[Detectability of lesions in multiplanar reformation (MPR) of the lung].

Abstract

Purpose: To evaluate the diagnostic ability of multiplanar reformation (MPR) images of the lung in comparison with thin-section source CT images.

Materials and methods: MPR images were reconstructed for 79 patients with suspected pulmonary disease. Slice thicknesses of source images were 2 mm in 24, 1 mm in 30, and 0.5 mm in 25 cases. The presence of centrilobular nodules, emphysema, bronchiectasis, ground-glass opacity (GGO), consolidation, interstitial thickening, and pulmonary nodule was evaluated on thin-section source images by an experienced chest radiologist to establish gold standards and then subsequently assessed on the MPR images independently by two radiologists. The sensitivity, specificity, and accuracy of each finding were calculated regarding the results of thin-section source images as the gold standards. Accuracy for the detection of findings was also statistically compared among the three groups of different source slice thicknesses using Fisher's exact test.

Results: Accuracy for the detection of findings was significantly less (p < 0.05) in 2 mm slice MPR for centrilobular nodule, GGO, and interstitial thickening than in 1 mm or 0.5 mm slice MPR. No statistically significant difference was observed for any of the findings between 0.5 mm and 1 mm slice MPR. Rates of sensitivity, specificity, and accuracy of the MPR images for detection of the findings were 89-100%, 73-95%, and 84-95%, respectively.

Conclusion: In comparison with thin-section source images, MPR images are comparably sensitive but not as specific for the detection of findings. When producing MPR images, the slice thickness of source images should be less than 2 mm.