Recent years have witnessed remarkable success of deep learning methods in quality enhancement for compressed video. To better explore temporal information, existing methods usually estimate optical flow for temporal motion compensation. However, since compressed video could be seriously distorted by various compression artifacts, the estimated optical flow tends to be inaccurate and unreliable, thereby resulting in ineffective quality enhancement. In addition, optical flow estimation for consecutive frames is generally conducted in a pairwise manner, which is computational expensive and inefficient. In this paper, we propose a fast yet effective method for compressed video quality enhancement by incorporating a novel Spatio-Temporal Deformable Fusion (STDF) scheme to aggregate temporal information. Specifically, the proposed STDF takes a target frame along with its neighboring reference frames as input to jointly predict an offset field to deform the spatio-temporal sampling positions of convolution. As a result, complementary information from both target and reference frames can be fused within a single Spatio-Temporal Deformable Convolution (STDC) operation. Extensive experiments show that our method achieves the state-of-the-art performance of compressed video quality enhancement in terms of both accuracy and efficiency.

No abstract

Purpose Epilepsy is one of the most disabling neurological disorders, which affects all age groups and often results in severe consequences. Since misdiagnoses are common, many pediatric patients fail to receive the correct treatment. Recently, 18F-fluorodeoxyglucose positron emission tomography (18F-FDG PET) imaging has been used for the evaluation of pediatric epilepsy. However, the epileptic focus is very difficult to be identified by visual assessment since it may present either hypo- or hyper-metabolic abnormality with unclear boundary. This study aimed to develop a novel symmetricity-driven deep learning framework of PET imaging for the identification of epileptic foci in pediatric patients with temporal lobe epilepsy (TLE). Methods We retrospectively included 201 pediatric patients with TLE and 24 age-matched controls who underwent 18F-FDG PET-CT studies. 18F-FDG PET images were quantitatively investigated using 386 symmetricity features, and a pair-of-cube (PoC)-based Siamese convolutional neural network (CNN) was proposed for precise localization of epileptic focus, and then metabolic abnormality level of the predicted focus was calculated automatically by asymmetric index (AI). Performances of the proposed framework were compared with visual assessment, statistical parametric mapping (SPM) software, and Jensen-Shannon divergence-based logistic regression (JS-LR) analysis. Results The proposed deep learning framework could detect the epileptic foci accurately with the dice coefficient of 0.51, which was significantly higher than that of SPM (0.24, P < 0.01) and significantly (or marginally) higher than that of visual assessment (0.31–0.44, P = 0.005–0.27). The area under the curve (AUC) of the PoC classification was higher than that of the JS-LR (0.93 vs. 0.72). The metabolic level detection accuracy of the proposed method was significantly higher than that of visual assessment blinded or unblinded to clinical information (90% vs. 56% or 68%, P < 0.01). Conclusion The proposed deep learning framework for 18F-FDG PET imaging could identify epileptic foci accurately and efficiently, which might be applied as a computer-assisted approach for the future diagnosis of epilepsy patients. Trial registration NCT04169581. Registered November 13, 2019 Public site: https://clinicaltrials.gov/ct2/show/NCT04169581

No abstract

Automatic segmentation of cardiac magnetic resonance imaging (MRI) facilitates efficient and accurate volume measurement in clinical applications. However, due to anisotropic resolution and ambiguous border (e.g., right ventricular endocardium), existing methods suffer from the degradation of accuracy and robustness in 3D cardiac MRI video segmentation. In this paper, we propose a novel Deformable U-Net (DeU-Net) to fully exploit spatio-temporal information from 3D cardiac MRI video, including a Temporal Deformable Aggregation Module (TDAM) and a Deformable Global Position Attention (DGPA) network. First, the TDAM takes a cardiac MRI video clip as input with temporal information extracted by an offset prediction network. Then we fuse extracted temporal information via a temporal aggregation deformable convolution to produce fused feature maps. Furthermore, to aggregate meaningful features, we devise the DGPA network by employing deformable attention U-Net, which can encode a wider range of multi-dimensional contextual information into global and local features. Experimental results show that our DeU-Net achieves the state-of-the-art performance on commonly used evaluation metrics, especially for cardiac marginal information (ASSD and HD).