Neuroimaging Reader Study on Clinical Sensitivity and Specificity Using Synthetic MRI Based on MR Quantification

Abstract

Background and purpose: Prior large prospective studies have shown 2D synthetic MR image quality to be similar to that of conventional MR imaging across a wide variety of normal and abnormal subjects. This study is the first large, prospective comparison of 3D synthetic to conventional MR imaging performed in routine clinical settings.

Materials and methods: A prospective, multicenter, multireader clinical investigation to evaluate the diagnostic performance of 3D synthetic images based on the novel synthetic MRI 3D-QALAS method compared with conventional 3D MRI was performed. Five blinded neuroradiologists evaluated the 3D synthetic and conventional MR images of 189 subjects who presented at clinical sites for neuroimaging. The objectives were to compare sensitivity and specificity of pathologic findings, accuracy of imaging findings, image quality, legibility of anatomic structures, artifact prevalence, and interrater and intermethod agreement of synthetic 3D MR images with conventional 3D images.

Results: Both sensitivity and specificity of 3D synthetic and conventional images for pathologic findings and diagnostic accuracy of radiologic findings were very similar, 66/68% and 85/85%, respectively. The individual values for the readers varied between 58%-73%/58%-75% for sensitivity and 72%-98%/77%-94% for specificity. No significant difference could be determined between the methods. Ratings (1-5 scale) of synthetic MR image quality were higher than for T1-weighted images (mean score of 4.6 ± 0.6, with 98.6% having a 3 or higher versus 4.5 ± 0.7, with 97.8% having a 3 or higher) and T2-weighted images (4.4 ± 0.7, with 98.0% having a 3 or higher versus 4.2 ± 0.8, with 97.5% having a 3 or higher).

Conclusions: In this study no significant differences were found in the sensitivity and specificity of pathologic findings and accuracy of imaging findings between 3D synthetic T1-weighted and T2-weighted images and 3D conventional images. In addition, 5 blinded neuroradiologists rated 3D synthetic MR images of higher image quality than conventional 3D images.