PET Quantification of Ultra Low Activity via Inhomogeneous Poisson Process Parameters Estimation Directly from Listmode Data

TL;DR

This paper introduces the I3PE method for PET quantification, modeling coincidence events as an Inhomogeneous Poisson Process directly from listmode data, enabling detection of ultra low activity levels in PET imaging.

Contribution

The paper presents a novel I3PE method that improves PET quantification accuracy at low activity levels by directly estimating Poisson process parameters from listmode data.

Findings

I3PE method effectively models PET coincidence events as an Inhomogeneous Poisson Process.

The method enables detection of nano-Curie activity levels with high accuracy.

Experimental results demonstrate improved quantification at ultra low activity levels.

Abstract

Metabolic imaging with PET/CT using $^{18}$F-Fludeoxyglucose ($^{18}$F-FDG) as well as other imaging biomarkers has achieved wide acceptance in oncology, cardiology and neurology not only because of the unique metabolic information generated by this modality, but also because of its ability to quantify biological processes. However, PET quantification is affected by many technical and physiologic factors, and then recognized as an important problem for diagnosis, determination of prognosis, and response monitoring in oncology. In this work, we investigated the effect of reduced PET emission count statistics on the accuracy and precision of tracer quantification, and proposed Inhomogeneous Poisson Process Parameter Estimation (I3PE) method. In I3PE method, we modelled the coincidence event as Inhomogeneous Poisson Process, and estimate its parameter directly from the streaming listmode data. To evaluate the effectiveness, a experiment using $^{18}$F-FDG was implemented on LIGHTNING. The results not only demonstrate I3PE method, but also evaluated the minimal detectable activity of the using PET machine. According $0.3\%$ mean error rate criterion, LIGHTNING can detect several nano-Curie, cooperated with I3PE method.