Michael F. McNitt-Gray PhD, Presenter: Institutional research agreement, Siemens AG Research grant, Siemens AG Instructor, Medical Technology Management Institute

Hyun Jung Kim PhD, Abstract Co-Author: Nothing to Disclose

Binsheng Zhao DSc, Abstract Co-Author: Nothing to Disclose

Lawrence H. Schwartz MD, Abstract Co-Author: Research, General Electric Company Research Consultant, Novartis AG

David A. Clunie MBBS, Abstract Co-Author: Officer, CoreLab Partners, Inc Owner, PixelMed Publishing LLC Research support, Siemens AG

Kristin Borradaile MS, Abstract Co-Author: Nothing to Disclose

Kevin J. Byrne MD, Abstract Co-Author: Employee, CoreLab Partners, Inc

S Kaplan, Abstract Co-Author: Employee, CoreLab Partners, Inc

Julie Lewin Barudin MD, Abstract Co-Author: Employee, CoreLab Partners, Inc

J Sherman, Abstract Co-Author: Employee, CoreLab Partners, Inc

K Slazak, Abstract Co-Author: Employee, CoreLab Partners, Inc

Nicholas Petrick PhD, Abstract Co-Author: Nothing to Disclose

Charles Fenimore PhD, Abstract Co-Author: Nothing to Disclose

John Lu PhD, Abstract Co-Author: Nothing to Disclose

Andrew J. Buckler MS, Abstract Co-Author: Stockholder, vascuVis Inc President, vascuVis Inc CEO, vascuVis Inc

To determine the minimum detectable change in size of lung lesions by investigating 1D, 2D, and 3D measurement variability when patients are repeatedly imaged with CT under a “no change” (“coffee break”) condition.

The CT datasets used were those of 32 non-small cell lung cancer patients scanned twice within 15 minutes and reconstructed as thin transverse slices (publicly available from NBIA in the RIDER collection, contributed by MSKCC). One lesion was identified for each patient (32 target lesions). Five experienced radiologists (KB, KS, JB, JS, SK) were shown each lesion in multiple sessions and performed separate measurements using three techniques (1D longest in-slice dimension; 2D product of the longest and perpendicular in-slice dimensions; 3D semi-automated segmented volume). Readers, scans and measurement techniques were randomized to reduce bias. Change in size was estimated by comparing measurements performed on both scans for the same lesion, for each reader and each measurement method. Results were pooled across lesions, across readers and across both readers and lesions, for each measurement method.

Mean percent difference (± standard deviation) when pooled across both readers and lesions for 1-D, 2-D and 3-D measurements was 5.8% ± 23.8, 15.2% ± 68.5 and 25.0% ± 117.9, respectively. The range of mean percent differences (across lesions) between readers was 0.19% to 17.0% for the 1-D measurements; 2.9% to 77.8% for 3-D measurements. The range of mean percent differences (across readers) between lesions ranged from -15.6% to 57.6% for 1-D measurements; -35.1% to 299.6% for 3-D measurements.

Even under a “no change” condition between scans, there is variation in lesion size measurements due to repeat scans and variations in reader, lesion and measurement method. Understanding this variability will lead to improved interpretation of lesion size when assessing response to treatment. Preliminary examination of outliers reveals that some lesions are more suitable than others for volume measurement reinforcing the importance of appropriate target lesion selection.

Characterizing performance, such as minimal detectable change, is a prerequisite for biomarker qualification and for establishing limitations when using measurements to determine therapeutic response.

McNitt-Gray, M, Kim, H, Zhao, B, Schwartz, L, Clunie, D, Borradaile, K, Byrne, K, Kaplan, S, Barudin, J, Sherman, J, Slazak, K, Petrick, N, Fenimore, C, Lu, J, Buckler, A, Estimating the Minimum Detectable Change of Lung Lesions Using Patient Datasets Acquired under a “No Change” Condition.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11014014.html