Publication Information

Basic Information

Access and Citation

AI Researcher Chatbot

Get quick answers to your questions about the article from our AI researcher chatbot

Primary Location

Authors

Topics

Keywords

Related Works

Title: $MP20-10 DEEP LEARNING WITH A CONVOLUTIONAL NEURAL NETWORK ALGORITHM FOR FULLY AUTOMATED DETECTION OF PROSTATE CANCER USING PRE-BIOPSY MRI
Abstract: You have accessJournal of UrologyImaging/Radiology: Uroradiology II1 Apr 2018MP20-10 DEEP LEARNING WITH A CONVOLUTIONAL NEURAL NETWORK ALGORITHM FOR FULLY AUTOMATED DETECTION OF PROSTATE CANCER USING PRE-BIOPSY MRI Junichiro Ishioka, Yoh Matsuoka, Sho Uehara, Yosuke Yasuda, Toshiki Kijima, Soichiro Yoshida, Minato Yokoyama, Kazutaka Saito, Tomo Kimura, Kosei Kudo, Itsuo Kumazawa, and Yasuhisa Fujii Junichiro IshiokaJunichiro Ishioka , Yoh MatsuokaYoh Matsuoka , Sho UeharaSho Uehara , Yosuke YasudaYosuke Yasuda , Toshiki KijimaToshiki Kijima , Soichiro YoshidaSoichiro Yoshida , Minato YokoyamaMinato Yokoyama , Kazutaka SaitoKazutaka Saito , Tomo KimuraTomo Kimura , Kosei KudoKosei Kudo , Itsuo KumazawaItsuo Kumazawa , and Yasuhisa FujiiYasuhisa Fujii View All Author Informationhttps://doi.org/10.1016/j.juro.2018.02.680AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookTwitterLinked InEmail INTRODUCTION AND OBJECTIVES Magnetic resonance imaging (MRI) provides a noninvasive assessment of the prostate that improves the detection of prostate cancer and can reduce unnecessary biopsies. The excessive variation in the performance and interpretation of MRI is, however, a major barrier to its widespread acceptance and use. In this study, we developed a computer-aided diagnostic system with a convolutional neural network algorithm for fully automated detection of prostate cancer using pre-biopsy MRI. METHODS We selected 187 T2WMRI-positive patients who were diagnosed with prostate cancer by MRI-targeted biopsy and 129 T2WMRI-negative patients who were not diagnosed with prostate cancer by systematic prostate biopsy. We used 1334 images from 165 patients diagnosed as prostate cancer and 3155 images from 107 patients diagnosed as non-cancer for deep learning. Model evaluation and validation were performed by calculating area under the curve (AUC) using 2 data sets consisting of 11 cancer images and 6 non-cancer images excluded from the training samples. To generate the input image, we sampled 20 million parallelogram images at the central region including the prostate, and performed an affine transformation to create a square of 261 × 261 pixels. Different loss functions were used in the cancer and non-cancer regions. Deep transfer learning for parameter optimization was performed using a convolutional neural network (U-net combined with ResNet50) (Figure). RESULTS Median PSA levels of cancer and non-cancer patients were 8.40 ng/ml and 6.43 ng/ml, respectively. The Gleason score of the targeted biopsy positive site was 3+3: 14, 3+4: 76, 4+3: 41, 4+4: 47, 9≤: 9, respectively. The time for deep learning was 5.5 hours, and the required time for detecting the cancer region using the pre-trained model was 30 milliseconds per image (GPU: GeForceGTX 1080®). As the learning of the image progressed, the AUC increased and finally reached 0.793 and 0.636 when the learning of 20 million images was finished. CONCLUSIONS Computer-aided diagnosis with a convolutional neural network algorithm for fully automated detection of prostate cancer regions in the pre-biopsy MRI image has the potential to provide reproducible interpretation and a greater level of standardization and consistency. © 2018FiguresReferencesRelatedDetails Volume 199Issue 4SApril 2018Page: e256 Advertisement Copyright & Permissions© 2018MetricsAuthor Information Junichiro Ishioka More articles by this author Yoh Matsuoka More articles by this author Sho Uehara More articles by this author Yosuke Yasuda More articles by this author Toshiki Kijima More articles by this author Soichiro Yoshida More articles by this author Minato Yokoyama More articles by this author Kazutaka Saito More articles by this author Tomo Kimura More articles by this author Kosei Kudo More articles by this author Itsuo Kumazawa More articles by this author Yasuhisa Fujii More articles by this author Expand All Advertisement Advertisement PDF downloadLoading ...