Area weighted convolutional interpolation for data reprojection in single photon emission computed tomography

Area of Science:

• Medical Imaging
• Computer Science
• Algorithm Development

Background:

• Single photon emission computed tomography (SPECT) imaging requires accurate data reprojection for image reconstruction.
• Existing interpolation algorithms like nearest-neighbor and linear methods have limitations in accuracy and computational efficiency.
• Parallel processing hardware offers potential for accelerating complex imaging algorithms.

Purpose of the Study:

• To develop and evaluate a novel data reprojection algorithm for SPECT imaging.
• To enhance the accuracy and computational efficiency of SPECT data reprojection on parallel processing systems.
• To introduce the area weighted convolution (AWC) method as an improved interpolation technique.

Main Methods:

• Developed a data reprojection algorithm utilizing a uniform square pixel model for accurate pixel activity representation.
• Implemented a pixel-driven nearest-neighbor projection to subdivided projection bins.
• Convolved the subdivided bin array with the angle-dependent projection of a uniform square pixel, followed by compression to original bin size (Area Weighted Convolution - AWC).

Main Results:

• The AWC algorithm demonstrated higher accuracy compared to nearest-neighbor and linear interpolation, approaching the accuracy of the line length algorithm.
• AWC provided a more efficient and easier implementation on parallel hardware.
• The AWC method achieved faster execution times than both line length and linear interpolation algorithms.

Conclusions:

• The Area Weighted Convolution (AWC) algorithm is a superior method for data reprojection in SPECT imaging.
• AWC offers significant improvements in accuracy and computational performance, particularly on parallel processing systems.
• This algorithm facilitates more efficient and accurate SPECT image reconstruction.