Simulated pulmonary nodules implanted in a dedicated porcine chest phantom: sensitivity of MR imaging for detection.

Abstract

To evaluate the diagnostic accuracy of common magnetic resonance (MR) imaging sequences for detection of small pulmonary nodules by using a chest phantom and porcine lungs containing simulated lesions. Fourteen porcine lungs containing 366 porcine myocardial tissue implants were inflated inside a phantom. Two-dimensional (2D) and three-dimensional (3D) gradient-echo (GRE), T2-weighted turbo spin-echo (SE), and T2-weighted single-shot SE train MR sequences were performed. Spiral computed tomography (CT) was performed for comparison. Blinded observers read the images and recorded the sizes and locations of visible nodules by consensus. The sensitivity of each imaging method for depicting single nodules of given sizes was calculated. Specificities, positive predictive values (PPVs), and negative predictive values (NPVs) for detection of one or more nodules of various sizes were calculated. Sensitivities of 3D GRE, 2D GRE, T2-weighted turbo SE, and T2-weighted single-shot SE train MR imaging and of CT were 0.50, 0.40, 0.12, 0.00, and 0.55, respectively, for detection of 1.4-mm nodules and 0.88, 0.84, 0.69, 0.06, and 0.96, respectively, for detection of 4.2-mm nodules. The 95% CIs for CT and GRE MR imaging overlapped, but those for turbo SE and single-shot SE train MR imaging differed significantly (P <.05). For detection of nodules larger than 5 mm, all examinations except single-shot SE train MR imaging yielded a specificity, PPV, and NPV of 1.00 each. For detection of nodules smaller than 5 mm, diagnostic accuracy of 3D GRE MR imaging was high: Specificity, PPV, and NPV all were approximately 0.90. Two-dimensional GRE MR imaging results were influenced by false-positive findings: Specificity was 0.64; PPV, 0.74; and NPV, 1.00. Common MR imaging sequences such as 3D GRE have high diagnostic accuracy in depicting small pulmonary nodules when artifacts from cardiac and respiratory motion are absent.