Acceleration of maximum likelihood reconstruction, using frequency amplification and attenuation compensation.

Area of Science:

• Medical Imaging
• Computer Science
• Physics

Background:

• Expectation-maximization algorithms for maximum likelihood (ML) and maximum a posteriori (MAP) reconstruction are standard in SPECT and PET.
• These algorithms offer theoretical advantages and guaranteed convergence but suffer from slow processing times due to slow convergence.

Purpose of the Study:

• To introduce and evaluate two novel heuristic acceleration methods for ML reconstruction in emission computed tomography (ECT).
• To compare the performance of these new methods against the established 1D line search method by Lewitt.

Main Methods:

• Developed a frequency-dependent amplification method to counteract backprojection's low-pass filtering effect.
• Incorporated an amplification factor to mitigate the impact of attenuation on updating factors.
• Compared these methods with Lewitt's 1D line search using test images.

Main Results:

• All three methods demonstrated acceleration of the ML algorithm.
• Lewitt's method provided the most significant acceleration among the individual methods tested.
• A combined approach using frequency amplification and line search yielded the best performance improvements.

Conclusions:

• The proposed heuristic methods effectively accelerate ML reconstruction for ECT.
• Combining frequency amplification with line search presents a promising strategy for enhancing ECT image reconstruction efficiency.
• Potential for further acceleration exists by selectively skipping ML calculations under specific conditions.