Marios A. Gavrielides PhD, Presenter: Nothing to Disclose

Rongping Zeng PhD, Abstract Co-Author: Nothing to Disclose

Berkman Sahiner PhD, Abstract Co-Author: Nothing to Disclose

Kyle Jean Myers PhD, Abstract Co-Author: Nothing to Disclose

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

To determine the effect of overlapping reconstruction on lung nodule volume estimation error using realistic phantom data. To the best of our knowledge this effect has not been previously examined for this task. 

We scanned an anthropomorphic phantom containing 6 spherical synthetic nodules attached to vasculature (+100, -630 HU, 5, 10, 20 mm in diameter), with a 64-slice scanner using 3 exposures (20, 100, 200 mAs), 2 pitches (0.9, 1.2), 3 slice thicknesses (0.75, 1.5, 3.0 mm), and a standard thorax reconstruction kernel (B40f). Ten repeat scans were acquired for each condition. Volumes were reconstructed in both the contiguous (no overlap, slice interval equal to slice thickness) and overlapping (50% overlap, slice interval equal to half slice thickness) modes. Volume estimation was performed using a matched filter-based method. Percentage bias, defined as the percentage difference between measured and true volume normalized by true volume, were calculated for all nodules under each imaging condition, and compared between contiguous and overlapping reconstructions. In addition, the effect of overlapping reconstruction was analyzed for three sub-groups with constant nodule size to slice thickness ratios (NSSTR), including: 1) NSSTR=3.3, consisting of 5mm nodules reconstructed at 1.5 mm slice thickness (STh) and 10mm nodules at 3mm STh, 2) NSSTR=6.7, consisting of 5mm nodules at 0.75 mm STh, 10mm nodules at 1.5 mm STh, and 20mm nodules at 3mm STh, and 3) NSSTR=13.3, consisting of 10mm nodules at 0.75mm STh and 20mm nodules at 1.5mm STh. Statistical differences were analyzed using t-test for paired data.

Using pair-wise analysis on all volume measures, bias was 1.8% for overlapping and 8.0% for non-overlapping reconstruction (p<0.001). Bias for overlapping and non-overlapping reconstruction was -1.2% and 7.3% respectively for NSSTR=3.3 (p<0.001), 0.5% and -11.5% for NSSTR=6.7 (p<0.001), and 2.0% and 4.2% for NSSTR=13.3 (p=0.01).

Our results indicate that reconstruction with 50% overlap can significantly reduce the magnitude of bias in volumetric estimation. Results show a trend for this effect to be less significant as the nodule size to slice thickness ratio is increased.

Our experiments support that CT reconstruction with 50% overlap reduces the bias in volumetric estimation of lung nodule size. This result can impact the choice of protocols for sizing lung nodules.

Gavrielides, M, Zeng, R, Sahiner, B, Myers, K, Petrick, N, Effect of Overlapping Reconstruction on Lung Nodule Volume Estimation with Multidetector CT.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11034303.html