A blind deconvolution method incorporated with anatomical-based filtering for partial volume correction: Validations...

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Image Processing

Background:

• Conventional partial volume correction (PVC) for SPECT imaging typically requires contrast-enhanced CT segmentation and point spread function (PSF) measurement.
• These prerequisites can be complex and time-consuming, limiting the widespread application of accurate SPECT quantification.

Purpose of the Study:

• To develop and validate a novel segmentation-free method for SPECT PVC.
• The method utilizes blind deconvolution (BD) combined with anatomical-based filtering to simultaneously estimate the restored image and the PSF.

Main Methods:

• An iterative BD algorithm was employed to estimate the restored SPECT image and the PSF.
• Anatomical-based filtering was integrated within each iteration to mitigate Gibbs artifacts and reduce noise amplification during deconvolution.
• Validation was performed using ¹²³I-metaiodobenzylguanidine (¹²³I-mIBG) SPECT/CT imaging in NCAT and physical cardiac phantoms, as well as a human study.

Main Results:

• The proposed method significantly reduced bias and noise in SPECT images, with mean relative bias improvements in simulated normal myocardium and myocardial defects.
• Signal-to-noise ratio (SNR) in simulated defects increased by approximately 38-41%.
• Correlation between SPECT-quantified and true heart-to-mediastinum ratios (HMRs) was markedly improved after applying the BD-based PVC, with estimation errors significantly reduced in both phantom studies and a human study.

Conclusions:

• The developed segmentation-free PVC method effectively enhances SPECT quantification accuracy and reduces image noise.
• This approach eliminates the need for pre-measuring the image PSF, offering a more streamlined and potentially more accessible method for SPECT image analysis.