Automated Matching of Pulmonary Nodules: Evaluation in Serial Screening Chest CT

Abstract

The purpose of our study was to evaluate the performance of a computer-aided program that performs automated matching of pulmonary nodules imaged in three serial screening chest MDCT studies. Forty subjects with pulmonary nodules depicted in three annual (T0, T1, T2) low-dose MDCT screening studies for lung cancer were selected from the National Lung Screening Trial database at a single institution. All CT images were reevaluated by two radiologists in consensus. One hundred forty-three nodules were identified and characterized by type (solid parenchymal, juxtavascular, juxtapleural, and ground-glass opacity) and size (< or = 4 mm, 4-6 mm, 6-8 mm, and > 8 mm). Using an automated program, nodules at T0 were matched to nodules at T1, and the same nodules at T1 were matched to nodules at T2. Associations between nodule matching rate (i.e., number of nodules matched by the program divided by the number of nodules determined to match by radiologists) and nodule type or size were analyzed. The combined matching rate of the nodules was 92.7% (T0 vs T1, 91.6%; T1 vs T2, 93.7%). By nodule type, the matching rate was 94.6% (parenchymal), 98.4% (juxtavascular), 85.8% (juxtapleural), and 100% (ground-glass opacity), with the rate significantly lower for juxtapleural nodules (p < 0.01). Matching rates were not significantly influenced by nodule size (p = 0.67). The automated matching rate for pulmonary nodules in screening MDCT scans was high (92.7%) and was not affected by the nodule size but was slightly lower with nodules at juxtapleural locations.