Three-region MRI-based whole-body attenuation correction for automated PET reconstruction

Abstract

Introduction

In this study we proposed and developed a simple attenuation mapping approach based on magnetic resonance imaging (MRI) for the purpose of reconstructing positron emission tomography (PET) images in PET/MRI imaging devices.

Methods

After experimental development, an in vivo calibration was performed by whole-body scanning of five beagles on both a PET/CT and an MRI. The attenuation was determined by using an automated segmentation algorithm to segment regions of background, lung, soft tissue and bone, and assigning them values of 0.002, 0.030, 0.098 and 0.130 cm⁻¹, respectively.

Results

The CT-attenuated and MRI-attenuated PET images had average standardized uptake values (SUVs) that differed by 1–6% for most regions of interest (ROIs). Also, mean relative differences (MRDs) between the images were between 5% and 9% for most regions. The only exception is bone, where the three-region MRI-attenuated PET images had an SUV 10% less on average than the CT-attenuation images, and the MRD averaged 14%. Also, additional segmentation of the bone in the four-region MRI-attenuated PET images reduced the SUV difference to 3% and the MRD to 6%.

Conclusion

Therefore, despite the improvements in the four-region segmentation, the three-region segmentation, without delineation of osseous tissues, produces high-quality images that are sufficient for most expected clinical and research purposes.

Introduction

Hybrid imaging, the combination of two distinct imaging systems, is a highly useful diagnostic tool that has been researched extensively in recent years [1], [2], [3]. Imaging systems such as positron emission tomography (PET)/computed tomography (CT) and single photon emission computed tomography (SPECT)/CT are widely used in both clinical and research settings due to the combined functional information concerning metabolic processes (PET, SPECT) and anatomical detail (CT). There is a lot of interest in simultaneous acquisition of high-resolution morphologic imaging with molecular imaging systems. Whereas PET/CT and SPECT/CT are acquired sequentially, the integration of PET/MRI is seen as one of the opportunities to do this in a simultaneous way [4], [5], [6]. Magnetic resonance imaging (MRI) also has several advantages over CT, including high soft tissue contrast, the ability to scan in all three dimensions, higher resolution, no ionizing radiation, and can be used with a wide variety of contrast materials. Several prototypes have been developed in research groups around the world including a 7-T MRI combined with an animal PET scanner that allows for simultaneous acquisition [7], [8], [9].

Attenuation correction is a necessary step in PET reconstruction to obtain adequate images [10]. Images without attenuation correction have poor contrast and artifacts, such as high signal in the lungs and on the surface of the body. The attenuation can be determined using a transmission scan, which uses one or more rotating radioactive sources to measure the attenuation along all the lines of response. The emergence of PET/CT hybrids has allowed the attenuation to be measured using CT, eliminating the need for an additional transmission scan [11], [12], [13], [14]. Likewise, PET/MRI systems could theoretically use the MRI scan to obtain the attenuation correction. This would eliminate the need for CT and transmission scans, which increase the radiation exposure to the patient and would require additional time and money to integrate them into the system. Therefore, the most feasible way to acquire attenuation information in PET/MRI systems is to approximate the attenuation from the MRI scan.

Acquiring attenuation information from MRI is challenging due to its measurement method. Whereas CT and transmission scans directly measure the attenuation of ionizing radiation passing through the body, MRI measures proton density and magnetic characteristics of various nuclei. One problem in particular is that proton MRI signal does not allow visualization of calcium and phosphorus, which are some of the strongest absorbers of ionizing radiation. This poses a technological problem due to the difficulty of distinguishing bone from surrounding tissues. However, some researchers have developed sequences for high bone contrast using ultrashort TE imaging, but the additional scan would be unfavorable due to the added time to the protocol [15], [16]. Zaidi et al. [17] and Hofmann et al. [18] have developed methods for segmenting the bone from surrounding tissues in the brain, but no one has yet presented a method for adequately segmenting bone from the entire body. It has also been suggested by Martinez-Möller et al. [19] that segmentation of the bone for attenuation corrections is not necessary.

This study concerns evaluating the effectiveness and feasibility of estimating the attenuation based on segmentation of the acquired MRI images. This is achieved by approximating the attenuation by segmenting and categorizing tissue groups based on the pixel intensities of certain regions. The resulting PET images reconstructed using approximate MRI-based attenuation maps are compared to PET images reconstructed using CT-based attenuation maps. The goal of the study was to measure the effectiveness of a simple approximation of the attenuation on the resulting PET images.