Image non-uniformity causes substantial problems especially in abdominal imaging using a 3T scanner.^1,2 We evaluated two commercially available methods for non-uniformity correction in Gd-EOB-DTPA-enhanced MR imaging using a 3T scanner.

MR imaging: Twenty patients who underwent Gd-EOB-DTPA-enhanced imaging on a GE 3T clinical scanner were retrospectively analyzed. Dynamic imaging and hepatobiliary-phase imaging were performed using a liver acquisition with volume acceleration (LAVA) sequence. For hepatobiliary-phase imaging, the same tuning parameters (receiver gain, transmitter gain, center frequency, and gradient shim) as dynamic imaging were entered manually to ensure direct comparison of signal intensities.

Non-uniformity correction: Image non-uniformity correction was performed using two methods provided by the manufacturer: surface coil intensity correction (SCIC) and phased-array uniformity enhancement (PURE)³. We generated uncorrected images first, and uniformity correction was applied retrospectively to compare three sets of images created from the same image data. PURE utilizes proton-density-weighted images acquired with both the body coil and the surface coil to calibrate the coil sensitivity. SCIC is an image-based method and requires no additional scan.

Data analysis: Superficial hyperintensity and focal hyperintensity in the liver were assessed visually in the precontrast and hepatobiliary-phase axial images to determine the signal uniformity in the uncorrected, SCIC, and PURE images. The most prefereble image set was selected for each phase of a given patient. The histogram of the liver signal intensity was assessed for the precontrast and hepatobiliary-phase images. The entire liver were demarcated manually on seven slices, and a histogram was created with 15-point smoothing. The signal range showing frequencies of more than half of the frequency at the mode were determined, and the width of the range were divided by the mode signal to calculate full-width at half-maximum (FWHM) of the histogram as a marker of uniformity. The signal intensities for the liver, muscle, and spleen were evaluated before and after contrast administration. Three circular ROIs were set for the liver, avoiding superficial hyperintensity on SCIC images, two circular ROIs were set for the back muscle, and one circular ROI was set for the spleen. The liver-to-muscle signal ratio (LMR) and liver-to-spleen signal ratio (LSR) were calculated at each phase. Contrast enhancement ratio was calculated from the liver signal (CER-Liver), LMR (CER-LMR), and LSR (CER-LSR), dividing the postcontrast value by the precontrast value.

Examples of images are presented in Figure 1. Superficial hyperintensity was observed in the uncorrected images of all patients with no correction but in no PURE images. With SCIC, it was shown in 8 and all 20 patients for the precontrast and hepatobiliary-phase images, respectively. Focal hyperintensity was noted only in the PURE images (16 and 13 patients in the precontrast and hepatobiliary-phase images, respectively), and was located in the lateral segment in all cases. The SCIC and PURE images were judged the most preferable in 14 and 6 patients for the precontrast images, respectively, and in 1 and 19 patient(s) in the hepatobiliary-phase images.

On histogram analysis, FWHM were significantly smaller for the SCIC and PURE images, indicating better uniformity, than for the uncorrected images (Fig. 2). In the comparison of the two correction methods, precontrast FWHM was significantly smaller for the SCIC images, whereas hepatobiliary-phase FWHM was significantly smaller for the PURE images.

Uniformity correction largely influenced the estimates of LMR and LSR (Fig. 3). The PURE images yielded larger values, especially for LMR, compared to the uncorrected and SCIC images. LMR and LSR remained relatively constant irrespective of the imaging phase. The CERs were identical between the uncorrected and PURE images as predicted theoretically. The SCIC images provided quite different profiles of signal enhancement (Fig. 4). CER-Liver in the SCIC images were definitely lower than in the uncorrected images. CER-LMR was close to 1 irrespective of the imaging phase, and CER-LSR also tended to converge to 1.

SCIC allowed favorable non-uniformity correction on the precontrast images; however, superficial hyperintensity was evident espeicially in the hepatobiliary phase. Without correction, underestimation of LMR is considered due to overestimation of the muscle signal located superficially. SCIC did not remove the underestimation of LMR, and distrurbed the assessement of the temporal changes in the quantitative parameters of contrast enhancement. Postcontrast SCIC images should be interpreted considering such effects and should not be used for quantitative evaluation. PURE improved uniformity in the precontrast and hepatobiliary-phase images, increasing LMR and preserving temporal signal changes after contrast administration. PURE is indicated to be a useful non-uniformity correction method in Gd-EOB-DTPA-enhanced MR imaging using a 3T scanner.