Reconstruction Method
   • Choose among the following filters for FBP:
     • ramp, shepp-logan, cosine, hamming, hann
   • Or select SART for algebraic reconstruction.
4. Calculate
   • Computes the sinogram and reconstructs the image based on the selected method.
   • Displays the sinogram and reconstructed image.
5. Show Animation
   • Iterates through angles in steps (from 10° up to the specified angle) to visualize progressive reconstruction.

2. Test Script

A separate test script (test.py) demonstrates the reconstruction steps in a non-GUI fashion:

cd src
python test.py

This script will:

• Load a specified image from the images/ directory (you can change the path in the code).
• Generate a sinogram using the Radon Transform.
• Perform Filtered Back-Projection (FBP) and SART reconstructions.
• Display the original image, the sinogram, and reconstruction results side by side.

---

Implementation Details

MI.py (Core Reconstruction Logic)

1. Class MI
   • Constructor: Accepts image, maxAngle, and filterName. Converts RGB to grayscale if necessary.
   • processImage: Rescales the image (default scale factor of 0.4) and prepares angles for radon transform.
   • radonTransform: Performs the Radon Transform (scikit-image.transform.radon).
   • filteredBackProjection: Uses iradon with the specified filter to reconstruct the image from its sinogram.
   • sart: Applies iradon_sart for algebraic reconstruction.

main.py (GUI)

1. PySide6 Window
   • Load Image button uses a QFileDialog to select and display an image.
   • Calculate calls either filteredBackProjection or sart, then renders the sinogram and reconstruction with matplotlib.
   • Show Animation demonstrates incremental reconstruction from small angles up to the user-specified maximum.

test.py (Testing & Demonstration)

1. Command-Line Runner
   • Reads an image from images/.
   • Generates sinogram and performs both FBP and SART reconstructions.
   • Displays comparative results via matplotlib.

---

Planned Improvements

1. Algorithmic Expansion

   • Investigate other iterative methods like ART, TV-based methods, or deep-learning approaches.

2. Performance Optimization

   • Potential GPU acceleration using CuPy or PyTorch for large-scale reconstructions.

3. Dataset & Demo

   • Include more diverse images in images/ to showcase various organ/tissue reconstructions.

4. Parameter Tuning

   • User-adjustable SART parameters (relaxation, iterations) through the GUI.

5. CLI Extension

   • Offer a robust command-line interface for batch processing or pipeline integration.

---

License

Distributed under the MIT License. Feel free to use, modify, and distribute this code for personal or commercial projects, subject to the terms of the license.

---

Disclaimer

This repository is intended for educational and research purposes. The authors do not guarantee the suitability of this software for clinical or diagnostic use.

---

Author

Developed by Elif Pulukçu.
Contributions and pull requests are welcome—please open an issue or submit a PR for improvements.