USPIO-enhanced MR imaging for nodal staging in patients with primary rectal cancer: predictive criteria.

ABSTRACT

PURPOSE: To prospectively determine diagnostic performance of predictive criteria for nodal staging with ultrasmall superparamagnetic iron oxide (USPIO)-enhanced magnetic resonance (MR) imaging in primary rectal cancer patients, with histopathologic findings as reference standard. MATERIALS AND METHODS: Institutional review board approval and informed consent were obtained. Twenty-eight rectal cancer patients (18 men, 10 women; mean age, 68 years) underwent USPIO-enhanced MR. Two observers with different experience evaluated each node on three-dimensional T2*-weighted images for border irregularity, short- and long-axis diameter, and estimated percentage (<30%, 30%-50%, or >50%) of white region within the node. Ratio of measured surface area of white region within the node to measured surface area of total node (ratio(A)) was calculated. Signal intensity (SI) of gluteus muscle (SI(GM)), total node (SI(TN)), and white (SI(WR)) and dark (SI(DR)) regions within the node were used to calculate SI(TN)/SI(GM) and SI(WR)/SI(DR) ratios. Lesion-by-lesion, receiver operating characteristic curve, and interobserver agreement analyses were performed. The most accurate and practical criterion was evaluated by observer 3. RESULTS: In 28 patients, 236 lymph nodes were examined. Area under the receiver operating characteristic curve (AUC) of estimated percentage of white region and ratio(A) were 0.96 and 0.99 (observer 1) and 0.95 and 0.97 (observer 2), respectively. AUC of estimated percentage criterion for observer 3 was 0.96. AUC of border, short- and long-axis diameter, and SI(TN)/SI(GM) and SI(WR)/SI(DR) ratios were 0.65, 0.75, 0.79, 0.85, and 0.75 (observer 1) and 0.58, 0.75, 0.79, 0.89, and 0.79 (observer 2), respectively. Interobserver agreement (kappa value) for estimated white region between observers 1 and 2, 1 and 3, and 2 and 3 were 0.77, 0.79, and 0.84, respectively. For observers 1 and 2, kappa value for border was 0.28. For observers 1 and 2, intraclass correlation coefficient for short- and long-axis diameters, ratio(A), and SI(TN)/SI(GM) and SI(WR)/SI(DR) ratios were 0.91, 0.96, 0.91, 0.72, and 0.92, respectively. CONCLUSION: Estimated percentage of white region and measured ratio(A) are the most accurate criteria for predicting malignant nodes with USPIO-enhanced MR imaging; the first criterion is the most practical.