Materials and Methods
We reviewed images from all available 3-T MRI examinations performed from March 2003 to December 2005 on all patients referred because of epilepsy. Inclusion criteria were that all patients had undergone both 1.5-T and 3-T whole-brain MRI according to our epilepsy protocol between January 2000 and December 2005. Indications for repeated evaluation at 3 T varied. Although some examinations were performed because of prev iously normal or equivocal results at 1.5 T, many others were performed for lesional follow-up and surgical planning and because of scheduling con straints related to availability of the MRI unit. Studies without directly comparable high-resolution sequences were excluded, so a total of 25 patients who underwent 50 MRI examinations were included in the review. This retrospective study was approved by our institutional review board with waiver of informed consent.
The reference standard for lesion localization in the 19 patients with partial complex epilepsy was surgical confirmation in 12 cases and electroencephalographic or PET localization in conjunction with clear clinical signs in seven cases. The other six patients did not have a focal epilepsy syndrome, and their cases were used only for the qualitative assessment portion of the analysis.
Both 1.5-T and 3-T MR images of the 25 patients were reviewed independently by four experienced neuroradiologists. The images were assessed digitally with a commercially available PACS workstation (Impax version 4.5, Agfa) with real-time multiplanar reformation capabilities available to all reviewers. The multiplanar reformation function operates with a localization marker on both the source and the reformatted images. This feature was particularly helpful in assessment of the 3D T1-weighted spoiled gradient-recalled echo (SPGR) images with nearly isovoxel resolution. With this tool, the radiologist was able to assess areas suggestive of cortical thickening in directly orthogonal or perpendicular planes to rule out artifacts related to in-plane cortex. Reviewers were blinded to clinical results, including data on seizure signs, electro encephalographic findings, and other forms of localization.
A six-channel sensitivity-encoding head coil was used on both clinical 3-T units (Achieva, Philips Healthcare) for all whole-brain epilepsy imaging. Our 1.5-T MRI units (Signa Horizon and Signa LX, GE Healthcare) had a transmit-receive single-channel head coil for whole-brain imaging. Parallel-processing head coils were impractical on our two 1.5-T units for several reasons but primarily owing to degradation in image quality from inadequate signal-to-noise ratio. Identical imaging parameters therefore were not possible. Directly comparable sequences (those of the same sequence type, plane, and approximate slice thickness) used for our epilepsy protocol on the 3-T and 1.5-T MRI units were reviewed. At the time of this study, our whole-brain epilepsy protocol on all units included the following sequence parameters.
The 1.5-T protocol consisted of one 3D and three 2D sequences. The 3D images were obtained with a coronal T1-weighted SPGR sequence (TR/TE, 24/9.2; acquisition matrix, 256 x 256; field of view, 230 mm2; flip angle, 25°; slice thickness, 1.5 mm with no space). The first 2D acquisition was an axial fast spin-echo T2-weighted sequence (5,000/96.1; acquisition matrix, 256 x 256; field of view, 230 mm2; flip angle, 90°; slice thickness, 4.0-5.0 mm with 1.0-mm space). Two-dimensional fast multiplanar inversion recovery (4,500/14; inversion time, 300 seconds; acquisition matrix, 256 x 256; field of view, 180-220 mm; slice thickness, 3.0 mm with no space) and coronal FLAIR (8,802/133; inversion time, 2,200 milliseconds; acquisition matrix, 256 x 256; field of view, 220-240 mm; slice thickness, 5.0 mm with 1.0-mm space) sequences also were performed.
The 3-T protocol also consisted of one 3D and three 2D sequences. The 3D images were obtained with a coronal T1-weighted SPGR sequence (30/6; acquisition matrix, 256 x 256; field of view, 230 mm; flip angle, 45°; slice thickness, 1.2 mm with no space). The first 2D acquisition was an axial turbo spin-echo T2-weighted sequence (3,000/90; acqui sition matrix, 256 x 256; field of view, 230 mm; flip angle, 90°; slice thickness, 4.0-5.0 mm with 1.0-mm space). Two-dimensional fast multiplanar inversion recovery (3,975/20; inversion time, 250 seconds; acquisition matrix, 256 x 256; field of view, 180-220 mm; slice thickness, 2.0 mm thick with 0.2-mm space) and coronal FLAIR (11,004/120; inversion time, 2,800 milliseconds; acquisi tion matrix, 256 x 256; field of view, 220-240 mm; slice thickness, 4.0 mm with 1.0-mm space) sequences also were performed.
Four neuroradiologists experienced in interpreting epilepsy studies were asked to independently review the images from the 1.5- and 3-T studies. The viewing order was random, and to allow them the opportunity for direct comparison, reviewers were not blinded in regard to viewing both studies at the same time. Reviewers were asked to rate the 1.5- and 3-T image sets separately for the four following features: lesion conspicuity, defined as the ease with which the suspected epileptogenic focus was visible, with a specific diagnosis when possible; normal tissue contrast between gray and white matter; technical artifacts resulting in image degradation; and artifacts related to patient motion. All reviewers were blinded to clinical findings, final diagnosis, and other reviewers' interpretations. All features were rated on a 4-point scale (1, worst; 4, best) for lesion conspicuity and tissue contrast (1, worst artifacts; 4, clinically insignificant or no artifacts) for image degradation due to technical factors for both overall imaging artifacts, such as phase and susceptibility artifacts, and motion.
Analysis of variance was used to assess differences in the reported scores of lesion characterization, tissue contrast, and technical and motion artifacts. Differences in reported identification also were compared because in some cases, anatomic abnormalities were visible only at 3 T. Individual scores were used as the response variable, and p = 0.05 was considered significant. Logistic regression was used to determine the diagnostic accuracy of 3-T com pared with 1.5-T MRI through the use of two models fitted for lesion characterization, tissue contrast, and technical and motion artifacts as responses. A value of p < 0.05 was considered significant. Intraclass correlation is a measure of interrater reliability for two or more reviewers, and the significance of this value can be interpreted in a manner similar to that for kappa statistics. The 95% CI for intraclass correlation was used to assess the reliability of the four independent reviewers' scores.
Am J Roentgenol. 2008;191(3):890-895. © 2008 American Roentgen Ray Society
Cite this: Qualitative Comparison of 3-T and 1.5-T MRI in the Evaluation of Epilepsy - Medscape - Sep 01, 2008.