Missing head and color banding in low-count SPECT reconstructions

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Image Reconstruction

Background:

• Low photon counts in (111)In SPECT scans can lead to significant image artifacts.
• Philips Extended Brilliance Workspace (EBW) and IntelliSpace Portal (ISP) workstations exhibited issues with image reconstruction algorithms.
• Integer image storage caused underflow problems, resulting in intensity resolution loss and color banding.

Purpose of the Study:

• To investigate and resolve image reconstruction artifacts, specifically missing data in SPECT scans with low photon counts.
• To address underflow issues and intensity scaling problems encountered on Philips EBW and ISP workstations.
• To evaluate different reconstruction algorithms and pre-processing techniques for improving low-count SPECT image quality.

Main Methods:

• Reconstruction algorithms compared included Ordered Subsets Expectation Maximization (OSEM) with and without Astonish (resolution recovery), Filtered Backprojection (FBP), and Maximum Likelihood Expectation Maximization (MLEM).
• Artifacts were analyzed on Philips EBW2.0 and ISP5.02 workstations.
• Solutions explored included applying a Hann pre-filter, upscaling projection data, and limiting the reconstructed volume.

Main Results:

• OSEM (with/without Astonish) produced missing head artifacts in low-count SPECT images, unlike FBP and MLEM.
• A Hann pre-filter effectively resolved the missing head artifact on Philips workstations.
• Underflow problems were partially mitigated in newer workstation versions, but upscaling raw projection data before reconstruction provided a more robust solution, albeit with potential minor statistical deviations.

Conclusions:

• Reconstruction of low-count SPECT studies requires careful algorithm selection and parameter tuning to avoid artifacts.
• The Hann pre-filter is a viable solution for missing data artifacts in OSEM reconstruction on Philips systems.
• While newer software versions improve underflow handling, pre-reconstruction data upscaling remains a reliable strategy for managing intensity scaling issues in low-count SPECT imaging.