19 March 2020 in IEEE Transactions on Neural Networks and Learning Systems
Among the various generative adversarial network (GAN)-based image inpainting methods, a coarse-to-fine network with a contextual attention module (CAM) has shown remarkable performance. However, due to two stacked generative networks, the coarse-to-fine network needs numerous computational resources, such as convolution operations and network parameters, which result in low speed. To address this problem, we propose a novel network architecture called parallel extended-decoder path for semantic inpainting (PEPSI) network, which aims at reducing the hardware costs and improving the inpainting performance. PEPSI consists of a single shared encoding network and parallel decoding networks called coarse and inpainting paths. The coarse path produces a preliminary inpainting result to train the encoding network for the prediction of features for the CAM. Simultaneously, the inpainting path generates higher inpainting quality using the refined features reconstructed via the CAM. In addition, we propose Diet-PEPSI that significantly reduces the network parameters while maintaining the performance. In Diet-PEPSI, to capture the global contextual information with low hardware costs, we propose novel rate-adaptive dilated convolutional layers that employ the common weights but produce dynamic features depending on the given dilation rates. Extensive experiments comparing the performance with state-of-the-art image inpainting methods demonstrate that both PEPSI and Diet-PEPSI improve the qualitative scores, i.e., the peak signal-to-noise ratio (PSNR) and structural similarity (SSIM), as well as significantly reduce hardware costs, such as computational time and the number of network parameters.
18 June 2021 in npj Precision Oncology
In 2020, it is estimated that 73,750 kidney cancer cases were diagnosed, and 14,830 people died from cancer in the United States. Preoperative multi-phase abdominal computed tomography (CT) is often used for detecting lesions and classifying histologic subtypes of renal tumor to avoid unnecessary biopsy or surgery. However, there exists inter-observer variability due to subtle differences in the imaging features of tumor subtypes, which makes decisions on treatment challenging. While deep learning has been recently applied to the automated diagnosis of renal tumor, classification of a wide range of subtype classes has not been sufficiently studied yet. In this paper, we propose an end-to-end deep learning model for the differential diagnosis of five major histologic subtypes of renal tumors including both benign and malignant tumors on multi-phase CT. Our model is a unified framework to simultaneously identify lesions and classify subtypes for the diagnosis without manual intervention. We trained and tested the model using CT data from 308 patients who underwent nephrectomy for renal tumors. The model achieved an area under the curve (AUC) of 0.889, and outperformed radiologists for most subtypes. We further validated the model on an independent dataset of 184 patients from The Cancer Imaging Archive (TCIA). The AUC for this dataset was 0.855, and the model performed comparably to the radiologists. These results indicate that our model can achieve similar or better diagnostic performance than radiologists in differentiating a wide range of renal tumors on multi-phase CT.
1 January 2021 in IEEE Transactions on Neural Networks and Learning Systems
In adversarial learning, the discriminator often fails to guide the generator successfully since it distinguishes between real and generated images using silly or nonrobust features. To alleviate this problem, this brief presents a simple but effective way that improves the performance of the generative adversarial network (GAN) without imposing the training overhead or modifying the network architectures of existing methods. The proposed method employs a novel cascading rejection (CR) module for discriminator, which extracts multiple nonoverlapped features in an iterative manner using the vector rejection operation. Since the extracted diverse features prevent the discriminator from concentrating on nonmeaningful features, the discriminator can guide the generator effectively to produce images that are more similar to the real images. In addition, since the proposed CR module requires only a few simple vector operations, it can be readily applied to existing frameworks with marginal training overheads. Quantitative evaluations on various data sets, including CIFAR-10, CelebA, CelebA-HQ, LSUN, and tiny-ImageNet, confirm that the proposed method significantly improves the performance of GAN and conditional GAN in terms of the Frechet inception distance (FID), indicating the diversity and visual appearance of the generated images.
23 May 2008 in IEEE Transactions on Broadcasting
Boundary matching algorithm (BMA) and decoder motion vector estimation (DMVE) are two well-known temporal error concealment methods using the matching-based approach. In these two methods, the motion vector of each missing block is estimated by choosing one among candidate motion vectors which minimizes a sum of absolute differences (SAD) between boundary pixels of the corrupted macroblock. In general, the performance of DMVE is better than that of BMA. However, depending on the location or pattern of the corrupted block, BMA produces higher visual quality than DMVE. In this paper, we propose two types of hybrid error concealment methods; switching method and blending method. The switching method chooses one of two results obtained by BMA and DMVE based on the normalized SAD values. In the blending method, the weighted sum of the results concealed by the aforementioned two methods is utilized to improve the performance of error concealment. In order to reduce blocking artifacts further, the modified overlapped-block motion compensation is adaptively applied to the concealed blocks. Simulation results show that the proposed methods outperform other techniques in terms of subjective visual quality as well as PSNR performance.
3 November 2022 in IEEE Journal of Biomedical and Health Informatics
Multi-phase CT is widely adopted for the di- agnosis of kidney cancer due to the complementary in- formation among phases. However, the complete set of multi-phase CT is often not available in practical clinical applications. In recent years, there have been some studies to generate the missing modality image from the available data. Nevertheless, the generated images are not guaran- teed to be effective for the diagnosis task. In this paper, we propose a unified framework for kidney cancer diagnosis with incomplete multi-phase CT, which simultaneously re- covers missing CT images and classifies cancer subtypes using the completed set of images. The advantage of our framework is that it encourages a synthesis model to ex- plicitly learn to generate missing CT phases that are helpful for classifying cancer subtypes. We further incorporate lesion segmentation network into our framework to exploit lesion-level features for effective cancer classification in the whole CT volumes. The proposed framework is based on fully 3D convolutional neural networks to jointly optimize both synthesis and classification of 3D CT volumes. Exten- sive experiments on both in-house and external datasets demonstrate the effectiveness of our framework for the diagnosis with incomplete data compared with state-of-the- art baselines. In particular, cancer subtype classification using the completed CT data by our method achieves higher performance than the classification using the given incomplete data.