THE INSTITUTE OF ELECTRONICS, INFORMATION AND COMMUNICATION ENGINEERS E-mail: PET/CT, , IEICE Technical Report [email protected] , , Improved scheme of automated detection of pulmonary nodules in PET/CT images using convolutional neural network and conventional characteristic features Atsushi TERAMOTO Hiroshi FUJITA Osamu YAMAMURO and Tsuneo TAMAKI Faculty of Radiological Technology, School of Health Sciences, Fujita Health University 1-98 Dengakugakubo, Kutsukake-cho, Toyoake Aichi, 470-1192 Japan Department of Intelligent Image Information, Graduate School of Medicine, Gifu University 1-1 Yanagido, Gifu, 501-1194 Japan East Nagoya Imaging Diagnosis Center, 3-4-26 Jiyugaoka, Chikusa-ku, Nagoya, Aichi, 464-0044 Japan E-mail: [email protected] Abstract In the existing method for automated detection of pulmonary nodules in PET/CT images, false positives (FPs) are eliminated by the combination of shape feature extraction and classifier. In this study, we develop an improved scheme of automated detection of pulmonary nodules by adding the convolutional neural network (CNN). The proposed scheme detects pulmonary nodules using both CT and PET images. In the CT images, a massive region is first detected using an active contour filter that we developed so far. Subsequently, high-uptake regions detected by the PET images. FP candidates are eliminated using shape features and CNN output, followed by two support vector machines. In the experiments, we evaluated the detection capability using 104 cases of PET/CT images. As a result, the proposed method eliminated approximately half the FPs existing in the previous study. Keywords PET/CT, pulmonary nodule, computer-aided detection, deep learning, convolutional neural network 1. - 17 - This article is a technical report without peer review, and its polished and/or extended version may be published elsewhere. This article is a technical report without peer review, and its polished and/or extended version may be published elsewhere. Copyright©201 ©20 6 by byIEICE IEICE Copyright 2.2. CT CT [6] 2 2.3. 2.4. 2. 2.1. 1 PET/CT CNN - 18 - Sensitivity FPs/case - 19 - [1] American Cancer Society, "Cancer Facts and Figures 2015," 2015. [2] M. Ide and Y. Suzuki, “Is whole-body FDG-PET valuable for health screening?,” Eur. J. Nucl. Med. Mol. Imaging, vol.32, no.3, pp.339-341, 2005. [3] Y. Cui, B. Zhao, TJ. Akhurst, J. Yan, LH. Schwartz, “CT-guided, automated detection of lung tumors on PET images,” Proc. of SPIE Medical Imaging 2008: Computer-Aided Diagnosis, vol.6915, pp.69152N-1 – 69152N-6, 2008. [4] Y. Song, W. Cai, H. Huang, X. Wang, Y. Zhou, M. Fulham, D. Feng, "Lesion Detection and Characterization With Context Driven Approximation in Thoracic FDG PET-CT Images of NSCLC Studies," IEEE Trans. Med Imag. vol.33, no.2, pp.408-421, 2014. [5] A. Teramoto, H. Fujita, K. Takahashi, O. Yamamuro, T. Tamaki, M. Nishio, T. Kobayashi, “Hybrid method for the detection of pulmonary nodules using positron emission tomography / computed tomography: A preliminary study,” Int. J. CARS, vol.9, no.1, pp.59-69, 2014. [6] A. Teramoto, H. Adachi, M. Tsujimoto, H. Fujita, K. Takahashi, O. Yamamuro, T. Tamaki, M. Nishio, “Automated detection of lung tumors in PET/CT images using active contour filter,” Proc. of SPIE Medical Imaging 2015: Computer-Aided Diagnosis, vol.9414, pp.94142V-1 - 94142V-6, 2015. [7] Y. LeCun, Y. Bengio, GE. Hinton, "Deep learning," Nature, vol.521, pp.436–444, 2015. [8] A. Krizhevsky, I. Sutskever, GE. Hinton, “Imagenet classification with deep convolutional neural networks,” Adv. Neur. In., vol.25, pp.1106-1114, 2012. [9] O. Russakovsky, J. Deng, H. Su, J. Krause, S. Satheesh, S. Ma, Z. Huang, A. Karpathy, A. Khosla, M. Bernstein, AC. Berg, L. Fei-Fei, “ImageNet large scale visual recognition challenge,” Int. J. Comp. Vis., vol.115, pp.211-252, 2015. [10] Y. Jia, E. Shelhamer, J. Donahue, S. Karayev, J. Long, R. Girshick, S. Guadarrama, T. Darrell, “Caffe: Convolutional architecture for fast feature embedding,” In ACM Conference on Multimedia, pp.675-678, 2014. - 20 -