4D non-local means post-filtering for cardiac gated SPECT

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Processing

Background:

• Cardiac-gated SPECT images often exhibit increased noise due to lower data counts compared to non-gated studies.
• This noise can compromise the accuracy of myocardial wall reconstruction and visualization of perfusion defects.

Purpose of the Study:

• To investigate a spatiotemporal post-processing approach using a non-local means (NLM) filter for noise suppression in gated SPECT images.
• To evaluate the effectiveness of the 4D NLM filter in improving image quality and quantitative accuracy, particularly in scenarios with reduced imaging dose.

Main Methods:

• A 4D non-local means (NLM) filter was developed, calculating voxel output based on a weighted average of its 4D neighborhood.
• Filter coefficients adaptively adjust based on local image pattern similarity, preserving image details.
• The approach was validated using simulated SPECT data and eight clinical acquisitions, including tests with a 50% reduced imaging dose.

Main Results:

• 4D NLM filtering significantly reduced noise in gated SPECT images, enabling more accurate myocardium reconstruction.
• Compared to spatial filtering alone, 4D NLM reduced mean-squared-error by 55.63% and improved left ventricle resolution by 19.92%.
• Improved visibility of perfusion defects was observed, with comparable results at standard and reduced imaging doses, and performance similar to computationally intensive motion-compensated reconstruction.

Conclusions:

• The 4D NLM filter is an effective post-processing technique for noise reduction in cardiac-gated SPECT imaging.
• This method enhances myocardial image quality and diagnostic accuracy without significant loss of detail, offering a computationally efficient alternative.
• The NLM filter demonstrates robustness and utility even under reduced radiation exposure protocols.