Koji Sakai, Presenter: Nothing to Disclose

Hiroyuki Sekiguchi, Abstract Co-Author: Nothing to Disclose

Masahiro Yakami MD, Abstract Co-Author: Nothing to Disclose

Koji Fujimoto MD, Abstract Co-Author: Nothing to Disclose

Naozo Sugimoto, Abstract Co-Author: Nothing to Disclose

Kaori Togashi MD, PhD, Abstract Co-Author: Nothing to Disclose

Yutaka Emoto MD, PhD, Abstract Co-Author: Nothing to Disclose

Takeshi Kubo MD, Abstract Co-Author: Nothing to Disclose

Ryo Sakamoto, Abstract Co-Author: Nothing to Disclose

Contour features computed by a method based on a normalized 2D elliptic Fourier descriptor showed linear correlations with subjective features reported by radiologists. Thus, these computed contour features would be feasible for use as supporting information in image-based differential diagnoses.

Information on contour features of solitary pulmonary nodules is important in image diagnosis and has been widely used in differential diagnoses. The present study attempted to translate the contours of solitary pulmonary nodules to a numerical scale using a method based on a normalized 2D elliptic Fourier descriptor.

The nodules included in this study were clinically confirmed as primary lung cancers, metastatic pulmonary nodules, or other benign nodules (diameter: 5-30 mm) which were obtained from 40 CT examinations (slice thickness: 0.5 mm, convolution kernel; Toshiba FC51). Two board-certificated radiologists graded the shape features for every nodule (n = 154). The conspicuity was rated on a 5-point scale (1 to 5).Two contour features (mean protrusion and mean intrusion) of the 2D solitary pulmonary nodules based on normalized elliptic Fourier descriptor were calculated and were compared with subjective features reported by radiologists. The two contour features were calculated as follows: mean protrusion=Σ(dn80Pi-dn1i)/nP, where nP is the number of points on both the 1st and 80th harmonics and dn80Pi is the distance from the origin (dn80Pi > dn1i); and mean intrusion=Σ(dn80Ii-dn1i)/nI, where nI is the number of points on both the 1st and 80th harmonics and dn80Ii is the distance from the origin (dn80Ii < dn1i). Linear correlations between the subjective ratings of contour features and the computed contour features were investigated, and correlation coefficients (R2) were calculated.

The smooth, ragged, coarse spiculation, notch, concave and convex contours were linearly correlated (R2 > 0.73) with the mean intrusion. The concave contour cases had the best correlation with the computed features (R2 = 0.83, 0.88). The corona radiata had a low R2 with the computed features (R2 = 0.30, 0.53).

Sakai, K, Sekiguchi, H, Yakami, M, Fujimoto, K, Sugimoto, N, Togashi, K, Emoto, Y, Kubo, T, Sakamoto, R, Computed Contour Features of Solitary Pulmonary Nodules Match with Radiologists' Observation: Feasibility Study of a Normalized 2D Elliptic Fourier Descriptor-based Method.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11013312.html