Jianxin Feng

Introduction: Early diagnosis of Alzheimer's disease (AD) remains challenging due to the high similarity among AD, mild cognitive impairment (MCI), and cognitively normal (CN) individuals, as well as confounding factors such as population heterogeneity, label noise, and variations in imaging acquisition. Although multimodal neuroimaging techniques like MRI and PET can provide complementary information, current approaches are limited in multimodal fusion and multi-scale feature aggregation.Methods: We propose a novel multimodal diagnostic framework, Alzheimer's Disease Multi-View Multimodal Diagnostic Network (ADMV-Net), to enhance recognition accuracy across all AD stages. Specifically, a dual-pathway Hybrid Convolution ResNet module is designed to fuse global semantic and local boundary information, enabling robust three-dimensional medical image feature extraction. Furthermore, a Multi-view Fusion Learning mechanism, which comprises a Global Perception Module, a Multi-level Local Cross-modal Aggregation Network, and a Bidirectional Cross-Attention Module, is introduced to efficiently capture and integrate multimodal features from multiple perspectives. Additionally, a Regional Interest Perception Module is incorporated to highlight brain regions strongly associated with AD pathology.Results: Extensive experiments on public datasets demonstrate that ADMV-Net achieves 94.83% accuracy and 95.97% AUC in AD versus CN classification, significantly outperforming mainstream methods. The framework also shows strong discriminative capability and excellent generalization performance in multi-class classification tasks.Discussion: These findings suggest that ADMV-Net effectively leverages multimodal and multi-view information to improve the diagnostic accuracy of AD. By integrating global, local, and regional features, the framework provides a promising tool for assisting early diagnosis and clinical decision-making in Alzheimer's disease. The implementation code is publicly available at https://github.com/zhaoxinyu-1/ADMV-Net.

As a form of discrete representation learning, Vector Quantized Variational Autoencoders (VQ-VAE) have increasingly been applied to generative and multimodal tasks due to their ease of embedding and representative capacity. However, existing VQ-VAEs often perform quantization in the spatial domain, ignoring global structural information and potentially suffering from codebook collapse and information coupling issues. This paper proposes a frequency quantized variational autoencoder (FQ-VAE) to address these issues. The proposed method transforms image features into linear combinations in the frequency domain using a 2D fast Fourier transform (2D-FFT) and performs adaptive quantization on these frequency components to preserve image’s global relationships. The codebook is dynamically optimized to avoid collapse and information coupling issue by considering the usage frequency and dependency of code vectors. Furthermore, we introduce a post-processing module based on graph convolutional networks to further improve reconstruction quality. Experimental results on four public datasets demonstrate that the proposed method outperforms state-of-the-art approaches in terms of Structural Similarity Index (SSIM), Learned Perceptual Image Patch Similarity (LPIPS), and Reconstruction Fréchet Inception Distance (rFID). In the experiments on the CIFAR-10 dataset, compared to the baseline method VQ-VAE, the proposed method improves the above metrics by 4.9%, 36.4%, and 52.8%, respectively.

Multivariate time series forecasting is widely used in traffic planning, weather forecasting, and energy consumption. Series decomposition algorithms can help models better understand the underlying patterns of the original series to improve the forecasting accuracy of multivariate time series. However, the decomposition kernel of previous decomposition-based models is fixed, and these models have not considered the differences in frequency fluctuations between components. These problems make it difficult to analyze the intricate temporal variations of real-world time series. In this paper, we propose a series decomposition-based Mamba model, DecMamba, to obtain the intricate temporal dependencies and the dependencies among different variables of multivariate time series. A variable-level adaptive kernel combination search module is designed to interact with information on different trends and periods between variables. Two backbone structures are proposed to emphasize the differences in frequency fluctuations of seasonal and trend components. Mamba with superior performance is used instead of a Transformer in backbone structures to capture the dependencies among different variables. A new embedding block is designed to capture the temporal features better, especially for the high-frequency seasonal component whose semantic information is difficult to acquire. A gating mechanism is introduced to the decoder in the seasonal backbone to improve the prediction accuracy. A comparison with ten state-of-the-art models on seven real-world datasets demonstrates that DecMamba can better model the temporal dependencies and the dependencies among different variables, guaranteeing better prediction performance for multivariate time series.

With the challenge posed by global warming, accurately estimating and managing carbon emissions becomes a key step for businesses, especially power generation companies, to reduce their environmental impact. Optuna–LightGBM–XGBoost, a novel power and carbon emission relationship model that aims to improve the efficiency of carbon emission monitoring and estimation for power generation companies, is proposed in this paper. Deeply exploring the intrinsic link between power production data and carbon emissions, this model paves a new path for “measuring carbon through electricity”, in contrast to the emission factor method commonly used in China. Unit data from power generation companies are processed into structured tabular data, and a parallel processing framework is constructed with LightGBM and XGBoost, and optimized with the Optuna algorithm. The multilayer perceptron (MLP) is used to fuse features to enhance prediction accuracy by capturing characters that the individual models cannot detect. Simulation results show that Optuna–LightGBM–XGBoost can achieve better performance compared to existing methods. The mean absolute error (MAE), mean squared error (MSE), mean absolute percentage error (MAPE), and coefficient of determination (R²) of the model are 0.652, 0.939, 0.136, and 0.994, respectively. This not only helps governments and enterprises to develop more scientific and reasonable emission reduction strategies and policies, but also lays a solid foundation for achieving global carbon neutrality goals.

We introduce an innovative deep learning approach specifically designed for the environment identification of intelligent vehicles under rainy conditions in this paper. In the construction of wireless vehicular communication networks, an innovative approach is proposed that incorporates additional multipath components to simulate the impact of raindrop scattering on the vehicle-to-vehicle (V2V) channel, thereby emulating the channel characteristics of vehicular environments under rainy conditions and an equalization strategy in OFDM-based systems is proposed at the receiver end to counteract channel distortion. Then, a rainy environment identification method for autonomous vehicles is proposed. The core of this method lies in utilizing the Channel State Information (CSI) shared within the vehicular network to accurately identify the diverse rainy environments in which the vehicle operates without relying on traditional sensors. The environmental identification task is considered as a multi-class classification problem and a dedicated Convolutional Neural Network (CNN) model is proposed. This CNN model uses the CSI estimated from CAM exchanged in vehicle-to-vehicle (V2V) communication as training features. Simulation results showed that our method achieved an accuracy rate of 95.7% in recognizing various rainy environments, which significantly surpasses existing classical classification models. Moreover, it only took microseconds to predict with high accuracy, surpassing the performance limitations of traditional sensing systems under adverse weather conditions. This breakthrough ensures that intelligent vehicles can rapidly and accurately adjust driving parameters even in complex weather conditions like rain to autonomous drive safely and reliably.

Traffic flow prediction is a crucial aspect of Intelligent Transport Systems, offering a scientific foundation for urban transport system management and planning. However, predicting traffic flow becomes challenging due to its susceptibility to diverse static and dynamic external factors, such as the presence of emergency vehicles that necessitate priority treatment in the road network. To tackle this issue, this paper introduces an Emergency Vehicle Priority Scheduling Model based on Heterogeneous Feature Fusion in Graph Convolutional Networks (EVHF-GCN). This model concurrently considers road network and emergency vehicle information, dynamically adjusting signal control strategies based on traffic flow prediction outcomes. This approach ensures the prioritized passage of emergency vehicles and mitigates traffic congestion. The model utilizes a heterogeneous feature fusion mechanism within a Graph Convolutional Network (GCN) to propagate features and aggregate information from intersection nodes. It also integrates a Gated Recurrent Unit (GRU) network to capture dynamic traffic flow features. Additionally, we propose a Dynamic Signal Control Strategy (DSCS) that determines intersection green light durations based on prediction results and selects different control strategies as per the situation. Experimental results demonstrate that the model enhances traffic flow prediction accuracy and improves traffic system efficiency and safety in scenarios with and without emergency vehicles.