Yulei Qin1, Hao Zheng1, Xiaolin Huang1, Jie Yang1, Yue-Min Zhu2. 1. Institute of Image Processing and Pattern Recognition, Shanghai Jiao Tong University, 800 Dongchuan RD. Minhang District, Shanghai, 200240, China. 2. University Lyon, INSA Lyon, CNRS, INSERM, CREATIS UMR 5220, U1206, F-69621, Lyon, France.
Abstract
PURPOSE: Segmentation of pulmonary nodules is critical for the analysis of nodules and lung cancer diagnosis. We present a novel framework of segmentation for various types of nodules using convolutional neural networks (CNNs). METHODS: The proposed framework is composed of two major parts. The first part is to increase the variety of samples and build a more balanced dataset. A conditional generative adversarial network (cGAN) is employed to produce synthetic CT images. Semantic labels are generated to impart spatial contextual knowledge to the network. Nine attribute scoring labels are combined as well to preserve nodule features. To refine the realism of synthesized samples, reconstruction error loss is introduced into cGAN. The second part is to train a nodule segmentation network on the extended dataset. We build a three-dimensional (3D) CNN model that exploits heterogeneous maps including edge maps and local binary pattern maps. The incorporation of these maps informs the model of texture patterns and boundary information of nodules, which assists high-level feature learning for segmentation. Residual unit, which learns to reduce residual error, is adopted to accelerate training and improve accuracy. RESULTS: Validation on LIDC-IDRI dataset demonstrates that the generated samples are realistic. The mean squared error and average cosine similarity between real and synthesized samples are 1.55 × 10 - 2 and 0.9534, respectively. The Dice coefficient, positive predicted value, sensitivity, and accuracy are, respectively, 0.8483, 0.8895, 0.8511, and 0.9904 for the segmentation results. CONCLUSIONS: The proposed 3D CNN segmentation framework, based on the use of synthesized samples and multiple maps with residual learning, achieves more accurate nodule segmentation compared to existing state-of-the-art methods. The proposed CT image synthesis method can not only output samples close to real images but also allow for stochastic variation in image diversity.
PURPOSE: Segmentation of pulmonary nodules is critical for the analysis of nodules and lung cancer diagnosis. We present a novel framework of segmentation for various types of nodules using convolutional neural networks (CNNs). METHODS: The proposed framework is composed of two major parts. The first part is to increase the variety of samples and build a more balanced dataset. A conditional generative adversarial network (cGAN) is employed to produce synthetic CT images. Semantic labels are generated to impart spatial contextual knowledge to the network. Nine attribute scoring labels are combined as well to preserve nodule features. To refine the realism of synthesized samples, reconstruction error loss is introduced into cGAN. The second part is to train a nodule segmentation network on the extended dataset. We build a three-dimensional (3D) CNN model that exploits heterogeneous maps including edge maps and local binary pattern maps. The incorporation of these maps informs the model of texture patterns and boundary information of nodules, which assists high-level feature learning for segmentation. Residual unit, which learns to reduce residual error, is adopted to accelerate training and improve accuracy. RESULTS: Validation on LIDC-IDRI dataset demonstrates that the generated samples are realistic. The mean squared error and average cosine similarity between real and synthesized samples are 1.55 × 10 - 2 and 0.9534, respectively. The Dice coefficient, positive predicted value, sensitivity, and accuracy are, respectively, 0.8483, 0.8895, 0.8511, and 0.9904 for the segmentation results. CONCLUSIONS: The proposed 3D CNN segmentation framework, based on the use of synthesized samples and multiple maps with residual learning, achieves more accurate nodule segmentation compared to existing state-of-the-art methods. The proposed CT image synthesis method can not only output samples close to real images but also allow for stochastic variation in image diversity.
Authors: Leon Lenchik; Laura Heacock; Ashley A Weaver; Robert D Boutin; Tessa S Cook; Jason Itri; Christopher G Filippi; Rao P Gullapalli; James Lee; Marianna Zagurovskaya; Tara Retson; Kendra Godwin; Joey Nicholson; Ponnada A Narayana Journal: Acad Radiol Date: 2019-08-10 Impact factor: 3.173