Masahiro Yanagawa MD, Presenter: Nothing to Disclose

Yuko Tanaka, Abstract Co-Author: Nothing to Disclose

Eiichi Morii, Abstract Co-Author: Nothing to Disclose

Meinoshin Okumura MD, Abstract Co-Author: Nothing to Disclose

Takeshi Johkoh MD, PhD, Abstract Co-Author: Research Consultant, Bayer AG Research Consultant, F. Hoffman-La Roche Ltd

Noriyuki Tomiyama MD, PhD, Abstract Co-Author: Nothing to Disclose

Hiromitsu Sumikawa MD, Abstract Co-Author: Nothing to Disclose

Ayano Kikuyama, Abstract Co-Author: Nothing to Disclose

Osamu Honda MD, PhD, Abstract Co-Author: Nothing to Disclose

Masayoshi Inoue, Abstract Co-Author: Nothing to Disclose

Mitsuhiro Koyama MD, Abstract Co-Author: Nothing to Disclose

Tomoko Gyobu, Abstract Co-Author: Nothing to Disclose

Yutaka Kawata, Abstract Co-Author: Nothing to Disclose

To analyze quantitatively preoperative CT data of pathological stage I pulmonary adenocarcinomas using a custom-developed software and to correlate the results with prognostic factors.

145 adenocarcinomas with pathological stage I were entered into the present study. A custom-developed software can segment solid portion and ground-glass opacity (GGO) using the previously-reported threshold selection method for segmenting gray-level images, and can eliminate vessels on CT using 3-D multi-scale line filter for segmentation and visualization of curvilinear structures such as vessels, resulting in calculating a rate of pure solid portion volume to whole tumor volume (%Solid) regardless of distribution of solid portions in nodule. The following groupings due to %Solid are also possible with this software: pure GGO, dense GGO, part-solid with various distributions of solid portions, and pure solid. Inter-observer agreement between radiologist and software groupings was examined using weighted-Kappa test. Prognostic factors included lymphoduct invasion (LI), vascular invasion (VI), pleural invasion (PI), 5-year overall survival (OS), and 5-year disease-free survival (DFS). Immunostaining methods by D2-40 and CD31 were used for examining LI and VI, respectively. Survival curves were calculated according to the Kaplan-Meier method. Prognostic factors were analyzed by Logistic regression model and Cox proportional hazard model.

There was a good agreement between radiologist and software groupings (weighted kappa=0.760). Multivariate logistic regression analyses revealed that %Solid was significantly useful in estimating LI (p=0.038) and PI (p=0.008). Cox proportional hazard model revealed that %Solid significantly contributed to DFS (p=0.0003). If the cutoff value of %Solid is 63%, there was a significant difference between survival curve of %Solid≥63% (DFS rate, 68.1%) and that of %Solid<63% (DFS rate, 97.6%) (p<0.0001).

Three-dimensional %Solid of adenocarcinomas with pathological stage I is feasible for estimating lymphoduct invasion, pleural invasion, and 5-year disease-free survival. Nodule with %Solid≥63% significantly contributed to short disease-free survival.

Quantitative CT analysis using the software might be able to provide results for estimating prognosis in pulmonary adenocarcinomas with pathological stage I.

Yanagawa, M, Tanaka, Y, Morii, E, Okumura, M, Johkoh, T, Tomiyama, N, Sumikawa, H, Kikuyama, A, Honda, O, Inoue, M, Koyama, M, Gyobu, T, Kawata, Y, Three-dimensional (3D) Quantitative Analysis of Preoperative CT Images in Pathological Stage I Pulmonary Adenocarcinomas: Correlation with Prognostic Factors.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11034213.html