Optimization of Butterworth filter for brain SPECT imaging

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Signal Processing

Background:

• Brain SPECT imaging utilizes filters like Butterworth to enhance image quality by separating signals from noise.
• The effectiveness of these filters is influenced by the total counts in projection images, a factor that varies between subjects.
• Current methods may not adequately account for subject-specific total counts when setting filter parameters.

Purpose of the Study:

• To investigate the relationship between total counts and optimal Butterworth filter cutoff frequencies in brain SPECT.
• To develop a method for optimizing filter cutoff frequencies based on individual subject data.
• To improve the visual quality and diagnostic accuracy of brain SPECT images.

Main Methods:

• Utilized 99mTc hexamethyl-propyleneamine oxime (HMPAO) in a normal volunteer to acquire projection sets with varying total counts.
• Generated high-quality reference images from projection sets with 300-second acquisition times per projection.
• Optimized Butterworth filter cutoff frequencies by assessing mean square errors and visually inspecting filtered reconstructed images.

Main Results:

• A clear dependence between total counts and optimal cutoff frequencies was established.
• A nomogram was created to visually represent this relationship, allowing for estimation of optimal cutoff frequencies.
• Individualized optimization of the Butterworth filter cutoff frequency based on total counts was shown to maximize visual image quality.

Conclusions:

• The cutoff frequency of the Butterworth filter in brain SPECT imaging should be determined on a per-study basis, referencing the total counts.
• Individualized filter optimization can lead to superior image quality in brain SPECT.
• This approach offers a practical method for enhancing diagnostic performance in SPECT studies.