Tomoki Uemura1,2, Janne J Näppi1, Yasuji Ryu3, Chinatsu Watari1, Tohru Kamiya2, Hiroyuki Yoshida4. 1. 3D Imaging Research, Department of Radiology, Massachusetts General Hospital and Harvard Medical School, 25 New Chardon Street, Suite 400C, Boston, MA, 02114, USA. 2. Department of Mechanical and Control Engineering, Kyushu Institute of Technology, 1-1 Sensui-cho, Kitakyushu, 804-8550, Japan. 3. Department of Radiology, Tonami General Hospital, 1-61 Shintomi-cho, Tonami, Toyama, 939-1395, Japan. 4. 3D Imaging Research, Department of Radiology, Massachusetts General Hospital and Harvard Medical School, 25 New Chardon Street, Suite 400C, Boston, MA, 02114, USA. yoshida.hiro@mgh.harvard.edu.
Abstract
PURPOSE: Deep learning can be used for improving the performance of computer-aided detection (CADe) in various medical imaging tasks. However, in computed tomographic (CT) colonography, the performance is limited by the relatively small size and the variety of the available training datasets. Our purpose in this study was to develop and evaluate a flow-based generative model for performing 3D data augmentation of colorectal polyps for effective training of deep learning in CADe for CT colonography. METHODS: We developed a 3D-convolutional neural network (3D CNN) based on a flow-based generative model (3D Glow) for generating synthetic volumes of interest (VOIs) that has characteristics similar to those of the VOIs of its training dataset. The 3D Glow was trained to generate synthetic VOIs of polyps by use of our clinical CT colonography case collection. The evaluation was performed by use of a human observer study with three observers and by use of a CADe-based polyp classification study with a 3D DenseNet. RESULTS: The area-under-the-curve values of the receiver operating characteristic analysis of the three observers were not statistically significantly different in distinguishing between real polyps and synthetic polyps. When trained with data augmentation by 3D Glow, the 3D DenseNet yielded a statistically significantly higher polyp classification performance than when it was trained with alternative augmentation methods. CONCLUSION: The 3D Glow-generated synthetic polyps are visually indistinguishable from real colorectal polyps. Their application to data augmentation can substantially improve the performance of 3D CNNs in CADe for CT colonography. Thus, 3D Glow is a promising method for improving the performance of deep learning in CADe for CT colonography.
PURPOSE: Deep learning can be used for improving the performance of computer-aided detection (CADe) in various medical imaging tasks. However, in computed tomographic (CT) colonography, the performance is limited by the relatively small size and the variety of the available training datasets. Our purpose in this study was to develop and evaluate a flow-based generative model for performing 3D data augmentation of colorectal polyps for effective training of deep learning in CADe for CT colonography. METHODS: We developed a 3D-convolutional neural network (3D CNN) based on a flow-based generative model (3D Glow) for generating synthetic volumes of interest (VOIs) that has characteristics similar to those of the VOIs of its training dataset. The 3D Glow was trained to generate synthetic VOIs of polyps by use of our clinical CT colonography case collection. The evaluation was performed by use of a human observer study with three observers and by use of a CADe-based polyp classification study with a 3D DenseNet. RESULTS: The area-under-the-curve values of the receiver operating characteristic analysis of the three observers were not statistically significantly different in distinguishing between real polyps and synthetic polyps. When trained with data augmentation by 3D Glow, the 3D DenseNet yielded a statistically significantly higher polyp classification performance than when it was trained with alternative augmentation methods. CONCLUSION: The 3D Glow-generated synthetic polyps are visually indistinguishable from real colorectal polyps. Their application to data augmentation can substantially improve the performance of 3D CNNs in CADe for CT colonography. Thus, 3D Glow is a promising method for improving the performance of deep learning in CADe for CT colonography.
Entities:
Keywords:
Artificial intelligence; Computer-aided detection; Data augmentation; Deep learning; Generative models; Virtual colonoscopy
Authors: Geert Litjens; Thijs Kooi; Babak Ehteshami Bejnordi; Arnaud Arindra Adiyoso Setio; Francesco Ciompi; Mohsen Ghafoorian; Jeroen A W M van der Laak; Bram van Ginneken; Clara I Sánchez Journal: Med Image Anal Date: 2017-07-26 Impact factor: 8.545
Authors: D C Rockey; E Paulson; D Niedzwiecki; W Davis; H B Bosworth; L Sanders; J Yee; J Henderson; P Hatten; S Burdick; A Sanyal; D T Rubin; M Sterling; G Akerkar; M S Bhutani; K Binmoeller; J Garvie; E J Bini; K McQuaid; W L Foster; W M Thompson; A Dachman; R Halvorsen Journal: Lancet Date: 2005 Jan 22-28 Impact factor: 79.321
Authors: Kirsten Bibbins-Domingo; David C Grossman; Susan J Curry; Karina W Davidson; John W Epling; Francisco A R García; Matthew W Gillman; Diane M Harper; Alex R Kemper; Alex H Krist; Ann E Kurth; C Seth Landefeld; Carol M Mangione; Douglas K Owens; William R Phillips; Maureen G Phipps; Michael P Pignone; Albert L Siu Journal: JAMA Date: 2016-06-21 Impact factor: 56.272
Authors: Daniele Regge; Cristiana Laudi; Giovanni Galatola; Patrizia Della Monica; Luigina Bonelli; Giuseppe Angelelli; Roberto Asnaghi; Brunella Barbaro; Carlo Bartolozzi; Didier Bielen; Luca Boni; Claudia Borghi; Paolo Bruzzi; Maria Carla Cassinis; Massimo Galia; Teresa Maria Gallo; Andrea Grasso; Cesare Hassan; Andrea Laghi; Maria Cristina Martina; Emanuele Neri; Carlo Senore; Giovanni Simonetti; Silvia Venturini; Giovanni Gandini Journal: JAMA Date: 2009-06-17 Impact factor: 56.272