DQDefu Qiu

Defu Qiu

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Defu Qiu has a total of 23 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • UMIC: Super-Resolution of Cine Cardiac MRI Using U-Shaped Network With Multi-Level Information Compensation

    by Defu Qiu, Kelvin K.L. Wong, Yuhu Cheng, Zhang Yi, Xuesong Wang

    In cine cardiac magnetic resonance imaging (CMRI), deep learning-based super-resolution (SR) reconstruction algorithms often suffer from feature information loss during feature extraction and lack effective mechanisms for feature compensation. These problems can lead to the lack of texture and edge details in the reconstructed image, making it difficult to obtain a clear cardiac image, which will increase the rate of misjudgment of cardiac disease by experts. To address these issues, we propose a U-shaped network with multi-level information compensation (UMIC). Specifically, the network first performs multi-level feature extraction on low-resolution (LR) inputs and reduces channel dimensionality via a downward channel branch. The compressed features are then fused through a bottom module to capture inter-channel dependencies. Finally, the relevant features are recovered and enhanced through an upward channel branch. Additionally, we introduce a multi-level information compensation module to mitigate detail loss incurred during channel compression and to assist in recovering difficult-to-restore LR image details in the reconstruction phase. Extensive experiments show that UMIC achieves better CMRI SR reconstruction performance compared to some state-of-the-art SR methods.


  • CCDM: Continuous-Time Conditional Diffusion Model for Blind CMRI Super-Resolution

    by Defu Qiu, Junliang Shang, Yuanke Zhang, Wenjun Zhang, Kelvin K. L. Wong, Zhang Yi

    Diffusion probabilistic models have effectively addressed the ill-posed nature of cardiac magnetic resonance imaging (CMRI) super-resolution (SR) by learning high-resolution image distributions from low-resolution inputs. However, the iterative sampling process in these models often suffers from slow inference speeds, as well as limitations in the quality and structural consistency of the generated images. To address these challenges, we propose a continuous-time conditional diffusion model (CCDM) for blind CMRI SR. Specifically, we propose a continuous-time conditional diffusion module that reduces the time consumption of the diffusion probability model by maintaining the mean and variance of the data in the forward process. Meanwhile, we design a cascaded residual attention network as a feature extractor to enhance the model’s discriminative power and feature representation capabilities. To further elevate image fidelity, we propose an image quality loss module that integrates a score matching loss, significantly improving detail reconstruction and overall perceptual quality. Furthermore, we develop a hybrid score predictor that approximates the conditional score function via a hybrid parameterized denoising network, facilitating efficient CMRI generation through probability flow sampling. Extensive experimental results demonstrate that compared to existing diffusion model-based SR methods, our CCDM achieves significant improvements in SR quality while substantially reducing time consumption.


  • DBSR: Quadratic Conditional Diffusion Model for Blind Cardiac MRI Super-Resolution

    by Defu Qiu, Yuhu Cheng, Kelvin K.L. Wong, Wenjun Zhang, Zhang Yi, Xuesong Wang

    Cardiac magnetic resonance imaging (CMRI) can help experts quickly diagnose cardiovascular diseases. Due to the patient's breathing and slight movement during the magnetic resonance imaging scan, the obtained CMRI may be severely blurred, affecting the accuracy of clinical diagnosis. To address this issue, we propose the quadratic conditional diffusion model for blind CMRI super-resolution (DBSR). Specifically, we propose a conditional blur kernel noise predictor, which predicts the blur kernel from low-resolution images by the diffusion model, transforming the unknown blur kernel in low-resolution CMRI into a known one. Meanwhile, we design a novel conditional CMRI noise predictor, which uses the predicted blur kernel as prior knowledge to guide the diffusion model in reconstructing high-resolution CMRI. Furthermore, we propose a cascaded residual attention network feature extractor, which extracts feature information from CMRI low-resolution images for blur kernel prediction and SR reconstruction of CMRI images. Extensive experimental results indicate that our proposed DBSR achieves better blind super-resolution reconstruction results than several state-of-the-art baselines. Our code and model are publicly available at https://github.com/Q-D-F/DBSR .


  • ABSR: Progressive Alternate Refinement for Blind Cardiac MRI Super-Resolution

    by Defu Qiu, Zhaoyang Song, Wenjun Zhang, Kelvin K. L. Wong, Zhang Yi

    Deep learning-based methods for super-resolution reconstruction of cardiac magnetic resonance imaging (CMRI) have achieved commendable reconstruction performance owing to the potent learning capability of neural networks. Nonetheless, these methods suffer from performance degradation when handling real-world CMRI images, failing to reconstruct high-fidelity CMRI high-resolution images. This degradation stems from the fact that real-world CMRI images are afflicted with blur and noise, whereas mainstream deep learning-based CMRI super-resolution algorithms are typically trained on images degraded using bicubic methods. To address this problem, we propose a progressive alternate refinement for blind CMRI super-resolution, which implements blind super-resolution reconstruction through progressive alternate refinement optimization of the CMRI image feature extraction process and blur kernel feature extraction process. Moreover, we propose a novel progressive blind super-resolution reconstruction sub-network, which utilizes the alternate residual attention block to perform deep feature extraction. Meanwhile, we propose an alternate residual attention block, which uses channel and pixel attention mechanisms to extract high-frequency features from the extracted CMRI and blur kernel features. Extensive experimental results demonstrate that our proposed ABSR outperforms the state-of-the-art SR methods, exhibiting superior reconstruction performance and the ability to reconstruct high-fidelity CMRI images.


  • Parallel Alternating Iterative Optimization for Cardiac Magnetic Resonance Image Blind Super-Resolution

    by Zhaoyang Song, Defu Qiu, Xiaoqiang Zhao, Ruijuan Liu, Yongyong Hui, Hongmei Jiang

    Cardiac magnetic resonance imaging (CMRI) super-resolution (SR) reconstruction technology can enhance the resolution and quality of CMRI, providing experts with clearer and more accurate information about cardiac structure and function. This technology aids in the rapid and accurate diagnosis of cardiac abnormalities and the development of personalized treatment plans. In the processing of CMRI, existing bicubic degradation-based SR methods often suffer from performance degradation, resulting in blurred SR images. To address the aforementioned problem, we present a parallel alternating iterative optimization for CMRI image blind SR method (PAIBSR). Specifically, we propose a parallel alternating iterative optimization strategy, which employs dynamically corrected blur kernels and dynamically extracted intermediate low-resolution features as prior knowledge for both the blind SR process and the blur kernel correction process. Meanwhile, we propose a blur kernel update module composed of a blur kernel extractor and a low-resolution kernel extractor to correct the blur kernel. Furthermore, we propose an enhanced spatial feature transformation residual block, leveraging the corrected blur kernel as prior knowledge for the blind SR process. Through extensive experiments conducted on synthetic datasets, we have validated the superiority of PAIBSR method. It outperforms state-of-the-art SR methods in terms of performance and produces visually pleasing results.


  • Progressive Feedback Residual Attention Network for Cardiac Magnetic Resonance Imaging Super-Resolution

    by Defu Qiu, Yuhu Cheng, Xuesong Wang

    Atrial fibrillation (AF) is an increasing medical burden worldwide, and its pathological manifestations are atrial tissue remodeling and low-pressure atrial tissue fibrosis. Due to the inherent defects of medical image data acquisition systems, the acquisition of high-resolution cardiac magnetic resonance imaging (CMRI) faces many problems. In response to these problems, we propose the Progressive Feedback Residual Attention Network (PFRN) for CMRI super-resolution. Specifically, we directly perform feature extraction on low-resolution images, retain feature information to a large extent, and then build multiple independent progressive feedback modules to extract high-frequency details. To accelerate network convergence and improve image reconstruction quality, we implement the MS-SSIM-L1 loss function. Furthermore, we utilize the residual attention stack module to explore the image's internal relevance and extract the low-resolution image's detailed features. Extensive benchmark evaluation shows that PFRN can improve the detailed information of the image SR reconstruction results, and the reconstructed CMRI has a better visual effect.


  • Residual Dense Attention Networks for COVID-19 Computed Tomography Images Super Resolution

    by Defu Qiu, Yuhu Cheng, Xuesong Wang

    The coronavirus disease 2019 (COVID-19) is highly infectious and pathogenic and poses a severe threat to people’s public safety. Owing to the low-resolution of computed tomography (CT) images, it is essential to use super-resolution reconstruction technology to improve the resolution of COVID-19 medical images. Aiming at the problems of the limited receptive field, low resolution, high complexity, and loss of edge information in the super-resolution reconstruction method of residual learning, we present a residual dense attention network (RDAN) for COVID-19 CT image super-resolution. Firstly, to better extract features and reduce the number of parameters, we design the residual dense network module to extract detailed information from images. Secondly, we add the channel attention mechanism to enable the network to have adequate high-frequency information with larger weights to reduce the computational cost of the model. Finally, we filter and reorganize the multi-layer image information by skip connections so that the network model can allow extensive use of image information of different depths. Comprehensive benchmark evaluation shows that our RDAN method dramatically improves the convergence speed of the network, solves the problem of missing information, and makes the reconstructed COVID-19 CT images have more apparent textures, richer details, and better visual effects that can effectively assist experts in diagnosis.