Kelly Piacsek PhD, Abstract Co-Author: Nothing to Disclose

Saad Sirohey PhD, Abstract Co-Author: Nothing to Disclose

Paul Edgar Licato MS, Presenter: Nothing to Disclose

Volumetric quantification of sub-solid pulmonary nodules is of significant clinical interest since recent evidence suggests that these nodules are more likely to represent malignant disease than solid nodules. We have developed a segmentation algorithm that delineates multiple sub-components of lung nodules. We have also developed a phantom that mimics the attenuation and geometry of sub-solid nodules within normal lung parenchyma. The goal of this study was to evaluate the performance of this approach on ground truth data obtained from the phantom.

Non-solid nodules were constructed from a hardened mixture of polyurethane foam and epoxy. Part solid nodules were constructed from acrylic spheres (2.4-9.5mm diameter), encased in the non-solid mixture. Nodules were embedded in synthetic foam that was inserted into a resin shell approximating attenuation of the chest cavity. The phantom was scanned using a GE Lightspeed 16-slice CT scanner (0.625mm slice thickness, 400mA, 120kV and standard kernel). Reference volume of the simulated nodules was approximated by known density and measured weight of individual nodules. Tolerance was 3.06% and 1.41% for the solid and non-solid nodule volume components, respectively. Voxels within a user-selected volume of interest are classified by identifying optimal boundaries between tissue classes (i.e. parenchyma, non-solid tissue and solid tissue) and are delineated based on density values as well as local texture criteria. Finally, total tumor volume within a 3D image volume is estimated from the sum of solid and non-solid components for each nodule.

Total nodule volume within the phantom was 4680mm3 and was underestimated by 1.9%. The total non-solid component of nodule volume (3794mm3) was underestimated by 0.2%. We validated this approach qualitatively on 12 radiologist-determined sub-solid nodules and confirmed that our automated segmentation met with radiologist agreement.

We have demonstrated a quantitative technique for tracking total volume of solid and non-solid nodule components. We validated this technique on a unique phantom consisting of materials that mimic CT attenuation and geometry of known lung nodules.

K.P.,S.S.,P.E.L.: Authors are employees of General Electric Healthcare

Piacsek, K, Sirohey, S, Licato, P, Accurate Volumetric Quantification of Sub-Solid Lung Nodules from 3D Computed Tomography.  Radiological Society of North America 2005 Scientific Assembly and Annual Meeting, November 27 - December 2, 2005 ,Chicago IL. http://archive.rsna.org/2005/4418666.html