Purpose: To estimate the bias and variance of radiologists measuring the size of spherical and complex synthetic nodules. Methods: This study did not require IRB approval. Six radiologists estimated the size of 10 synthetic nodules embedded within an anthropomorphic thorax phantom from CT scans at 0.8 and 5 mm slice thicknesses. The readers measured the nodule size using three sizing techniques (1D longest in-slice dimension; 2D area from longest in-slice and longest perpendicular dimension; 3D semi-automated volume). Inter-comparisons of bias (difference between average and true size) and variance among methods were performed. Results: The relative biases of radiologists with the 3D tool were -3.6%, -0.5%, -1.6%, 1.9%, -2.6% for 10 mm spherical, 20 mm spherical, 10 mm elliptical, 10 mm lobulated and 10 mm spiculated nodules compared with 1.4%, -0.1%, -26.5%, -7.8%, -39.8% for 1D. 3D was significantly less biased than 1D for elliptical and spiculated nodules. The relative standard deviation for 3D was 21.9%, 11.9%, 10.7%, 29.4%, 24.0% compared with 5.7%, 2.6%, 20.2%, 5.3%, 16.3% for 1D. 1D sizing was significantly less variable than 3D for spherical and lobulated nodules and significantly more variable for the ellipsoid. 2D results were similar to 1D. 3D bias and variability were smaller for thin 0.8 mm slice data compared with thick 5.0 mm data. Conclusion: Radiologists 3D sizing had low bias across all nodule shapes, while 1D and 2D sizing underestimated the true largest diameter and area of complex nodules. Radiologists were generally less variable in their 1D and 2D size measurements.
Citation: Academic Radiology
Pub Type: Journals