Iron-oxide-enhanced magnetic resonance imaging of atherosclerotic plaques: postmortem analysis of accuracy, inter-observer agreement, and pitfalls.

Abstract

Contrast-enhanced magnetic resonance (MR) imaging using ultra small superparamagnetic iron oxide (USPIO) particles is a new noninvasive modality for imaging inflammatory atherosclerotic plaques. We determined the accuracy, interobserver agreement, and potential sources of error of this technique by means of postmortem MR imaging of aortic preparations. Anesthetized atherosclerotic Watanabe heritable hyperlipidemic (WHHL) rabbits were studied after administration of different dosages of intravenous USPIO (DDM 43/34, IDF Berlin, Germany) and different postcontrast time intervals. A (n = 5) received 0 micromol Fe/kg. B (n = 5) received 50 micromol Fe/kg, 8-hour postcontrast interval. C (n = 5) received 50 micromol, 24 hours. D received 200 micromol, 48 hours. The aortas were removed and 3-mm segments prepared for postmortem examination by MR imaging using a T2-weighted gradient-echo sequence (TR/TE/FA; 41 milliseconds/11 milliseconds/15 degrees ), radiography (mammography), and histology (iron staining). USPIO accumulation was defined as the presence of 20 iron-positive cells per microscopic view (x100 magnification). Two independent readers analyzed the MR images and rated their confidence level for a positive MRI finding, defined as a focal signal loss, on a 5-point scale. The results were evaluated by receiver-operator characteristic (ROC) analysis. Of a total of 621 vessel segments technically acceptable for evaluation, 534 were histologically negative and 87 positive. Accuracy, expressed as the area under the ROC curve, was 0.85 for reader 1 and 0.88 for reader 2. Interobserver agreement was 0.67. False-positive findings were established by at least one reader for 121 of the 621 segments, false-negative findings for only 15 segments. Calcifications and mural thrombi were identified as potential sources of error of the method. Postmortem USPIO-enhanced MR imaging of atherosclerotic plaques showed a high accuracy and good interobserver agreement in the animal model used here. Further optimization of the method should aim at reducing the rather high percentage of false-positive results.