Comparison of 2D and 3D Views for Evaluation of Flat Lesions in CT Colonography

Abstract

Flat lesions in the colon may result in false-negative computed tomography colonography interpretations. It is unknown whether flat lesions are better measured on two-dimensional (2D) or three-dimensional (3D) images and which settings are optimal for enhanced reproducibility and decreased variability. We evaluated these factors to determine whether 2D or 3D is best for flat lesion measurements. Eighty-eight lesions in 66 patients from a previously published clinical trial were analyzed. Lesions were viewed with four methods including 2D at three window/level settings and 3D endoluminal view. Lesions in either supine or prone were counted as one dataset. Long axis and height were measured. Criteria of "height" (<or=3 mm high) or "ratio" (height <or=half the long axis) were applied. A subset of lesions was subject to inter- and intra-observer variability analysis. With the "height" criterion, more datasets were classified as flat in 2D flat (n = 76), 2D soft tissue (n = 82), and 3D (n = 73) views than in the 2D lung (n = 49) view. If long axis is used as the key metric, endoluminal 3D (12.1%) views significantly showed the least inter-observer variability compared to lung (18.9%) or soft tissue (20.2%) views. Intra-observer variability was low overall for all methods. When characterizing lesions as flat, a consistent viewing method should be used. To minimize inter-observer variability (such as when following a patient over time), it is best to use the ratio criterion for flat lesion definition incorporating the single longest dimension on 3D views as the key metric.