Excessive Smoothing Research Articles

Triangular Mesh Surface Subdivision Based on Graph Neural Network

Mesh subdivision is a common mesh-processing algorithm used to improve model accuracy and surface smoothness. Its classical scheme adopts a fixed linear vertex update strategy and is implemented iteratively, which often results in excessive mesh smoothness. In recent years, a nonlinear subdivision method that uses neural network methods, called neural subdivision (NS), has been proposed. However, as a new scheme, its application scope and the effect of its algorithm need to be improved. To solve the above problems, a graph neural network method based on neural subdivision was used to realize mesh subdivision. Unlike fixed half-flap structures, the non-fixed mesh patches used in this paper naturally expressed the interior and boundary of a mesh and learned its spatial and topological features. The tensor voting strategy was used to replace the half-flap spatial transformation method of neural subdivision to ensure the translation, rotation, and scaling invariance of the algorithm. Dynamic graph convolution was introduced to learn the global features of the mesh in the way of stacking, so as to improve the subdivision effect of the network on the extreme input mesh. In addition, vertex neighborhood information was added to the training data to improve the robustness of the subdivision network. The experimental results show that the proposed algorithm achieved a good subdivision of both the general input mesh and extreme input mesh. In addition, it effectively subdivided mesh boundaries. In particular, using the general input mesh, the algorithm in this paper was compared to neural subdivision through quantitative experiments. The proposed method reduced the Hausdorff distance and the mean surface distance by 27.53% and 43.01%, respectively.

Elucidation of FDG-PET Imaging Characteristics that Reflect the Heterogeneity within Tumors Due to Variability in Metabolic Activity: A Phantom Study

Abstract Objective Focusing on the heterogeneity within cancer lesions and revealing cancer lesions with a high signal-to-noise ratio will help improve the quality of positron emission tomography (PET) images. This study aimed to understand how glucose metabolic activity can be shown with less statistical noise using a quality assessment phantom modeled after cancer lesions with a two-layer structure and a Clear adaptive Low-noise Method (CaLM) filter. Materials and Methods A National Electrical Manufacturers Association phantom with two spheres with a two-layer structure filled with 2-deoxy-2[18F]fluoro-D-glucose (inner and outer diameters of the spheres were 13/22 and 22/37 mm, respectively; radioactivity ratio of the background [BG] to outer sphere layer was 1:4; and BG-to-inner sphere layer ratios were 1:1, 1:2, and 1:3) was evaluated. The acquisition time was set at 120 seconds, and imaging was repeated five times. The image data in photomultiplier tube (PMT)-PET were reconstructed using a time-of-flight (TOF) ordered subset expectation maximization (OSEM) algorithm (point spread function [PSF] − , iteration: 3, subset: 10) with a 4-mm Gaussian filter (GF) for normal images. Silicon photomultiplier (SiPM)-PET data were reconstructed using a TOF OSEM algorithm (PSF − , iteration: 2, subset: 12) and a 3-mm GF for normal images. The target images were reconstructed using three CaLM parameters (mild, standard, and strong). All the obtained images were investigated quantitatively, with calculation of maximum standard uptake value (SUVmax), coefficient of variation (CV), and contrast-to-noise ratio (CNR), after setting regions of interest on the lesions. Statistical analysis using the Dunnett's test compared normal images (control group) and target images (treatment group). Statistical significance was considered at p < 0.05. Results Quantitative assessment revealed that the SUVmax of target images (standard and strong) was equivalent to that of normal images in PMT-PET, with SUVmax of 3 to 3.5 for both layers. The SUVmax in SiPM-PET was similar across all CaLM types, ranging from 3 to 4 for all spheres. The target images (standard and strong) had a significantly reduced CV and improved CNR compared with normal images. Conclusion The boundary of the 22/37-mm spheres was visible with CaLM (strong) at radioactivity ratios of 1:4:1 and 1:4:2 on both scanners. For the 13/22-mm sphere boundary, visibility with CaLM (strong) was observed only with SiPM-PET, with the SUVmax equivalent to normal images. CaLM (strong) was deemed the optimal postprocessing filter for PMT-PET due to significant improvements in CV and CNR, while CaLM (standard) was suggested as the optimal filter for SiPM-PET due to excessive BG smoothing.

DGGNets: Deep Gradient-Guidance Networks for Speckle Noise Reduction

Speckle noise is a granular interference that degrades image quality in coherent imaging systems, including underwater sonar, Synthetic Aperture Radar (SAR), and medical ultrasound. This study aims to enhance speckle noise reduction through advanced deep learning techniques. We introduce the Deep Gradient-Guidance Network (DGGNet), which features an architecture comprising one encoder and two decoders—one dedicated to image recovery and the other to gradient preservation. Our approach integrates a gradient map and fractional-order total variation into the loss function to guide training. The gradient map provides structural guidance for edge preservation and directs the denoising branch to focus on sharp regions, thereby preventing over-smoothing. The fractional-order total variation mitigates detail ambiguity and excessive smoothing, ensuring rich textures and detailed information are retained. Extensive experiments yield an average Peak Signal-to-Noise Ratio (PSNR) of 31.52 dB and a Structural Similarity Index (SSIM) of 0.863 across various benchmark datasets, including McMaster, Kodak24, BSD68, Set12, and Urban100. DGGNet outperforms existing methods, such as RIDNet, which achieved a PSNR of 31.42 dB and an SSIM of 0.853, thereby establishing new benchmarks in speckle noise reduction.

Wavelet detection method for transient electrical energy disturbance with noise

Abstract Electric energy signals are often contaminated by noise during the collection process, which affects the accurate positioning of disturbances. The discontinuity of traditional hard threshold functions can lead to oscillations, while excessive smoothness of soft threshold functions can lead to signal distortion. The optimized threshold function and scale adaptive processing on the selection of the threshold are proposed. The wavelet thresholding method is used to process voltage sag, rise, and interrupt signals containing noise. The denoised signal is decomposed into five scale wavelets using db7 wavelet and detected by using wavelet transform modulus maximum and signal singularity principle. Experiments have shown that optimized threshold and threshold function not only effectively denoise but also retain the effective mutation point information in the original signal, and the new method is reliable in detection.

SAR-NTV-YOLOv8: A Neural Network Aircraft Detection Method in SAR Images Based on Despeckling Preprocessing

Monitoring aircraft using synthetic aperture radar (SAR) images is a very important task. Given its coherent imaging characteristics, there is a large amount of speckle interference in the image. This phenomenon leads to the scattering information of aircraft targets being masked in SAR images, which is easily confused with background scattering points. Therefore, automatic detection of aircraft targets in SAR images remains a challenging task. For this task, this paper proposes a framework for speckle reduction preprocessing of SAR images, followed by the use of an improved deep learning method to detect aircraft in SAR images. Firstly, to improve the problem of introducing artifacts or excessive smoothing in speckle reduction using total variation (TV) methods, this paper proposes a new nonconvex total variation (NTV) method. This method aims to ensure the effectiveness of speckle reduction while preserving the original scattering information as much as possible. Next, we present a framework for aircraft detection based on You Only Look Once v8 (YOLOv8) for SAR images. Therefore, the complete framework is called SAR-NTV-YOLOv8. Meanwhile, a high-resolution small target feature head is proposed to mitigate the impact of scale changes and loss of depth feature details on detection accuracy. Then, an efficient multi-scale attention module was proposed, aimed at effectively establishing short-term and long-term dependencies between feature grouping and multi-scale structures. In addition, the progressive feature pyramid network was chosen to avoid information loss or degradation in multi-level transmission during the bottom-up feature extraction process in Backbone. Sufficient comparative experiments, speckle reduction experiments, and ablation experiments are conducted on the SAR-Aircraft-1.0 and SADD datasets. The results have demonstrated the effectiveness of SAR-NTV-YOLOv8, which has the most advanced performance compared to other mainstream algorithms.

SRARDA: A lightweight adaptive residual dense attention generative adversarial network for image super-resolution

Image super-resolution (SR) is the task of inferring a high resolution (HR) image from one/multiple single low resolution (LR) input(s). Traditional networks are evaluated by pixel-level metrics such as Peak-Signal-to-Noise Ratio (PSNR) etc., which do not always align with human perception of image quality. They often produce excessively smooth images that lack high-frequency texture and appear unnatural. Therefore, in this paper, we propose a lightweight adaptive residual dense attention generative adversarial network (SRARDA) for image SR. Firstly, our generator adopts the residual in residual (RIR) structure but redesigns the basic module. By using dynamic residual connection (ARC) to dynamically adjust the importance of residual and main paths, we design a novel adaptive residual dense attention block (ARDAB) that enhances the feature extraction capability of the generator. In addition, we build a high-frequency filtering unit (HFU) to extract more high-frequency features from the LR space. Finally, to fully utilize the discriminator, we use WGAN to compute the difference between the HR image and the reconstructed image. Experiments demonstrate that SRARDA effectively addresses the issue of excessive smoothing in reconstructed images, while also enhancing visual quality.

Leakage monitoring of carbon dioxide injection well string using distributed optical fiber sensor

Whether the oil and gas fields in the Carbon Capture, Utilization, and Storage (CCUS) project use underground storage or energy supplementation to enhance oil recovery, they must be injected or monitored through the wellbore. Thus, the foundation and requirement for the safety of carbon dioxide (CO2) storage is the wellbore's integrity. When CO2 is dissolved in water, carbonic acid is created, and this acid strongly corrodes underground pipes. Therefore, the integrity issue with CO2 injection wells is more noticeable than with other wellbores. An annular pressure during gas injection is the primary symptom of gas injection string leakage in CO2 injection wells. This study aims to provide real-time pipe string monitoring using a distributed optical fiber temperature sensing system (DTS) and a distributed optical fiber acoustic sensing system (DAS). Variations in temperature and vibration are caused by annulus pressure relief or gas injection. Optical fiber logging, in contrast to traditional logging, has better performance indicators for optical fiber sensing apparatus. To adapt to complex wellbore conditions, it is necessary to enhance the temperature accuracy of DTS and the sensitivity and signal-to-noise ratio of DAS in CO2 drive injection wells based on the features of the gas injection string. To differentiate the leakage signal from the regular fluid flow signal, the energy calculation in the frequency band is done for DAS based on noise reduction, and the signal processing in the frequency band is done by the spectrum characteristics of the CO2 wellbore signal. The translation invariant wavelet algorithm is the primary denoising method for DTS, overcoming the shortcomings of traditional wavelet threshold algorithms such as excessive smoothing and the pseudo-Gibbs phenomenon. Furthermore, the depth correction during the optical cable lowering process is also examined in this paper. A CO2 gas injection well field experiment was conducted using this technology. A 1671m well was dug, and 1631m of optical cable were installed in the tubing. The tubing leakage position was successfully identified through gas injection, annulus pressure relief, and a comparison of DAS and DTS data. The field results demonstrate the accuracy with which the gas injection string integrity can be accurately monitored in real-time using distributed optical fiber sensing technology for CO2 injection wells.

Sound event detection in traffic scenes based on graph convolutional network to obtain multi-modal information

Sound event detection involves identifying sound categories in audio and determining when they start and end. However, in real-life situations, sound events are usually not isolated. When one sound event occurs, there are often other related sound events that take place as co-occurrences or successive occurrences. The timing relationship of sound events can reflect their characteristics. Therefore, this paper proposes a sound event detection method for traffic scenes based on a graph convolutional network, which considers this timing relationship as a form of multimodal information. The proposed method involves using the acoustic event window method to obtain co-occurrences or successive occurrences of relationship information in the sound signal while filtering out possible noise relationship information. This information is then represented as a graphical structure. Next, the graph convolutional neural network is improved to balance relationship weights between neighbors and itself and to avoid excessive smoothing. It is used to learn the relationship information in the graph structure. Finally, the convolutional recurrent neural network is used to learn the acoustic feature information of sound events, and the relationship information of sound events is obtained by multi-modal fusion to enhance the performance of sound event detection. The experimental results show that using multi-modal information with the proposed method can effectively improve the performance of the model and enhance the perception ability of smart cars in their surrounding environment while driving.

Research on traffic flow prediction method based on adaptive multi-channel graph convolutional neural networks

In order to address the issues of predefined adjacency matrices inadequately representing information in road networks, insufficiently capturing spatial dependencies of traffic networks, and the potential problem of excessive smoothing or neglecting initial node information as the layers of graph convolutional neural networks increase, thus affecting traffic prediction performance, this paper proposes a prediction model based on Adaptive Multi-channel Graph Convolutional Neural Networks (AMGCN). The model utilizes an adaptive adjacency matrix to automatically learn implicit graph structures from data, introduces a mixed skip propagation graph convolutional neural network model, which retains the original node states and selectively acquires outputs of convolutional layers, thus avoiding the loss of node initial states and comprehensively capturing spatial correlations of traffic flow. Finally, the output is fed into Long Short-Term Memory networks to capture temporal correlations. Comparative experiments on two real datasets validate the effectiveness of the proposed model.

Multi-task learning with self-learning weight for image denoising

BackgroundImage denoising technology removes noise from the corrupted image by utilizing different features between image and noise. Convolutional neural network (CNN)-based algorithms have been the concern of the recent progress on diverse image restoration problems and become an efficient solution in image denoising.ObjectiveAlthough a quite number of existing CNN-based image denoising methods perform well on the simplified additive white Gaussian noise (AWGN) model, their performance often degrades severely on the real-world noisy images which are corrupted by more complicated noise.MethodsIn this paper, we utilized the multi-task learning (MTL) framework to integrate multiple loss functions for collaborative training of CNN. This approach aims to improve the denoising performance of CNNs on real-world images with non-Gaussian noise. Simultaneously, to automatically optimize the weights of individual sub-tasks within the MTL framework, we incorporated a self-learning weight layer into the CNN.ResultsExtensive experiments demonstrate that our approach effectively enhances the denoising performance of CNN-based image denoising algorithms on real-world images. It reduces excessive image smoothing, improves quantitative metrics, and enhances visual quality in the restored images.ConclusionOur method shows the effectiveness of the improved performance of denoising CNNS for real-world image denoising processing.

KGIE: Knowledge graph convolutional network for recommender system with interactive embedding

In recent years, knowledge graphs (KGs) have gained considerable traction across various domains, especially in the realm of recommender systems, where their integration has garnered significant interest. These integrations aim to enhance the accuracy of recommendations by leveraging user–item interaction data and item attributes. However, existing methods encounter several challenges, including excessive smoothing, sparse data, redundancy, inadequate consideration of auxiliary information, and limitations in constructing deep networks. To address these challenges and enhance knowledge-graph-based recommendation methods that ignore auxiliary information and framework redundancy in neighborhood information aggregation, this study proposes a novel graph neural network recommendation model based on interactive embedding. This model capitalizes on both the KG and user–item interaction matrix to extract valuable information, refine aggregation methods, and optimize the overall performance. Specifically, user–relation interactive embedding is formed by establishing connections between users and relations using the user–item interaction matrix and KG as the foundation. This interactive embedding merges with the user through convolutional neural networks (CNNs), independently participating in the aggregation of neighborhood information to provide more contextual information for recommendation. The set of users who interact with items is extracted employing the user–item interaction matrix and creating item–user interactive embedding. It is then merged with the item using a CNN. Lastly, the final representations of the users and items for prediction are obtained. Experimental evaluations conducted on six real recommendation datasets demonstrate that our proposed model outperforms existing baselines.

CT respiratory motion synthesis using joint supervised and adversarial learning

Objective. Four-dimensional computed tomography (4DCT) imaging consists in reconstructing a CT acquisition into multiple phases to track internal organ and tumor motion. It is commonly used in radiotherapy treatment planning to establish planning target volumes. However, 4DCT increases protocol complexity, may not align with patient breathing during treatment, and lead to higher radiation delivery. Approach. In this study, we propose a deep synthesis method to generate pseudo respiratory CT phases from static images for motion-aware treatment planning. The model produces patient-specific deformation vector fields (DVFs) by conditioning synthesis on external patient surface-based estimation, mimicking respiratory monitoring devices. A key methodological contribution is to encourage DVF realism through supervised DVF training while using an adversarial term jointly not only on the warped image but also on the magnitude of the DVF itself. This way, we avoid excessive smoothness typically obtained through deep unsupervised learning, and encourage correlations with the respiratory amplitude. Main results. Performance is evaluated using real 4DCT acquisitions with smaller tumor volumes than previously reported. Results demonstrate for the first time that the generated pseudo-respiratory CT phases can capture organ and tumor motion with similar accuracy to repeated 4DCT scans of the same patient. Mean inter-scans tumor center-of-mass distances and Dice similarity coefficients were 1.97 mm and 0.63, respectively, for real 4DCT phases and 2.35 mm and 0.71 for synthetic phases, and compares favorably to a state-of-the-art technique (RMSim). Significance. This study presents a deep image synthesis method that addresses the limitations of conventional 4DCT by generating pseudo-respiratory CT phases from static images. Although further studies are needed to assess the dosimetric impact of the proposed method, this approach has the potential to reduce radiation exposure in radiotherapy treatment planning while maintaining accurate motion representation. Our training and testing code can be found at https://github.com/cyiheng/Dynagan.

Machine learning rather than trial and error to close morphodynamical tuneable parameters: Application to a two-phase/two-layer model

ABSTRACT Physics-based numerical models often depend on several parameters to close. Some of them can be expressed using established theoretical or empirical closure formulations. However, some others aggregate complex physical processes and are hence left as tuneable parameters, and can only be calibrated by trial and error. Yet, calibration data are not always available to do so, which prevents these models from being applied to wide ranges of laboratory or river flows. We hence propose a machine learning-based methodology to close any group of unclosed and correlated parameters, applied here to a two-phase/two-layer (2P2L) morphodynamical model. The methodology combines a numerical experiment with a known theoretical solution and machine learning. It is applied to the considered model to close two friction parameters for which generalizable and vastly acknowledged closure formulations lack in the literature. The resulting hybrid model, combining the original 2P2L model and the closure models, is tested against two laboratory dam break test cases. Despite excessive smoothness and underestimation of the concentration in sediment, the hybrid model performed similarly to other models from the literature requiring trial and error calibration and showed high stability and accuracy regarding the estimation of the water-sediment mixture's inertia.

Discrete residual diffusion model for high-resolution prostate MRI synthesis

Objective. High-resolution magnetic resonance imaging (HR MRI) is an effective tool for diagnosing PCa, but it requires patients to remain immobile for extended periods, increasing chances of image distortion due to motion. One solution is to utilize super-resolution (SR) techniques to process low-resolution (LR) images and create a higher-resolution version. However, existing medical SR models suffer from issues such as excessive smoothness and mode collapse. In this paper, we propose a novel generative model avoiding the problems of existing models, called discrete residual diffusion model (DR-DM). Approach. First, the forward process of DR-DM gradually disrupts the input via a fixed Markov chain, producing a sequence of latent variables with increasing noise. The backward process learns the conditional transit distribution and gradually match the target data distribution. By optimizing a variant of the variational lower bound, training diffusion models effectively address the issue of mode collapse. Second, to focus DR-DM on recovering high-frequency details, we synthesize residual images instead of synthesizing HR MRI directly. The residual image represents the difference between the HR and LR up-sampled MR image, and we convert residual image into discrete image tokens with a shorter sequence length by a vector quantized variational autoencoder (VQ-VAE), which reduced the computational complexity. Third, transformer architecture is integrated to model the relationship between LR MRI and residual image, which can capture the long-range dependencies between LR MRI and the synthesized imaging and improve the fidelity of reconstructed images. Main results. Extensive experimental validations have been performed on two popular yet challenging magnetic resonance image super-resolution tasks and compared to five state-of-the-art methods. Significance. Our experiments on the Prostate-Diagnosis and PROSTATEx datasets demonstrate that the DR-DM model significantly improves the signal-to-noise ratio of MRI for prostate cancer, resulting in greater clarity and improved diagnostic accuracy for patients.

IMPROVING ULTRASOUND IMAGE EDGE ENHANCEMENT WITH AADFP METHOD

Image processing is a constantly evolving field that is actively seeking innovative solutions to improve the quality and accuracy of image analysis. Smoothing the edges of ultrasound images using filters is a common processing method that reduces noise and improves image clarity. The choice of filtering method and its parameters depends on the specific clinical or research requirements and the nature of the ultrasound images being processed. However, traditional image processing approaches, especially smoothing, have their limitations. An example is medical imaging, where the preservation of clear anatomical image structures is of paramount importance. Excessive smoothing in such cases leads to the erasure of important anatomical details, potentially affecting diagnostic accuracy. In "computer-aided imaging", excessive smoothing of the edges of the investigated objects may prevent their accurate recognition, which is particularly important when the object is characterised by a variety of features. Due to the limitations of traditional methods, the introduction of better approaches is required to address this problem. The AADFP (Adaptive Anisotropic Diffusion with Feature Priority) method is a technique for improving the quality of ultrasound images, which allows a new approach to their processing, by simultaneously smoothing edges and reducing noise, while preserving important anatomical features. The method is based on adaptive anisotropic diffusion, which adjusts the diffusion process depending on image filling, combined with the integration of features previously extracted from a database of annotated ultrasound images. The use of AADFP allows to significantly improve image quality by selectively smoothing out areas with noise while preserving the main structures in the image. Keywords: Ultrasound image processing, AADFP Method, Image Processing, Medical Imaging, Filtering methods, Feature recognition.

Moving from theory to application: Evaluating the numerical implementation of void shape effects and damage delocalization in the modeling of ductile fracture in porous plastic metals

In this paper, we present a robust exploration of the Gologanu–Leblond–Devaux (GLD) model, an advanced iteration of the Gurson model, designed to predict ductile fractures in porous metals. Going beyond the limits of the original Gurson model, the GLD model accounts for cavity shape effects and non-local strain localization, marking a significant leap in fracture mechanics. We also present a comprehensive exposition of the GLD model and its non-local extension, establishing their compatibility with the concept of GSMs. Notably, we emphasize the uniqueness of solutions in the numerical implementation, underlining the imperative need for a meticulously devised mixed implicit/explicit algorithm. Furthermore, we set out to validate the GLD model through rigorous comparisons of our numerical simulations with experimental data. Employing a damage delocalization approach rooted in the natural logarithm of porosity, our study provides compelling evidence of the model’s performance. This approach mitigates issues observed with the original porosity rate, preventing excessive smoothing of porosity and maintaining the fidelity of stress–strain curves. In addition, we gave a profound theoretical elucidation of this phenomenon via Fourier’s analysis of porosity rate. Through this work, we not only enhance our understanding of ductile fracture behavior but also establish a robust numerical framework for its predictive modeling. The GLD model emerges as a powerful tool for the accurate analysis and prediction of fracture phenomena in porous materials, further advancing the field of materials science and engineering.

Arbitrary scale super-resolution diffusion model for brain MRI images

Given the constraints posed by hardware capacity, scan duration, and patient cooperation, the reconstruction of magnetic resonance imaging (MRI) images emerges as a pivotal aspect of medical imaging research. Currently, deep learning-based super-resolution (SR) methods have been widely discussed in medical image processing due to their ability to reconstruct high-quality, high resolution (HR) images from low resolution (LR) inputs. However, most existing MRI SR methods are designed for specific magnifications and cannot generate MRI images at arbitrary scales, which hinders the radiologists from fully visualizing the lesions. Moreover, current arbitrary scale SR methods often suffer from issues like excessive smoothing and artifacts. In this paper, we propose an Arbitrary Scale Super-Resolution Diffusion Model (ASSRDM), which combines implicit neural representation with the denoising diffusion probabilistic model to achieve arbitrary-scale, high-fidelity medical images SR. Moreover, we formulate a continuous resolution regulation mechanism, comprising a multi-scale LR guidance network and a scaling factor. The scaling factor finely adjusts the resolution and dynamically influences the weighting of LR details and synthesized features in the final output. This capability allows the model to seamlessly adapt to the requirements of continuous resolution adjustments. Additionally, the multi-scale LR guidance network provides the denoising block with multi-resolution LR features to enrich texture information and restore high-frequency details. Extensive experiments conducted on the IXI and fastMRI datasets demonstrate that our ASSRDM exhibits superior performance compared to existing techniques and has tremendous potential in clinical practice.

EWRD: Entropy‐Weighted Low‐Light Image Enhancement via Reverse Diffusion Model

Low‐light enhancement significantly aids vision tasks under poor illumination conditions. Existing methods primarily focus on enhancing severely degraded low‐light areas, improving illumination accuracy, or noise suppression, yet they often overlook the loss of additional details and color distortion during calculation. In this paper, we propose an innovative entropy‐weighted low‐light image enhancement method via the reverse diffusion model, aiming at addressing the limitations of the traditional Retinex decomposition model in preserving local pixel details and handling excessive smoothing issues. This method integrates an entropy‐weighting mechanism for improved image quality and entropy, along with a reverse diffusion model to address the detail loss in total variation regularization and refine the enhancement process. Furthermore, we utilize long short‐term memory networks for the learning reverse process and the simulation of image degradation, based on a thermodynamics‐based nonlinear anisotropic diffusion model. Comparative experiments reveal the superiority of our method over conventional Retinex‐based approaches in terms of detail preservation and visual quality. Extensive tests across diverse datasets demonstrate the exceptional performance of our method, evidencing its potential as a robust solution for low‐light image enhancement.

Predicting miRNA-disease association via graph attention learning and multiplex adaptive modality fusion

miRNAs are a class of small non-coding RNA molecules that play important roles in gene regulation. They are crucial for maintaining normal cellular functions, and dysregulation or dysfunction of miRNAs which are linked to the onset and advancement of multiple human diseases. Research on miRNAs has unveiled novel avenues in the realm of the diagnosis, treatment, and prevention of human diseases. However, clinical trials pose challenges and drawbacks, such as complexity and time-consuming processes, which create obstacles for many researchers. Graph Attention Network (GAT) has shown excellent performance in handling graph-structured data for tasks such as link prediction. Some studies have successfully applied GAT to miRNA-disease association prediction. However, there are several drawbacks to existing methods. Firstly, most of the previous models rely solely on concatenation operations to merge features of miRNAs and diseases, which results in the deprivation of significant modality-specific information and even the inclusion of redundant information. Secondly, as the number of layers in GAT increases, there is a possibility of excessive smoothing in the feature extraction process, which significantly affects the prediction accuracy. To address these issues and effectively complete miRNA disease prediction tasks, we propose an innovative model called Multiplex Adaptive Modality Fusion Graph Attention Network (MAMFGAT). MAMFGAT utilizes GAT as the main structure for feature aggregation and incorporates a multi-modal adaptive fusion module to extract features from three interconnected networks: the miRNA-disease association network, the miRNA similarity network, and the disease similarity network. It employs adaptive learning and cross-modality contrastive learning to fuse more effective miRNA and disease feature embeddings as well as incorporates multi-modal residual feature fusion to tackle the problem of excessive feature smoothing in GATs. Finally, we employ a Multi-Layer Perceptron (MLP) model that takes the embeddings of miRNA and disease features as input to anticipate the presence of potential miRNA-disease associations. Extensive experimental results provide evidence of the superior performance of MAMFGAT in comparison to other state-of-the-art methods. To validate the significance of various modalities and assess the efficacy of the designed modules, we performed an ablation analysis. Furthermore, MAMFGAT shows outstanding performance in three cancer case studies, indicating that it is a reliable method for studying the association between miRNA and diseases. The implementation of MAMFGAT can be accessed at the following GitHub repository: https://github.com/zixiaojin66/MAMFGAT-master.

A cascade-based dual-domain data correction network for sparse view CT image reconstruction

Computed tomography (CT) provides non-invasive anatomical structures of the human body and is also widely used for clinical diagnosis, but excessive ionizing radiation in X-rays can cause harm to the human body. Therefore, the researchers obtained sparse sinograms reconstructed sparse view CT images (SVCT) by reducing the amount of X-ray projection, thereby reducing the radiological effects caused by radiation. This paper proposes a cascade-based dual-domain data correction network (CDDCN), which can effectively combine the complementary information contained in the sinogram domain and the image domain to reconstruct high-quality CT images from sparse view sinograms. Specifically, several encoder-decoder subnets are cascaded in the sinogram domain to reconstruct artifact-free and noise-free CT images. In the encoder-decoder subnets, spatial-channel domain learning is designed to achieve efficient feature fusion through a group merging structure, providing continuous and elaborate pixel-level features and improving feature extraction efficiency. At the same time, to ensure that the original sinogram data collected can be retained, a sinogram data consistency layer is proposed to ensure the fidelity of the sinogram data. To further maintain the consistency between the reconstructed image and the reference image, a multi-level composite loss function is designed for regularization to compensate for excessive smoothing and distortion of the image caused by pixel loss and preserve image details and texture. Quantitative and qualitative analysis shows that CDDCN achieves competitive results in artifact removal, edge preservation, detail restoration, and visual improvement for sparsely sampled data under different views.