Chapter 51 - Generalized Linear Least-Squares Modeling Algorithm for Optimally Sampled PET Image Data

Abstract

Positron emission tomography (PET) is distinguished from several other medical imaging modalities as it quantifies physiological and biochemical functions within the body. With the application of parametric imaging techniques, visual representations of these functions may be obtained. The original generalized linear least-squares (GLLS) algorithm is used to obtain parametric images as it is computationally fast, tracer and model configuration independent, and can produce estimation as accurate as the standard model-fitting method. However, when PET data are sampled according to an optimal sampling schedule (OSS) to reduce measurement samples while preserving the amount of information needed for accurate estimation of physiological parameters, the direct application of GLLS is not reliable as instantaneous measurements can no longer be approximated by averaging of accumulated measurements over sampling intervals.