An accelerated ordered subsets reconstruction algorithm using an accelerating power factor for emission tomography.

Area of Science:

• Medical Imaging
• Computational Science
• Algorithm Development

Background:

• Ordered-subset expectation maximization (OSEM) algorithms are widely used in medical imaging for tomographic reconstruction.
• Existing accelerated algorithms like E-COSEM offer limited speed enhancement.
• Convergent algorithms like COSEM ensure stable performance but can be slow.

Purpose of the Study:

• To develop a novel speed-enhanced tomographic reconstruction algorithm, ACOSEM (accelerated complete-data ordered-subset expectation-maximization).
• To accelerate the convergent COSEM algorithm without compromising image quality or requiring user-defined relaxation parameters.
• To compare ACOSEM's performance against other state-of-the-art algorithms like COSEM, E-COSEM, and RAMLA.

Main Methods:

• Derived the ACOSEM algorithm by modifying the COSEM update rule with an accelerating power factor.
• Systematically compared ACOSEM's reconstruction speed and image quality against COSEM, E-COSEM, and RAMLA.
• Evaluated ACOSEM's performance using log-likelihood values and noise studies, including comparison with the clinical standard OSEM.

Main Results:

• ACOSEM achieved comparable image quality to COSEM but with a two-fold increase in reconstruction speed using a power factor of 2.0.
• Performance was significantly faster than COSEM and E-COSEM.
• An upper limit of a power factor of 5 was observed for ACOSEM, maintaining monotonically increasing log-likelihood.

Conclusions:

• ACOSEM offers a substantial acceleration in tomographic reconstruction speed compared to existing convergent algorithms.
• The algorithm maintains image quality and convergence properties similar to COSEM.
• ACOSEM presents a promising advancement for faster and efficient medical image reconstruction.