Mean relative errors for standard uptake values in PET images reconstructed using MR attentuation correction (AC) with reference images reconstructed using CTAC for VOIs in the thorax. Image source: J Nucl Med 2011;52(9):1392-1399.

The MRI-based attenuation correction (MRAC) method using atlas registration and pattern recognition (AT&PR) provides better overall PET quantification accuracy than the basic MR image segmentation approach, according to a study in the September issue of the Journal of Nuclear Medicine.

PET/MRI is an emerging dual-modality imaging technology that requires new approaches to PET attenuation correction (AC). The researchers assessed two algorithms for whole-body MRAC: a basic MR image segmentation algorithm and a method based on AT&PR.

Matthias Hofmann, MD, from the department of preclinical imaging and radiopharmacy at Eberhard Karls University in Tübingen, Germany, and colleagues assessed 11 patients. Each underwent a whole-body PET/CT study and a separate multi-bed whole-body MRI study.

The MR image segmentation algorithm uses a combination of image thresholds, Dixon fat-water segmentation and component analysis to detect the lungs. MR images are segmented into five tissue classes (not including bone), and each class is assigned a default linear attenuation value. The AT&PR algorithm uses a database of previously aligned pairs of MRI/CT image volumes.

For each patient, these pairs are registered to the patient MRI volume, and machine-learning techniques are used to predict attenuation values on a continuous scale. According to the researchers, MRAC methods are compared via the quantitative analysis of AC PET images using volumes of interest in normal organs and on lesions. They assumed the PET/CT values after CT-based AC to be the reference standard.

“In practice, it is not immediately clear what tissue class an MR image segmentation approach will predict for bone areas,” the study authors wrote. “In our MR image segmentation approach, for example, the bone marrow in the vertebrae of the spine was often identified as fat via Dixon segmentation. This misclassification is one of the reasons why the PET quantification errors obtained for images reconstructed with AC using the MRI-derived pseudo-CT image PsCT_MRSEG were higher than for those reconstructed via AC using the CT-derived CT_seg when both were compared with the PET images reconstructed after CT AC.”

In regions of normal physiologic uptake, Hofmann and colleagues reported that the average error of the mean standardized uptake value was 14.1 percent and 7.7 percent for the segmentation and the AT&PR methods, respectively. Lesion-based errors were 7.5 percent for the segmentation method and 5.7 percent for the AT&PR method.
 
“This better quantification was due to the significantly reduced volume of errors made regarding volumes of interest within or near bones and the slightly reduced volume of errors made regarding areas outside the lungs,” the authors concluded.