Diagnostic accuracy of magnetic resonance imaging for the detection of pulmonary nodules simulated in a dedicated porcine chest phantom

• Published in: PLOS ONE Vol. 15; no. 12; p. e0244382
• Main Authors: Bonert, Madeleine, Schneider, Moritz, Solyanik, Olga, Hellbach, Katharina, Bondesson, David, Gaass, Thomas, Thaens, Natalie, Ricke, Jens, Benkert, Thomas, Dinkel, Julien
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science (PLoS) 23.12.2020; Public Library of Science
• Subjects: Animals; Biology and Life Sciences; Cancer; Diagnosis; Early Detection of Cancer; Evaluation; Humans; Lung Neoplasms; Lung Neoplasms - diagnostic imaging; Lung tumors; Magnetic Resonance Imaging; Magnetic Resonance Imaging - instrumentation; Medicine; Medicine and Health Sciences; Multiple Pulmonary Nodules; Multiple Pulmonary Nodules - diagnostic imaging; Phantoms, Imaging; Q; R; Radiographic Image Interpretation, Computer-Assisted; Radiographic Image Interpretation, Computer-Assisted - methods; Research and Analysis Methods; Science; Sensitivity and Specificity; Swine; Tomography, X-Ray Computed; Tomography, X-Ray Computed - instrumentation; Germany
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0244382

CT serves as gold standard for the evaluation of pulmonary nodules. However, CT exposes patients to ionizing radiation, a concern especially in screening scenarios with repeated examinations. Due to recent technological advances, MRI emerges as a potential alternative for lung imaging using 3D steady state free precession and ultra-short echo-time sequences. Therefore, in this study we assessed the performance of three state-of-the-art MRI sequences for the evaluation of pulmonary nodules. Lesions of variable sizes were simulated in porcine lungs placed in a dedicated chest phantom mimicking a human thorax, followed by CT and MRI examinations. Two blinded readers evaluated the acquired MR-images locating and measuring every suspect lesion. Using the CT-images as reference, logistic regression was performed to investigate the sensitivity of the tested MRI-sequences for the detection of pulmonary nodules. For nodules with a diameter of 6 mm, all three sequences achieved high sensitivity values above 0.91. However, the sensitivity dropped for smaller nodules, yielding an average of 0.83 for lesions with 4 mm in diameter and less than 0.69 for lesions with 2 mm in diameter. The positive predictive values ranged between 0.91 and 0.96, indicating a low amount of false positive findings. Furthermore, the size measurements done on the MR-images were subject to a bias ranging from 0.83 mm to -1.77 mm with standard deviations ranging from 1.40 mm to 2.11 mm. There was no statistically significant difference between the three tested sequences. While showing promising sensitivity values for lesions larger than 4 mm, MRI appears to be not yet suited for lung cancer screening. Nonetheless, the three tested MRI sequences yielded high positive predictive values and accurate size measurements; therefore, MRI could potentially figure as imaging method of the chest in selected follow-up scenarios, e.g. of incidental findings subject to the Fleischner Criteria.