Usformer: A small network for left atrium segmentation of 3D LGE MRI.

Left atrial (LA) fibrosis plays a vital role as a mediator in the progression of atrial fibrillation. 3D late gadolinium-enhancement (LGE) MRI has been proven effective in identifying LA fibrosis. Image analysis of 3D LA LGE involves manual segmentation of the LA wall, which is both lengthy and challenging. Automated segmentation poses challenges owing to the diverse intensities in data from various vendors, the limited contrast between LA and surrounding tissues, and the intricate anatomical structures of the LA. Current approaches relying on 3D networks are computationally intensive since 3D LGE MRIs and the networks are large. Regarding this issue, most researchers came up with two-stage methods: initially identifying the LA center using a scaled-down version of the MRIs and subsequently cropping the full-resolution MRIs around the LA center for final segmentation. We propose a lightweight transformer-based 3D architecture, Usformer, designed to precisely segment LA volume in a single stage, eliminating error propagation associated with suboptimal two-stage training. The transposed attention facilitates capturing the global context in large 3D volumes without significant computation requirements. Usformer outperforms the state-of-the-art supervised learning methods in terms of accuracy and speed. First, with the smallest Hausdorff Distance (HD) and Average Symmetric Surface Distance (ASSD), it achieved a dice score of 93.1% and 92.0% in the 2018 Atrial Segmentation Challenge and our local institutional dataset, respectively. Second, the number of parameters and computation complexity are largely reduced by 2.8x and 3.8x, respectively. Moreover, Usformer does not require a large dataset. When only 16 labeled MRI scans are used for training, Usformer achieves a 92.1% dice score in the challenge dataset. The proposed Usformer delineates the boundaries of the LA wall relatively accurately, which may assist in the clinical translation of LA LGE for planning catheter ablation of atrial fibrillation.

Fast and Robust Cascade Model for Multiple Degradation Single Image Super-Resolution.

Single Image Super-Resolution (SISR) is one of the low-level computer vision problems that has received increased attention in the last few years. Current approaches are primarily based on harnessing the power of deep learning models and optimization techniques to reverse the degradation model. Owing to its hardness, isotropic blurring or Gaussians with small anisotropic deformations have been mainly considered. Here, we widen this scenario by including large non-Gaussian blurs that arise in real camera movements. Our approach leverages the degradation model and proposes a new formulation of the Convolutional Neural Network (CNN) cascade model, where each network sub-module is constrained to solve a specific degradation: deblurring or upsampling. A new densely connected CNN-architecture is proposed where the output of each sub-module is restricted using some external knowledge to focus it on its specific task. As far we know, this use of domain-knowledge to module-level is a novelty in SISR. To fit the finest model, a final sub-module takes care of the residual errors propagated by the previous sub-modules. We check our model with three state-of-the-art (SOTA) datasets in SISR and compare the results with the SOTA models. The results show that our model is the only one able to manage our wider set of deformations. Furthermore, our model overcomes all current SOTA methods for a standard set of deformations. In terms of computational load, our model also improves on the two closest competitors in terms of efficiency. Although the approach is non-blind and requires an estimation of the blur kernel, it shows robustness to blur kernel estimation errors, making it a good alternative to blind models.

Generative Adversarial Networks and Perceptual Losses for Video Super-Resolution.

Video super-resolution (VSR) has become one of the most critical problems in video processing. In the deep learning literature, recent works have shown the benefits of using adversarial-based and perceptual losses to improve the performance on various image restoration tasks; however, these have yet to be applied for video super-resolution. In this paper, we propose a generative adversarial network (GAN)-based formulation for VSR. We introduce a new generator network optimized for the VSR problem, named VSRResNet, along with new discriminator architecture to properly guide VSRResNet during the GAN training. We further enhance our VSR GAN formulation with two regularizers, a distance loss in feature-space and pixel-space, to obtain our final VSRResFeatGAN model. We show that pre-training our generator with the mean-squared-error loss only quantitatively surpasses the current state-of-the-art VSR models. Finally, we employ the PercepDist metric to compare the state-of-the-art VSR models. We show that this metric more accurately evaluates the perceptual quality of SR solutions obtained from neural networks, compared with the commonly used PSNR/SSIM metrics. Finally, we show that our proposed model, the VSRResFeatGAN model, outperforms the current state-of-the-art SR models, both quantitatively and qualitatively.