Extraction of time activity curves from gated FDG-PET images for small animals’ heart studies

Abstract

We introduce a new approach to extract the input function and the tissue time activity curve from dynamic ECG-gated ¹⁸F-FDG PET images. These curves are mandatory to model the myocardium metabolic rate of glucose for heart studies. The proposed method utilizes coupled active contours to track the myocardium and the blood pool deformations. Furthermore, a statistical approach is developed to model the blood and tissue activities and to correct for spillovers. The developed algorithm offers a reliable alternative to serial blood sampling for small animal cardiac PET studies. Indeed, the calculated MMRG value differs by 1.54% only from the reference value.

Introduction

The plasma time activity curve, also called Input Function (IF), is mandatory to calculate metabolic and physiologic parameters through kinetic modeling [1], [2]. The reference method to determine IF is invasive, as it is a manual blood sampling procedure [3], [4]. For small animals, the [¹⁸F]-2-deoxy-2-fluoro-d-glucose-fluoro (FDG) PET studies procedure is more complex and challenging because of their small blood vessels size and the limited blood volume. Furthermore, blood loss may perturb the physiology and confound the experimental outcome [5]. As an alternative, image and population-based IF is used for diverse studies in PET imaging [6], [7]. The IF can also be obtained directly from data by means of regions of interest (ROI) drawn over the blood pool in the case of cardiac images [8], [9], [10]. In principle, this method is relatively simple to implement; however, in small animal imaging, hearts and arteries are small compared to the scanner's spatial resolution. Moreover, cardiac and lung motions degrade the image by blurring the vascular radioactivities into adjacent tissues. The extracted IF curve from the ROI is therefore a mixture of the true IF and the surrounding tissue activity. Alternative methods of estimating IF from images have been proposed. Buvat et al. [11] have used the factor analysis of dynamic structures (FADS) method, which decomposes the dynamic sequences into several factor images. The IF is then extracted from the blood factor image [12], [13]. The Principal Component Analysis (PCA) and the Independent Component Analysis (ICA) techniques have also been applied to extract IFs from PET images [14], [15], [16]. ICA acts in similar fashion as FADS, but uses a de-mixing matrix to isolate components from a data mixture. In some studies, authors have presented a method of IF extraction from the so-called image-derived input function (IDIF) [17], [18]. IF can also be obtained by a simultaneous estimation method based on a multi-exponential time-activity function scaled to the measured activity concentration in a limited number of blood samples [19].

This paper proposes a new approach to extract the IF and the tissue time activity curve (TAC) for the purpose of kinetic modeling from PET images using a coupled active contour (AC) and a statistical analysis step. To reduce the blurring effect that result from heart and lung motions, we propose to use ECG-gated PET measurements, that are divided into 16 gates describing the cardiac cycle. In this framework, the TACs extraction is carried out independently for each gate. However, given the deformation of the heart during the cardiac cycle, the ROIs where the TACs are extracted must be updated dynamically. To carry out this task, a coupled active contour model is developed and applied to track the blood pool and the myocardium's motion. This model integrates through its parameterization prior knowledge on the cardiac systolic and diastolic phases. The coupled AC requires only a single initialization in the first gate. For the subsequent gates, the initialization is done automatically. Furthermore, our solution allows for sub-pixel precision to compensate for the low resolution of the small animals PET images.

The blood and tissue regions delineated by the active contours are statistically analyzed in order to estimate the TACs for spillover correction. For this purpose, a Bayesian approach is developed to extract the fractions of true blood and true tissue activities for each pixel in the ROIs. These fractions are estimated using prior knowledge on the tracer‘s behavior over time in the blood's and in the tissue's responses according to the kinetic model. Unlike the ICA and FADS based solutions, our method does not require a reduction of the data's dimension, nor does it need intermediate steps as oblique analysis for FADS or a cost function definition for ICA. Consequently, it provides a Myocardial Metabolic Rate of Glucose (MMRG) value for each pixel in the image, which allows a spatial localization of any sector abnormalities.