CAAI Transactions on Intelligence Technology最新文献

Acute Bilirubin Encephalopathy (ABE) is a significant threat to neonates and it leads to disability and high mortality rates. Detecting and treating ABE promptly is important to prevent further complications and long-term issues. Recent studies have explored ABE diagnosis. However, they often face limitations in classification due to reliance on a single modality of Magnetic Resonance Imaging (MRI). To tackle this problem, the authors propose a Tri-M2MT model for precise ABE detection by using tri-modality MRI scans. The scans include T1-weighted imaging (T1WI), T2-weighted imaging (T2WI), and apparent diffusion coefficient maps to get indepth information. Initially, the tri-modality MRI scans are collected and preprocessesed by using an Advanced Gaussian Filter for noise reduction and Z-score normalisation for data standardisation. An Advanced Capsule Network was utilised to extract relevant features by using Snake Optimization Algorithm to select optimal features based on feature correlation with the aim of minimising complexity and enhancing detection accuracy. Furthermore, a multi-transformer approach was used for feature fusion and identify feature correlations effectively. Finally, accurate ABE diagnosis is achieved through the utilisation of a SoftMax layer. The performance of the proposed Tri-M2MT model is evaluated across various metrics, including accuracy, specificity, sensitivity, F1-score, and ROC curve analysis, and the proposed methodology provides better performance compared to existing methodologies.

In the era of big data, personalised recommendation systems are essential for enhancing user engagement and driving business growth. However, traditional recommendation algorithms, such as collaborative filtering, face significant challenges due to data sparsity, algorithm scalability, and the difficulty of adapting to dynamic user preferences. These limitations hinder the ability of systems to provide highly accurate and personalised recommendations. To address these challenges, this paper proposes a clustering-based recommendation method that integrates an enhanced Grasshopper Optimisation Algorithm (GOA), termed LCGOA, to improve the accuracy and efficiency of recommendation systems by optimising cluster centroids in a dynamic environment. By combining the K-means algorithm with the enhanced GOA, which incorporates a Lévy flight mechanism and multi-strategy co-evolution, our method overcomes the centroid sensitivity issue, a key limitation in traditional clustering techniques. Experimental results across multiple datasets show that the proposed LCGOA-based method significantly outperforms conventional recommendation algorithms in terms of recommendation accuracy, offering more relevant content to users and driving greater customer satisfaction and business growth.

Internal learning-based video inpainting methods have shown promising results by exploiting the intrinsic properties of the video to fill in the missing region without external dataset supervision. However, existing internal learning-based video inpainting methods would produce inconsistent structures or blurry textures due to the insufficient utilisation of motion priors within the video sequence. In this paper, the authors propose a new internal learning-based video inpainting model called appearance consistency and motion coherence network (ACMC-Net), which can not only learn the recurrence of appearance prior but can also capture motion coherence prior to improve the quality of the inpainting results. In ACMC-Net, a transformer-based appearance network is developed to capture global context information within the video frame for representing appearance consistency accurately. Additionally, a novel motion coherence learning scheme is proposed to learn the motion prior in a video sequence effectively. Finally, the learnt internal appearance consistency and motion coherence are implicitly propagated to the missing regions to achieve inpainting well. Extensive experiments conducted on the DAVIS dataset show that the proposed model obtains the superior performance in terms of quantitative measurements and produces more visually plausible results compared with the state-of-the-art methods.

Skin cancer, a severe health threat, can spread rapidly if undetected. Therefore, early detection can lead to an advanced and efficient diagnosis, thus reducing mortality. Unsupervised classification techniques analyse extensive skin image datasets, identifying patterns and anomalies without prior labelling, facilitating early detection and effective diagnosis and potentially saving lives. In this study, the authors aim to explore the potential of unsupervised learning methods in classifying different types of skin lesions in dermatoscopic images. The authors aim to bridge the gap in dermatological research by introducing innovative techniques that enhance image quality and improve feature extraction. To achieve this, enhanced super-resolution generative adversarial networks (ESRGAN) was fine-tuned to strengthen the resolution of skin lesion images, making critical features more visible. The authors extracted histogram features to capture essential colour characteristics and used the Davies–Bouldin index and silhouette score to determine optimal clusters. Fine-tuned k-means clustering with Euclidean distance in the histogram feature space achieved 87.77% and 90.5% test accuracies on the ISIC2019 and HAM10000 datasets, respectively. The unsupervised approach effectively categorises skin lesions, indicating that unsupervised learning can significantly advance dermatology by enabling early detection and classification without extensive manual annotation.

Heterogeneous graphs generally refer to graphs with different types of nodes and edges. A common approach for extracting useful information from heterogeneous graphs is to use meta-graphs, which can be seen as a special kind of directed acyclic graph with same node and edge types as the heterogeneous graph. However, how to design proper meta-graphs is challenging. Recently, there have been many works on learning suitable meta-graphs from a heterogeneous graph. Existing methods generally introduce continuous weights for edges that are independent of each other, which ignores the topological structures of meta-graphs and can be ineffective. To address this issue, the authors propose a new viewpoint from tensor on learning meta-graphs. Such a viewpoint not only helps interpret the limitation of existing works by CANDECOMP/PARAFAC (CP) decomposition, but also inspires us to propose a topology-aware tensor decomposition, called TENSUS, that reflects the structure of DAGs. The proposed topology-aware tensor decomposition is easy to use and simple to implement, and it can be taken as a plug-in part to upgrade many existing works, including node classification and recommendation on heterogeneous graphs. Experimental results on different tasks demonstrate that the proposed method can significantly improve the state-of-the-arts for all these tasks.

Convolutional neural network (CNN) with the encoder–decoder structure is popular in medical image segmentation due to its excellent local feature extraction ability but it faces limitations in capturing the global feature. The transformer can extract the global information well but adapting it to small medical datasets is challenging and its computational complexity can be heavy. In this work, a serial and parallel network is proposed for the accurate 3D medical image segmentation by combining CNN and transformer and promoting feature interactions across various semantic levels. The core components of the proposed method include the cross window self-attention based transformer (CWST) and multi-scale local enhanced (MLE) modules. The CWST module enhances the global context understanding by partitioning 3D images into non-overlapping windows and calculating sparse global attention between windows. The MLE module selectively fuses features by computing the voxel attention between different branch features, and uses convolution to strengthen the dense local information. The experiments on the prostate, atrium, and pancreas MR/CT image datasets consistently demonstrate the advantage of the proposed method over six popular segmentation models in both qualitative evaluation and quantitative indexes such as dice similarity coefficient, Intersection over Union, 95% Hausdorff distance and average symmetric surface distance.

Efficient detection of surface defects is primary for ensuring product quality during manufacturing processes. To enhance the performance of deep learning-based methods in practical applications, the authors propose Dense-YOLO, a fast surface defect detection network that combines the strengths of DenseNet and you only look once version 3 (YOLOv3). The authors design a lightweight backbone network with improved densely connected blocks, optimising the utilisation of shallow features while maintaining high detection speeds. Additionally, the authors refine the feature pyramid network of YOLOv3 to increase the recall of tiny defects and overall positioning accuracy. Furthermore, an online multi-angle template matching technique is introduced based on normalised cross-correlation to precisely locate the detection area. This refined template matching method not only accelerates detection speed but also mitigates the influence of the background. To validate the effectiveness of our enhancements, the authors conduct comparative experiments across two private datasets and one public dataset. Results show that Dense-YOLO outperforms existing methods, such as faster R-CNN, YOLOv3, YOLOv5s, YOLOv7, and SSD, in terms of mean average precision (mAP) and detection speed. Moreover, Dense-YOLO outperforms networks inherited from VGG and ResNet, including improved faster R-CNN, FCOS, M2Det-320 and FRCN, in mAP.

Person Image Synthesis has been widely used in fashion with extensive application scenarios. The point of this task is how to synthesise person image from a single source image under arbitrary poses. Prior methods generate the person image with target pose well; however, they fail to preserve the fine style details of the source image. To address this problem, a robust style injection (RSI) model is proposed, which is a coarse-to-fine framework to synthesise target the person image. RSI develops a simple and efficient cross-attention based module to fuse the features of both source semantic styles and target pose for achieving the coarse aligned features. The adaptive instance normalisation is employed to enhance the aligned features in conjunction with source semantic styles. Subsequently, source semantic styles are further injected into the positional normalisation scheme to avoid the fine style details erosion caused by massive convolution. In training losses, optimal transport theory in the form of energy distance is introduced to constrain data distribution to refine the texture style details. Additionally, the authors’ model is capable of editing the shape and texture of garments to the target style separately. The experiments demonstrate that the authors’ RSI achieves better performance over the state-of-art methods.

Active Disturbance Rejection Control (ADRC) possesses robust disturbance rejection capabilities, making it well-suited for longitudinal velocity control. However, the conventional Extended State Observer (ESO) in ADRC fails to fully exploit feedback from first-order and higher-order estimation errors and tracking error simultaneously, thereby diminishing the control performance of ADRC. To address this limitation, an enhanced car-following algorithm utilising ADRC is proposed, which integrates the improved ESO with a feedback controller. In comparison to the conventional ESO, the enhanced version effectively utilises multi-order estimation and tracking errors. Specifically, it enhances convergence rates by incorporating feedback from higher-order estimation errors and ensures the estimated value converges to the reference value by utilising tracking error feedback. The improved ESO significantly enhances the disturbance rejection performance of ADRC. Finally, the effectiveness of the proposed algorithm is validated through the Lyapunov approach and experiments.

The goal of point cloud completion is to reconstruct raw scanned point clouds acquired from incomplete observations due to occlusion and restricted viewpoints. Numerous methods use a partial-to-complete framework, directly predicting missing components via global characteristics extracted from incomplete inputs. However, this makes detail recovery challenging, as global characteristics fail to provide complete missing component specifics. A new point cloud completion method named Point-PC is proposed. A memory network and a causal inference model are separately designed to introduce shape priors and select absent shape information as supplementary geometric factors for aiding completion. Concretely, a memory mechanism is proposed to store complete shape features and their associated shapes in a key-value format. The authors design a pre-training strategy that uses contrastive learning to map incomplete shape features into the complete shape feature domain, enabling retrieval of analogous shapes from incomplete inputs. In addition, the authors employ backdoor adjustment to eliminate confounders, which are shape prior components sharing identical semantic structures with incomplete inputs. Experiments conducted on three datasets show that our method achieves superior performance compared to state-of-the-art approaches. The code for Point-PC can be accessed by https://github.com/bizbard/Point-PC.git.