Hoen-Oh Shin MD, PRESENTER: Nothing to Disclose

Abstract: HTML Purpose: To develop a software tool for insertion of synthetic lung nodules into CT data for evaluation of different display modalities and computer-aided detection algorithms. Methods and Materials: 40 software generated nodules were inserted at random location and size on multi-detector row CT studies of the chest (scan parameters: 4 x 1-2.5 mm slice collimation, reconstruction with 1-3 mm slice thickness at 1-2 mm interval). Our study population consisted of 20 patients with known lung nodules. On each patient data, 2 synthetic nodules were inserted (size: 3-25 mm). Size, shape, margin, and attenuation could arbitrarily varied and was individually adjusted to match real lesions of each patient. Additionally, noise and a random pattern simulating local density variations was added to synthetic nodules. Three blinded readers evaluated 40 real and 40 simulated pulmonary nodules according to a 5-point confidence scale (1= synthetic, 2= probably synthetic, 3= indefinite, 4= probably real, 5= real). The readers were allowed to use cine view, multi-planar reformation, and volume rendering. Additionally, a zooming function and interactive change of window settings were permitted. A multivariate analysis of covariance using the 3 readers as dependent variables and the size of the simulated lesions as covariable was performed for statistical assessment (SPSS 11.5, Chicago, Il). Results: Real and simulated lesions were indistinguishable for all 3 readers (Pillai's Trace statistic: P= .881). However, nodule size was a statistically significant covariable regarding differentiation of synthetic against real nodule. Larger simulated nodules were easier to detect than smaller ones (Pillai's Trace statistic: P < .05). Conclusion: The developed algorithm allowed for synthetic generation of ground truth experimental data for evaluation of different display modalities and automatic detection algorithms. However, with increasing size, simulated nodules were easier to detect as the surrounding lung parenchyma was not included in the synthesis (e.g. displacement of vessels).       Questions about this event email: shin.hoen-oh@mh-hannover.de

Shin MD, H, Insertion of Synthetic Pulmonary Nodules in CT Data of the Chest: Development of a Software Tool.  Radiological Society of North America 2003 Scientific Assembly and Annual Meeting, November 30 - December 5, 2003 ,Chicago IL. http://archive.rsna.org/2003/3100496.html