Vascular tree reconstruction with discrete tomography: intensity based camera correction for 3D reconstruction

Area of Science:

• Medical Imaging
• Computational Biology
• Image Reconstruction

Background:

• Vascular tree reconstruction from limited projections is crucial for medical diagnosis.
• Discrete tomography methods face challenges with computational cost and data inconsistency.
• Existing methods lack robustness against positional errors and motion artifacts.

Purpose of the Study:

• To develop a robust method for vascular tree reconstruction from few, potentially inconsistent X-ray projections.
• To improve the accuracy and reduce the computational cost of vascular tree reconstruction.
• To handle positional inconsistencies and small motion artifacts effectively.

Main Methods:

• A data-driven approach minimizing reconstruction residual for enhanced robustness.
• Integration of an intensity-based 2D/3D-registration method with iterative reconstruction.
• A volume-splitting scheme for parallel processing to accelerate reconstruction speed.

Main Results:

• Accurate reconstruction of liver vascular trees from few, inconsistent projections.
• A 13% decrease in average reconstruction residual for real data after preprocessing.
• Reconstruction time decreased linearly with increased processor cores, demonstrating significant speed-up.

Conclusions:

• The method effectively reduces inconsistencies from positioning errors and motion artifacts.
• Intensity-based algorithms eliminate the need for prior segmentation or correspondence detection.
• Enables robust, high-quality vascular tree reconstructions with reduced radiation dose.