Kazuo Awai MD, Abstract Co-Author: Nothing to Disclose

Kohei Murao PhD, Presenter: Nothing to Disclose

Akio Ozawa BS, Abstract Co-Author: Nothing to Disclose

Yoshiharu Nakayama MD, Abstract Co-Author: Nothing to Disclose

Yoshinori Funama PhD, Abstract Co-Author: Nothing to Disclose

Yasuyuki Yamashita MD, Abstract Co-Author: Nothing to Disclose

Natsuki Maruyama MD, Abstract Co-Author: Nothing to Disclose

et al, Abstract Co-Author: Nothing to Disclose

To evaluate the effect of computer-aided diagnosis (CAD) on the performance of radiologists estimating the malignancy of pulmonary nodules at thin-slice helical CT.

We developed a CAD system which could quantify three-dimensional morphological features of pulmonary nodules using thin slice helical CT data and estimate their malignancy. Our CAD system showed radiologists various quantified index of pulmonary nodules such as maximum diameter, averaged CT number, volume, margin irregularity, internal homogeneity, averaged density gradient of the margin, sphericity, volume ratio of internal calcification, and volume ratio of cavity in the nodule. Our CAD system also demonstrated estimated values of likelihood of malignancy (0: absolutely benign, 1: absolutely malignant). We evaluated 33 pulmonary nodules less than 3.0 cm in maximum diameter (18 malignant and 15 benign nodules). Receiver operating characteristic (ROC) analysis with a continuous rating scale was used to compare the observers’ performance in estimating the likelihood of malignancy of pulmonary nodules with and without CAD. The participants were 10 board-certified radiologists and 9 radiology residents. After rating the probability of malignancy of pulmonary nodules without CAD they repeated their assessment using CAD.

For all　participants, the mean area under the best-fit ROC curves (Az) without and with CAD was 0.85 ± 0.08 and 0.91 ± 0.05, respectively. The difference was significant at p<0.01. The mean Az values without and with CAD were 0.91 ± 0.05 and 0.94± 0.04, respectively (p<0.01) for the 10 board-certified radiologists and 0.79 ± 0.06 and 0.88 ± 0.05, respectively (p<0.01), for the 9 radiology residents.

Use of CAD improved the performance of all 19 observers assessing the malignancy of pulmonary nodules.

A.O.,K.M.: Kohei Murao and Akio Ozawa are employee of Fujitsu LTD.

Awai, K, Murao, K, Ozawa, A, Nakayama, Y, Funama, Y, Yamashita, Y, Maruyama, N, et al, , Estimation of Malignancy of Pulmonary Nodules at Thin-slice Helical CT: The Performance of Radiologists Is Improved by Computer-aided Diagnosis.  Radiological Society of North America 2004 Scientific Assembly and Annual Meeting, November 28 - December 3, 2004 ,Chicago IL. http://archive.rsna.org/2004/4407860.html