Least-squares dual characterization for ROI assessment in emission tomography

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Quantitative Analysis

Background:

• Accurate statistical property estimation in emission tomography regions of interest (ROIs) is crucial for quantitative analysis.
• Traditional methods often rely on image reconstruction, which can introduce errors and biases.

Purpose of the Study:

• To present an original dual formulation method for direct estimation of ROI activity and variance from measured emission tomography data.
• To compare the proposed method with classical ROI estimation techniques and Huesman's method.

Main Methods:

• Developed a dual formulation of the ROI estimation problem based on approximate inverse theory.
• Defined an ROI characteristic function derived from co-registered morphological images, with optional smoothing for resolution-variance tradeoff.
• Implemented an iterative least-squares dual (LSD) characterization and a linear extrapolation scheme (LSD-ex) to reduce bias.

Main Results:

• LSD characterization demonstrated performance comparable to classical methods in terms of root mean square (RMS) error.
• LSD-ex reduced estimation bias by up to 14% and RMS error by up to 8.5% for small tumor regions.
• LSD with smoothing effectively managed the resolution-variance tradeoff for a large non-specific region.

Conclusions:

• The proposed dual formulation (LSD and LSD-ex) offers an effective alternative for ROI statistical property estimation in emission tomography.
• LSD-ex significantly improves bias and RMS error for small ROIs, outperforming classical methods.
• The method provides a flexible approach to optimize the resolution-variance tradeoff.