Adaptive Autoregressive Model for Reduction of Noise in SPECT.

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Signal Processing

Background:

• Single-photon emission computed tomography (SPECT) imaging is susceptible to noise, which can degrade image quality.
• Traditional noise reduction techniques may compromise spatial resolution.
• Developing advanced filtering methods is crucial for improving SPECT image analysis.

Purpose of the Study:

• To evaluate an improved autoregressive modeling (AR) technique for noise reduction in SPECT images.
• To compare the performance of the AR-OSEM-AR method against established techniques like Butterworth filtering with filtered back projection (BW-FBP) and OSEM reconstruction (OSEM-BW).
• To assess the impact of the AR-OSEM-AR method on image quality metrics, specifically contrast resolution (CR%) and full width at half maximum (FWHM).

Main Methods:

• An autoregressive (AR) filter was applied both as a pre-filter to projection data and a post-filter to ordered subset expectation maximization (OSEM) reconstructed images, termed the AR-OSEM-AR method.
• Performance was benchmarked against the Butterworth filtering followed by filtered back projection (BW-FBP) and OSEM reconstruction followed by Butterworth filtering (OSEM-BW).
• Image quality was quantitatively assessed using a mathematical cylinder phantom with hot and cold objects across three simulated SPECT datasets, measuring CR% and FWHM.

Main Results:

• The BW-FBP method yielded the highest contrast resolution (CR%) for cold objects, while the AR-OSEM-AR method showed the lowest CR% for cold objects.
• For hot objects, the BW-FBP method demonstrated superior CR% compared to the OSEM-BW method.
• The BW-FBP method achieved the lowest full width at half maximum (FWHM) for cold objects, whereas the AR-OSEM-AR method exhibited the lowest FWHM for hot objects.

Conclusions:

• The AR-OSEM-AR method presents a viable approach for effective noise suppression in SPECT imaging.
• This method demonstrates good spatial resolution, particularly for hot objects, suggesting its utility in clinical applications.
• The findings indicate that AR-OSEM-AR can enhance the diagnostic quality of SPECT images by reducing noise while preserving resolution.