Quantitative accuracy of the closed-form least-squares solution for targeted SPECT

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Computational Science

Background:

• Single photon emission computed tomography (SPECT) is a crucial nuclear medicine imaging technique.
• Quantitative accuracy in SPECT reconstructions is often limited by computational costs and image resolution.
• Targeted reconstruction offers a potential solution by focusing on regions of interest (ROI).

Purpose of the Study:

• To investigate the quantitative accuracy of the closed-form least-squares solution (LSS) for SPECT.
• To evaluate the effectiveness of targeted reconstruction for oncology SPECT scans with localized high tracer uptake.
• To compare ROI-based LSS with the conventional Maximum Likelihood Expectation Maximization (MLEM) approach.

Main Methods:

• Applied targeted reconstruction by disconnecting system equations and extracting ROI-specific components.
• Generated closed-form LSS for each target ROI.
• Compared ROI-based LSS with MLEM reconstructions in phantom experiments mimicking oncology SPECT scans.

Main Results:

• LSS approach demonstrated superior performance in reducing noise levels within ROIs compared to MLEM.
• MLEM achieved better total activity recovery with a sufficient number of iterations.
• ROI-based LSS provided superior spatial activity distribution in ROIs for experiments without background activity.

Conclusions:

• Targeted reconstruction using ROI-based LSS shows promise for improving quantitative accuracy in SPECT, particularly in oncology.
• LSS offers advantages in noise reduction and spatial resolution within targeted regions.
• Further research may explore optimizing LSS for various clinical scenarios and comparing it with other advanced reconstruction algorithms.