Joint Loss Research Articles

View-Aware Contrastive Learning for Incomplete Tabular Data with Low-Label Regimes

To address the challenges of label sparsity and feature incompleteness in structured data, a self-supervised representation learning method based on multi-view consistency constraints is proposed in this paper. Robust modeling of high-dimensional sparse tabular data is achieved through integration of a view-disentangled encoder, intra- and cross-view contrastive mechanisms, and a joint loss optimization module. The proposed method incorporates feature clustering-based view partitioning, multi-view consistency alignment, and masked reconstruction mechanisms, thereby enhancing the model’s representational capacity and generalization performance under weak supervision. Across multiple experiments conducted on four types of datasets, including user rating data, platform activity logs, and financial transactions, the proposed approach maintains superior performance even under extreme conditions of up to 40% feature missingness and only 10% label availability. The model achieves an accuracy of 0.87, F1-score of 0.83, and AUC of 0.90 while reducing the normalized mean squared error to 0.066. These results significantly outperform mainstream baseline models such as XGBoost, TabTransformer, and VIME, demonstrating the proposed method’s robustness and broad applicability across diverse real-world tasks. The findings suggest that the proposed method offers an efficient and reliable paradigm for modeling sparse structured data.

KANDiff: Low-light Image Enhancement Based on Diffusion Models

Low-light image enhancement technique aims to improve the contrast and brightness of low-light images. Diffusion models, attributed to adeptness at capturing intricate details, have achieved good results in image enhancement, but there are problems such as inadequate estimation of noise characteristics and the emergence of color bias in the enhanced images. To address the aforementioned problems, this paper proposes a diffusion models-based method for low-light image enhancement, termed KANDiff. In the diffusion model architecture, this paper adds the nonlinear learnable activation function Kolmogorov–Arnold network to the noise estimation network U-Net to generate higher quality enhanced images. Additionally, KANDiff mitigates color bias in the enhanced images through a joint loss function and employs a patch-based image restoration strategy to significantly enhance the model generalization capability. The experimental results show that the KANDiff algorithm proposed in this paper can achieve high-quality image enhancement and achieve better enhancement effects compared to other algorithms.

Metabolomics and Network Pharmacology Analyses Reveal the Mechanism of Moxibustion in Knee Osteoarthritis.

This study aimed to explore the mechanism of moxibustion in the knee by combining osteoarthritis metabolomics and network pharmacology. A rat knee osteoarthritis (KOA) model was established by intra-articular injection of papain. The efficacy of moxibustion in KOA rats was evaluated by swelling degree, pathological progress, and mobility loss of knee joint. On this basis, the metabolic mechanism of moxibustion in relieving knee osteoarthritis was analyzed by metabolomics analysis. Moxibustion significantly reduced joint swelling and inflammation in the knee joint of KOA rats. Sixteen metabolites and nine metabolic pathways were found to be associated with the mechanism of action of moxibustion in metabolomics analysis results. According to network pharmacology, 3186 KOA disease targets, 158 drug targets, and 89 intersecting targets were obtained. The key targets included MAPK-3, AKT-1, RELA, MAPK-8, MAPK-14, etc. Signal pathways were found to be involved in mechanisms of moxibustion in knee osteoarthritis, such as alanine, aspartate, and glutamate metabolism, cysteine and methionine metabolism, and arginine and proline metabolism. The mechanism of moxibustion in knee osteoarthritis may involve alanine, aspartate, and glutamate metabolism, cysteine and methionine metabolism, arginine and proline metabolism, amino tRNA biosynthesis, and D-glutamine and D-glutamate metabolism signaling pathways with MAPK-3, AKT-1, RELA, MAPK-8, and MAPK-14 as core targets. More precise mechanisms need to be verified by further systematic molecular biology experiments.

BioSeq_Ksite: Multi-perspective feature-driven prediction of protein succinylation based on an adaptive attention module with SSBCE loss strategy.

BioSeq_Ksite: Multi-perspective feature-driven prediction of protein succinylation based on an adaptive attention module with SSBCE loss strategy.

Towards High-Quality MRI Reconstruction With Anisotropic Diffusion-Assisted Generative Adversarial Networks and Its Multi-Modal Images Extension.

Recently, fast Magnetic Resonance Imaging reconstruction technology has emerged as a promising way to improve the clinical diagnostic experience by significantly reducing scan times. While existing studies have used Generative Adversarial Networks to achieve impressive results in reconstructing MR images, they still suffer from challenges such as blurred zones/boundaries and abnormal spots caused by inevitable noise in the reconstruction process. To this end, we propose a novel deep framework termed Anisotropic Diffusion-Assisted Generative Adversarial Networks, which aims to maximally preserve valid high-frequency information and structural details while minimizing noises in reconstructed images by optimizing a joint loss function in a unified framework. In doing so, it enables more authentic and accurate MR image generation. To specifically handle unforeseeable noises, an Anisotropic Diffused Reconstruction Module is developed and added aside the backbone network as a denoise assistant, which improves the final image quality by minimizing reconstruction losses between targets and iteratively denoised generative outputs with no extra computational complexity during the testing phase. To make the most of valuable MRI data, we extend its application to support multi-modal learning to boost reconstructed image quality by aggregating more valid information from images of diverse modalities. Extensive experiments on public datasets show that the proposed framework can achieve superior performance in polishing up the quality of reconstructed MR images. For example, the proposed method obtains average PSNR and mSSIM values of 35.785 dB and 0.9765 on the MRNet dataset, which are at least about 2.9 dB and 0.07 higher than those from the baselines.

FISIOTERAPIA NA REABILITAÇÃO DA MARCHA E EQUILÍBRIO EM INDIVÍDUOS COM NEUROPATIA DIABÉTICA: UMA REVISÃO INTEGRATIVA

ABSTRACT Introduction: Individuals with type 2 Diabetes Mellitus (DM2) tend to present complications such as Peripheral Neuropathy, which causes deficits in vibratory sensitivity, proprioception, kinesthesia and tactile sensitivity that lead to loss of joint and movement perception. This loss of sensitivity becomes a factor that contributes to decreased balance, changes in gait and proprioception. Changes in gait make individuals with Diabetic Neuropathy more susceptible to falls. The objective of the present study was to integrate knowledge about interventions that have results on individuals with Diabetic Neuropathy. Material and Methods: This is an integrative review of the literature, including clinical trials published between 2013 and 2021. The Physiotherapy Evidence Database (PEDro) scale was used to assess the methodological quality of the articles. Nine studies were included and analyzed. Results and Discussion: a total of 184 articles were found, of which 162 were excluded, resulting in a total of 22 for inclusion in the analysis. Among the eligible studies after reading the abstract, 11 remained and after applying the PEdro scale, 09 were included. The clinical trial studies presented mainly addressed physiotherapeutic interventions to improve balance and gait, however some trials also showed benefits for other symptoms. such as pain, tenderness and muscle weakness. Conclusion: It was concluded that Physiotherapy for patients with Diabetic Peripheral Neuropathy had a significantly positive impact on balance results and other findings inherent to the condition studied. Keywords: Diabetic Neuropathies; Rehabilitation; Postural balance; Gait.

M2FNet: multi-modality multi-level fusion network for segmentation of acute and sub-acute ischemic stroke

Ischemic stroke, a leading cause of death and disability, necessitates accurate detection and automatic segmentation of lesions. While diffusion weight imaging is crucial, its single modality limits the detection of subtle lesions and artifacts. To address this, we propose a multi-modality, multi-level fusion network (M2FNet) that aggregates salient features from different modalities across various levels. Our method uses a multi-modal independent encoder to extract modality-specific features from images of different modalities, thereby preserving key details and ensuring rich features. In order to suppress noise while ensuring the best preservation of modality-specific information, we effectively integrate features of different modalities through a cross-modal encoder fusion module. In addition, a cross-modal decoder fusion module and a multi-modality joint loss are designed to further improve the fusion of high-level and low-level features in the up-sampling stage, dynamically utilizing complementary information from multiple modalities to improve segmentation accuracy. To verify the effectiveness of our proposed method, M2FNet was validated on two public magnetic resonance imaging ischemic stroke lesion segmentation benchmark datasets. Whether single or multi-modality, M2FNet performed better than ten other baseline methods. This highlights the effectiveness of M2FNet in multi-modality segmentation of ischemic stroke lesions, making it a promising and powerful quantitative analysis tool for rapid and accurate diagnostic support. The codes of M2FNet are available at https://github.com/ShannanChen/MMFNet.

Outcome of Pulley Release via a Radial Mid-Lateral Approach for the Trigger Finger at the A2 Pulley.

Background: Trigger finger occurs primarily at the A1 pulley. However, triggering of the flexor tendon can also occur at the A2 pulley. The full release of the A2 pulley should be avoided because it may cause bowstringing and flexion weakness. To reduce the incidence of postoperative complications, we have performed a complete release of the A1 pulley via a volar approach and a complete release of the A2 pulley at the radial edge via a radial mid-lateral approach. This study aimed to investigate the outcome of our procedure. Methods: Seventeen fingers underwent A2 pulley release at the radial edge via a mid-lateral approach after complete release of the A1 pulley via a volar approach. The mean age of patients was 68 years. Regarding the Quinnell classification, 15 fingers were grade IV, 1 finger was grade II with triggering at the A2 pulley level and 1 finger was grade I with a ganglion on the A1 pulley. Results: The mean extension loss of the PIP joint was 23° preoperatively and 2° at the final examination, and the respective mean tip-palmar distances were 19 mm and 3 mm. No symptoms or signs suggested the bowstring of the flexor tendon during the observation period (mean: 99 weeks; range: 19-180 weeks). Conclusions: Our procedure is preferable for managing the tenosynovitis at the A1 and A2 pulleys. The release of the A2 pulley at the radial edge via a radial mid-lateral approach yielded a favourable result without bowstringing. Level of Evidence: Level IV (Therapeutic).

A Generic Image Steganography Recognition Scheme with Big Data Matching and an Improved ResNet50 Deep Learning Network

Image steganalysis has been a key technology in information security in recent years. However, existing methods are mostly limited to the binary classification for detecting steganographic images used in digital watermarking, privacy protection, illicit data concealment, and security images, such as unaltered cover images or surveillance images. They cannot identify the steganography algorithms used in steganographic images, which restricts their practicality. To solve this problem, this paper proposes a general steganography algorithms recognition scheme based on image big data matching with improved ResNet50. The scheme first intercepts the image region with the highest complexity and focuses on the key features to improve the analysis efficiency; subsequently, the original image of the image to be detected is accurately located by the image big data matching technique and the steganographic difference feature image is generated; finally, the ResNet50 is improved by combining the pyramid attention mechanism and the joint loss function, which achieves the efficient recognition of the steganography algorithm. To verify the feasibility and effectiveness of the scheme, three experiments are designed in this paper: verification of the selection of the core analysis region, verification of the image similarity evaluation based on Peak Signal-to-Noise Ratio (PSNR), and performance verification of the improved ResNet50 model. The experimental results show that the scheme proposed in this paper outperforms the existing mainstream steganalysis models, such as ZhuNet and YeNet, with a detection accuracy of 96.11%, supports the recognition of six adaptive steganography algorithms, and adapts to the needs of analysis of multiple sizes and image formats, demonstrating excellent versatility and application value.

Optimizing AlexNet for accurate tree species classification via multi-branch architecture and mixed-domain attention

Accurate identification of tree species is essential for effective forestry management and conservation. Simple deep-learning models, such as AlexNet and VGG16, often struggle with fine-grained texture extraction and feature distinction, especially in complex environments. While more advanced models, such as ResNet34 and deeper architectures, offer superior feature extraction capabilities, they come with the trade-off of significantly longer training times and higher computational costs. To address these challenges in tree species classification, an optimized AlexNet architecture, MMCAlexNet, is proposed to address these challenges for tree species classification. The model integrates a multi-branch convolutional module, a mixeddomain attention module, and a joint loss function to improve feature extraction and class separation. The multi-branch convolutional module extracts diverse features by processing input with branches of different kernel sizes, capturing both fine and global details. The mixeddomain attention module enhances feature representation by focusing on critical spatial and channel-wise features. The joint loss function ensures better class separation and prevents overfitting by balancing classification accuracy and feature consistency. Experimental results demonstrate that MMCAlexNet outperforms the baseline AlexNet by 7.61%–9.69% in classification accuracy. Although the introduction of optimized structures increases computational complexity, it reduces the model size by 12MB. Furthermore, MMCAlexNet decreases the model size by 306.24MB and improves accuracy by 3.94%–6.35% compared to VGG16, demonstrating a balance between improved accuracy and computational efficiency.

Ice Thickness Detection of Transmission Lines Based on Cross-Guide-UNet

The presence of ice on transmission lines threatens their integrity. The thickness of ice is crucial with respect to the autonomous operation of de-icing robots. Therefore, we propose a Cross-Guide-UNet (CG-UNet) model for detecting the thickness of ice on ice-covered transmission lines. The CG-UNet model uses an encoder–decoder architecture and incorporates a cross-guide module (CGM) at the crosslayer connection of the model. Initially, we redesign the attention calculation method to suit the shape characteristics of transmission lines. Channel and spatial attention are computed for the encoder and decoder features, respectively. Subsequently, the attention weight is cross-applied to the encoder and decoder features, allowing for comprehensive utilization of feature information to guide the generation of ice-covered edges. Additionally, we introduce a joint loss function comprising cross-entropy loss and boundary loss for model training, enhancing the distinction of easily confused edge pixels. The CG-UNet model effectively detects ice-covered edges through experimental analysis, achieving optimal dataset scale (ODS) and optimal image scale (OIS) scores of 0.934 and 0.938, respectively. The thickness detection error is constrained within 7.2%. These results confirm that the proposed algorithm can provide a control basis for the high-precision operation of de-icing robots.

Boundary-enhanced cascaded network for joint optic disc and optic cup semantic segmentation

Clinically, calculating the Cup-to-Disc Ratio (CDR) from fundus images is a straightforward method to screen and diagnose glaucoma. However, the accurate calculation of CDR heavily depends on precise segmentation of the Optic Disc (OD) and Optic Cup (OC). Most U-net-based OD and OC segmentation methods have achieved good results, but these approaches may lead to outliers, poor integrity, inaccurate and unsmooth contours. To alleviate these issues, we propose a Boundary-Enhanced Cascaded Deep Convolutional Neural Network (BEC-DCNN) for robust and accurate OD and OC segmentation in fundus images, which is followed by a two-stage framework. The first stage employs the DeepLabv3+ network with attention mechanism to detect the OD. In the second stage, we integrate a boundary auxiliary branch, which can obtain multi-scale contour feature maps from the backbone network to produce smoother OD and OC segmentation results. This paper designs a simple and efficient joint loss in training to constrain both the segmentation branch and the boundary branch, which is helpful to improve the segmentation precision and the smoothness of the boundary. Our experiments conducted on the three public datasets, Rim-One-r3, Drishti-GS, and REFUGE, demonstrate that our proposed BEC-DCNN outperforms many previous algorithms. Through qualitative and quantitative comparisons with other methods, our proposal produces more accurate and smoother contours. Our method is also applicable to other medical image segmentation tasks that require accurate edges.

DDPM-EMF: a denoising diffusion probabilistic model-based feature-enhancement fusion network for medical image fusion

Medical image fusion is crucial in clinical applications, combining data from various medical imaging modalities into a single high-quality image to enhance diagnosis. However, existing fusion algorithms suffer from several limitations, including inadequate feature extraction, leading to detail loss, poor inheritance of complementary information between modalities, and insufficient evaluation of color information in color and grayscale fusion tasks. To address these challenges, we propose a novel, to our knowledge, medical image fusion framework based on the denoising diffusion probabilistic model. Our model adopts a two-stage training strategy: feature extraction and image reconstruction. An edge-enhancement dense block is designed to work with a denoising diffusion probabilistic model as a feature extractor, learning and extracting joint features from multimodal medical images to ensure comprehensive feature extraction. To further integrate meaningful information and enhance the visual quality of fused images, we design a feature-enhanced reconstruction network that amplifies features during the reconstruction process. Additionally, we develop distinct joint loss functions based on the strengths and weaknesses of different modalities, ensuring effective retention of complementary information. In the color and grayscale fusion task, we introduce a multi-channel joint learning method to ensure the retention of complementary information and incorporate a color difference formula to evaluate color retention. Experimental results demonstrate that our proposed method significantly outperforms existing state-of-the-art techniques, producing fused images with improved clarity, enhanced detail preservation, and more effective inheritance of complementary information across modalities.

Mutual learning for domain adaptation: Self-distillation image dehazing network with sample-cycle

Mutual learning for domain adaptation: Self-distillation image dehazing network with sample-cycle

MCTASmRNA: A deep learning framework for alternative splicing events classification.

MCTASmRNA: A deep learning framework for alternative splicing events classification.

Multi-center brain age prediction via dual-modality fusion convolutional network.

Multi-center brain age prediction via dual-modality fusion convolutional network.

Fine-scale striatal parcellation using diffusion MRI tractography and graph neural networks.

Fine-scale striatal parcellation using diffusion MRI tractography and graph neural networks.

TD-STrans: Tri-domain sparse-view CT reconstruction based on sparse transformer.

TD-STrans: Tri-domain sparse-view CT reconstruction based on sparse transformer.

Optimized UNet framework with a joint loss function for underwater image enhancement

As the water economy advances and the concepts of water ecology protection and sustainable development take root in people’s minds, underwater imaging equipment has made remarkable progress. However, due to various factors, underwater images still suffer from low quality. How to enhance the quality of underwater images so that people can understand them quickly has become a crucial issue. Therefore, aiming at the degradation problems such as detail blurring, color imbalance, and noise interference in low-quality underwater images, this paper proposes an optimized UNet framework with a joint loss function (OUNet-JL). Firstly, to alleviate the problem of detail blurring, we construct a multi-residual module (MRM) to enhance the ability to represent detail features by using serially stacked convolutional blocks and residual connections. Secondly, we build a spatial multi-scale feature extraction module fused with channel attention (SMFM) to address the color imbalance issue through multi-scale dilated convolution and channel attention. Thirdly, to improve the signal-to-noise ratio of the enhanced image and solve the problem of blurring distortion, a strengthen-operate-subtract feature reconstruction module (SOSFM) is presented. Fourthly, to guide the network to perform training more efficiently and help it converge rapidly, a joint loss function is designed by integrating four different loss functions. Extensive experiments conducted on the well-known UIEB and UFO-120 datasets have shown the superiority of our OUNet-JL compared with several state-of-the-art algorithms. Moreover, ablation studies have also verified the effectiveness of the proposed modules. Our source code is publicly available at https://github.com/WangXin81/OUNet_JL.

Multitask Learning-Based Facial Landmark Detection via Signal Processing: Eliminating Artificial Auxiliary Labeling for Autonomous Driving

This study introduces a novel facial landmark detection framework that leverages signal processing techniques to eliminate the dependency on artificial auxiliary labeling. Built upon multi-task learning, the proposed method constructs auxiliary information directly from labeled facial landmarks, including inter-landmark connection distances and facial triangle angle features. By incorporating multi-task collaborative optimization, the method enforces explicit constraints on global face shape and enhances feature representation, significantly improving detection accuracy. A tailored joint loss function integrates geometric features and addresses data distribution imbalance, ensuring robust penalty for rare samples while further constraining global face geometry. Experimental evaluations on benchmark datasets, including 300W and AFLW, demonstrate that the proposed approach outperforms the existing state-of-the-art methods in terms of accuracy and reliability. Moreover, our study reveals that efficient facial landmark detection can be achieved without artificial auxiliary annotation information through a multi-task learning framework. These results establish the efficacy of the method for practical applications, such as autonomous driving systems.