A tailored ML-EM algorithm for reconstruction of truncated projection data using few view angles.

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Reconstruction

Background:

• Dedicated cardiac single photon emission computed tomography (SPECT) systems offer high speed and sensitivity.
• These systems can suffer from data truncation and limited view angles due to a small field of view (FOV).
• Traditional maximum-likelihood expectation-maximization (ML-EM) algorithms produce artifacts like streaks and distortion with truncated SPECT data.

Purpose of the Study:

• To develop a modified ML-EM algorithm to mitigate artifacts in cardiac SPECT reconstruction.
• To address image quality issues arising from data truncation and insufficient angular sampling.

Main Methods:

• A tailored ML-EM algorithm was proposed, adjusting image updating step sizes for pixels outside the FOV.
• The algorithm's convergence speed for external pixels was intentionally decelerated.
• The method was tested on analytical, Monte Carlo simulation, and real cardiac SPECT emission data with varying view numbers.

Main Results:

• The tailored ML-EM algorithm effectively suppressed streak artifacts originating outside the FOV.
• Distortion within the FOV was also reduced compared to the conventional ML-EM algorithm.
• The proposed method demonstrated superior performance in reconstructing images from truncated projection data with limited views.

Conclusions:

• The tailored ML-EM algorithm is a viable solution for improving cardiac SPECT image quality.
• This modification effectively handles artifacts caused by data truncation and limited angular sampling.
• The enhanced reconstruction quality can lead to more accurate cardiac diagnoses.