To assess the diagnostic accuracy of signal loss on In-Phase (IP) Gradient-Echo (GRE) images for differentiation between Renal Cell Carcinomas (RCCs) and Lipid-Poor Angiomyolipomas (lpAMLs).
We retrospectively searched our institutional database for histologically-proven small RCCs (<5.0 cm) and AMLs without microscopic fat (lpAMLs). Two experienced radiologists assessed MR images qualitatively, to depict signal loss foci on In-Phase GRE images. A third radiologist drew Regions of Interest (ROIs) on the same lesions, on In- and Out-of-Phase (OP) images to calculate the ratio of signal loss. Diagnostic accuracy parameters were calculated for both techniques and the inter reader agreement for the qualitative analysis was evaluated using the κ test.
15 (38.4%) RCCs lost their signal on IP images, with a sensitivity of 38.5% (95% CI = 23.4-55.4), a specificity of 100% (71.1-100), a Positive Predictive Value (PPV) of 100% (73.4-100), a Negative Predictive Value (NPV) of 31.4% (26.3-37.0), and an overall accuracy of 52% (37.4-66.3%). In terms of the quantitative analysis, the Signal Intensity Index (SII= [(SI – SI) / SI] x 100) for RCCs was -0.132 ± 0.05, while for AMLs it was -0.031 ± 0.02, = 0.26. The AUC was 0.414 ± -0.09 (0.237-0.592). Using 19% of signal loss as the threshold, sensitivity was 16% and specificity was 100%. The κ value for subjective analysis was 0.63.
Signal loss in “in-phase” images, assessed subjectively, was highly specific for distinction between RCCs and lpAMLs, although with low sensitivity. The findings can be used to improve the preoperative diagnostic accuracy of Magnetic Resonance Imaging (MRI) for renal masses.
1. Signal loss on “In-Phase” gradient-echo images is a reliable sign for differentiation between Renal Cell Carcinoma and lipid-poor angiomyolipoma.