Traditional recommendation algorithms usually rely on the user's existing data and historical behavioral records to make recommendations, which often leads to low recommendation accuracy and insufficient personalized experience. To solve these problems, this paper proposes an innovative recommendation algorithm model, neural collaborative filtering with multiple attention mechanism (NCF-MAH). The goal of this model is to enhance the effectiveness of the recommender system. The specific implementation includes constructing a negative sample set and applying matrix decomposition techniques to map user and item IDs to a low-dimensional embedding vector space. In addition, the model processes these embedding vectors using a multi-head attention mechanism to transform them into query vectors, key vectors, and value vectors, and further computes the attention scores and the corresponding weighted sums. Finally, the score prediction is accomplished by fusing the output of the multi-head attention mechanism with the results of the multilayer perceptual machine. The experimental results show that the NCF-MAH model exhibits significant advantages over the baseline model in two key evaluation metrics, hit rate and normalized discount cumulative gain (NDCG), on the MOOC platform and other datasets. Specifically, hit rate and NDCG improved by 13% vs. 9.8% and 15.7% vs. 12.8% when Top-k was set to 10 and 20, respectively.

All-in-One Degradation-Aware Fusion Models (ADFMs) as one of multi-modal image fusion models, which aims to address complex scenes by mitigating degradations from source images and generating high-quality fused images. Mainstream ADFMs rely on end-to-end learning and heavily synthesized datasets to achieve degradation awareness and fusion. This rough learning strategy and non-real world scenario dataset dependence often limit their upper-bound performance, leading to low-quality results. To address these limitations, we present LURE, a Learning-driven Unified REpresentation model for infrared and visible image fusion, which is degradation-aware. LURE learns a Unified Latent Feature Space (ULFS) to avoid the dependency on complex data formats inherent in previous end-to-end learning pipelines. It further improves image fusion quality by leveraging the intrinsic relationships between multi-modalities. A novel loss function is also proposed to drive the learning of unified latent representations more stable.More importantly, LURE seamlessly incorporates existing high-quality real-world image restoration datasets. To further enhance the model's representation capability, we design a simple yet effective structure, termed internal residual block, to facilitate the learning of latent features. Experiments show our method outperforms state-of-the-art (SOTA) methods across general fusion, degradation-aware fusion, and downstream tasks. The code is available in the supplementary materials.

To effectively solve the problem of a small proportion of substantia nigra and midbrain regions in Magnetic Resolution Imaging (MRI) images of Parkinson's disease (PD) patients, unclear boundaries with surrounding tissues, and difficulty in accurately delineating boundaries, an improved U-2-Net algorithm for Parkinson's substantia nigra midbrain image segmentation was proposed. This algorithm first improves the Residual U-blocks (RSU) and RSU-4F modules using the Shuffle Attention (SA) module, enhancing the network's attention to Parkinson's substantia nigra and midbrain blurry regions in the encoding and decoding layers. Next, replace some ordinary convolutions in RSU-4F with DynamicConv (DyConv), capture local features through dynamic convolution kernels, and improve the segmentation performance of Parkinson's images. The experimental data used clinical data, and the improved algorithm achieved Dice, Precision, Recall, and mIou of 80.27%, 89.99%, 88.88%, and 82.38% in substantia nigra segmentation, respectively. The experimental results show that this algorithm can achieve more precise segmentation of Parkinson's images.

Generative models are widely utilized to model the distribution of fused images in the field of infrared and visible image fusion. However, current generative models based fusion methods often suffer from unstable training and slow inference speed. To tackle this problem, a novel fusion method based on consistency model is proposed, termed as CoMoFusion, which can generate the high-quality images and achieve fast image inference speed. In specific, the consistency model is used to construct multi-modal joint features in the latent space with the forward and reverse process. Then, the infrared and visible features extracted by the trained consistency model are fed into fusion module to generate the final fused image. In order to enhance the texture and salient information of fused images, a novel loss based on pixel value selection is also designed. Extensive experiments on public datasets illustrate that our method obtains the SOTA fusion performance compared with the existing fusion methods.

Image fusion aims to integrate complementary information from different images to provide richer scene details. However, in real-world scenarios, low-light illumination conditions not only affect the brightness, contrast, and color of visible images but also impact the quality of fusion results. Therefore, this paper proposes an illumination-free infrared and visible image fusion method, named IFFusion. Specifically, the whole task is considered as a multitasking problem: fusion and luminance adjustment. Firstly, based on the Retinex theory, the visible image can be decoupled into reflectance and illumination components, with reflectance should be consistent under different natural lighting conditions. Therefore, we encourage the network to learn the prior of reflectance consistency from pairs of visible images. Secondly, to obtain a fused image with appropriate brightness, a luminance perception network (LP-Net) is designed to perceive the scene brightness of the input visible image. Here, we use the visible image with normal brightness to guide LP-Net for adaptively adjusting the illumination component. Extensive experiments on public datasets show that the proposed method has better performance than the state-of-the-art fusion methods. Last but not least, we performed downstream task experiments on object detection and the improved performance verified the effectiveness of IFFusion.

In steady-state visual evoked potential (SSVEP)-based brain-computer interface (BCI), improving the recognition performance for new subjects without calibration data is the key challenge for practical application. Unsupervised transfer learning is an effective way to overcome it. However, existing studies focus solely on what to transfer, rather than how to effectively transfer, resulting in unsatisfactory transfer effectiveness or even negative transfer. In this study, an innovative unsupervised cross-subject transfer learning method for SSVEP-BCI was proposed, named SUTL. It involves that subject transferability estimation (STE) and a multi-domain alignment method were proposed to alleviate the potential interference of differences in SSVEP signal distribution among subjects. STE screens appropriate transferable subjects from the source subject pool, while domain alignment directly makes all subjects more similar. Then, SUTL sufficiently exploits the information of the selected source subjects, learning and transferring both generalization knowledge and subject-specific knowledge to boost the recognition performance for the new subject. The performance of SUTL was evaluated on two public SSVEP datasets (benchmark dataset and BETA dataset) with 40 classes, the extensive experimental results reveal that SUTL markedly boosts the effectiveness of SSVEP cross-subject transfer and dramatically outperforms the state-of-art methods. SUTL significantly enhances the recognition performance of SSVEP-BCI for new subjects and facilitates its practical application.

Multi-focus image fusion (MFIF) explores the positioning and reorganization of the focused parts from the input images. Focused and defocused parts have similar representations in color, contour and other appearance information, which degrades the fusion quality due to the influence of these redundant information. Currently, most MFIF methods have not identified an effective way to remove redundant information before fusion stage. Thus, in this paper, we introduce a structural map extraction strategy for multi-focus image fusion. Compared to the source image, structural map reduces redundant information, and the clearer parts of the image retain more abundant structural features. Consequently, the differences between focused part and defocused part become more pronounced based on the extracted structural map. Specifically, the proposed fusion method adopts a two-stage training strategy. Firstly, the structural map is extracted by the proposed structural map extraction network (SMENet) from the source images. Secondly, the structural map is thus applied to train the decision map generation network (DMGNet) to obtain the decision map which is utilized to generate the final fusion image. Qualitative and quantitative experiments on three public datasets demonstrate the superiority of the proposed method, compared with the advanced image fusion algorithms.

Due to the increasing scarcity of tea pickers, the implementation of intelligent harvesting for premium tea is a crucial prerequisite for the sustainable development of the premium tea industry. The initial step towards achieving intelligent and precise harvesting is the accurate detection of tender shoots, which consist of one bud and one leaf. However, accurately identifying tea shoots poses a challenging visual task due to their small size, variable shapes, as well as similar colors and backgrounds. The existing model, based on RGB images, can only detect partial targets. To address this issue and further enhance the detection of tea buds, this study proposes the utilization of multi-modal features encompassing red, green, blue, and depth (RGB-D) for identification. In addition, a unidirectional complementary multi-modal fusion method is introduced to minimize the adverse effects caused by low-quality depth information. Firstly, an RGB-D dataset comprising high-quality tea leaves is constructed, and the samples are carefully calibrated. Subsequently, an enhanced end-to-end RGB-D multi-modal object detection network, referred to as YOLO-RGBDtea, is developed based on You Only Look Once version 7 (YOLOv7). This model incorporates a parallel lightweight depth image feature extraction backbone network and incorporates a self-attention mechanism to prioritize contextual information. Lastly, a cross-modal spatial attention fusion module (CSFM) is devised to collaboratively integrate depth features with RGB features in a unidirectional manner. The experimental results reveal that YOLO-RGBDtea achieves an AP50 of 91.12% when confronting complex outdoor tea shoots, exhibiting significant performance improvements compared to YOLOv7, especially in scenarios involving small targets, overlapping target groups, and highly overexposed images. Notably, the parameter increment in YOLO-RGBDtea compared to the original YOLOv7 model is merely 17.8%, and the additional components can be seamlessly transferred to other models. Overall, this study introduces a straightforward yet effective multi-modal fusion method that bears theoretical and practical significance in advancing the detection of high-quality tea shoots in complex outdoor environments.

Many industrial experiments involve factors with levels more difficult to change or control than others, which leads to the development of two-level fractional factorial split-plot (FFSP) designs. Recently, mixed-level FFSP designs were proposed due to the requirement of different-level factors. In this paper, we generalize the Bayesian optimal criterion for mixed two- and four-level FFSP designs, and then provide Bayesian minimum aberration (MA) criterion to rank FFSP designs. Bayesian MA criterion can give a natural ordering for the effects involving two-level factors and three components of a four-level factor. We also discuss the relationship between the Bayesian optimal and Bayesian MA criteria. Furthermore, we consider the designs with both qualitative and quantitative factors.

Background 18 F-FDG positron emission tomography (PET) plays a crucial part in the evaluation for pediatric epileptic patients prior to therapy. Short-term scanning holds significant importance, especially for pediatrics epileptic individuals who exhibited involuntary movements. The aim was to evaluate the effects of short acquisition time on image quality and lesion detectability in pediatric epileptic patients using total-body (TB) PET/CT. A total of 25 pediatric patients who underwent TB PET/CT using uEXPLORER scanner with an 18 F-FDG administered dose of 3.7 MBq/kg and an acquisition time of 600 s were retrospectively enrolled. Short acquisition times (60 s, 150 and 300 s) were simulated by truncating PET data in list mode to reduce count density. Subjective image quality was scored on a 5-point scale. Regions of interest analysis of suspected epileptogenic zones (EZs), corresponding locations contralateral to EZs, and healthy cerebellar cortex were used to compare the semi-quantitative uptake indices of short-time images and then were compared with 600 s images. The comparison of EZs detectability based on time-dependent PET images was performed. Results Our study demonstrated that a short acquisition time of 150 s is sufficient to maintain subjective image quality and lesion significance. Statistical analysis revealed no significant difference in subjective PET image quality between imaging at 300 s and 150 s ( P > 0.05). The overall impression scores of image quality and lesion conspicuity in G60s were both greater than 3 (overall quality, 3.21 ± 0.46; lesion conspicuity, 4.08 ± 0.74). As acquisition time decreased, the changes of SUVmax and SD in the cerebellar cortex gradually increased ( P < 0.01). There was no significant difference in asymmetry index (AI) difference between the groups and the AIs of EZs were > 15% in all groups. In 26 EZs of 25 patients, the lesion detection rate was still 100% when the time was reduced to 60 s. Conclusions This study proposed that TB PET/CT acquisition time could be reduced to 60 s with acceptable lesion detectability. Furthermore, it was suggested that a 150 s acquisition time would be sufficient to achieve diagnostic performance and image quality for children with epilepsy.

The fusion of visible light and infrared images has garnered significant attention in the field of imaging due to its pivotal role in various applications, including surveillance, remote sensing, and medical imaging. Therefore, this paper introduces a novel fusion framework using Res2Net architecture, capturing features across diverse receptive fields and scales for effective extraction of global and local features. Our methodology is structured into three fundamental components: the first part involves the Res2Net-based encoder, followed by the second part, which encompasses the fusion layer, and finally, the third part, which comprises the decoder. The encoder based on Res2Net is utilized for extracting multi-scale features from the input image. Simultaneously, with a single image as input, we introduce a pioneering training strategy tailored for a Res2Net-based encoder. We further enhance the fusion process with a novel strategy based on the attention model, ensuring precise reconstruction by the decoder for the fused image. Experimental results unequivocally showcase our method’s unparalleled fusion performance, surpassing existing techniques, as evidenced by rigorous subjective and objective evaluations.

Background:Ischemic stroke (IS) is one of the most common serious secondary diseases of atrial fibrillation (AF) within 1 year after its occurrence, both of which have manifestations of ischemia and hypoxia of the small vessels in the early phase of the condition. The fundus is a collection of capillaries, while the retina responds differently to light of different wavelengths. Predicting the risk of IS occurring secondary to AF, based on subtle differences in fundus images of different wavelengths, is yet to be explored. This study was conducted to predict the risk of IS occurring secondary to AF based on multi-spectrum fundus images using deep learning. Methods:A total of 150 AF participants without suffering from IS within 1 year after discharge and 100 IS participants with persistent arrhythmia symptoms or a history of AF diagnosis in the last year (defined as patients who would develop IS within 1 year after AF, based on fundus pathological manifestations generally prior to symptoms of the brain) were recruited. Fundus images at 548, 605, and 810 nm wavelengths were collected. Three classical deep neural network (DNN) models (Inception V3, ResNet50, SE50) were trained. Sociodemographic and selected routine clinical data were obtained. Results:The accuracy of all DNNs with the single-spectral or multi-spectral combination images at the three wavelengths as input reached above 78%. The IS detection performance of DNNs with 605 nm spectral images as input was relatively more stable than with the other wavelengths. The multi-spectral combination models acquired a higher area under the curve (AUC) scores than the single-spectral models. Conclusions:The probability of IS secondary to AF could be predicted based on multi-spectrum fundus images using deep learning, and combinations of multi-spectrum images improved the performance of DNNs. Acquiring different spectral fundus images is advantageous for the early prevention of cardiovascular and cerebrovascular diseases. The method in this study is a beneficial preliminary and initiative exploration for diseases that are difficult to predict the onset time such as IS.

OBJECTIVE:Improving the Information Transfer Rate (ITR) is a popular research topic in steady-state visual evoked potential (SSVEP)-based brain-computer interfaces (BCIs). The higher recognition accuracy of short-time SSVEP signal is critical to improving ITR and achieving high-speed SSVEP-BCIs. However, the existing algorithms have unsatisfactory performance on recognizing short-time SSVEP signals, especially for calibration-free methods.METHOD:This study for the first time proposed improving the recognition accuracy of short-time SSVEP signals based on the calibration-free method by extending the SSVEP signal length. A signal extension model based on Multi-channel adaptive Fourier decomposition with different Phase (DP-MAFD) is proposed to achieve signal extension. Then the Canonical Correlation Analysis based on signal extension (SE-CCA) is proposed to complete the recognition and classification of SSVEP signals after extension.RESULT:The similarity study and SNR comparison analysis on public SSVEP datasets demonstrate that the proposed signal extension model has the ability to extend SSVEP signals. The classification results show that the proposed method outperforms Canonical Correlation Analysis (CCA) and Filter Bank Canonical Correlation Analysis (FBCCA) significantly in the measure of classification accuracy and information transmission rate (ITR), especially for short-time signals. The highest ITR of SE-CCA is improved to 175.61 bits/min at around 1s, while CCA is 100.55 bits/min at 1.75s and FBCCA is 141.76 bits/min at 1.25s.CONCLUSION:The signal extension method can improve the recognition accuracy of short-time SSVEP signals and further improve the ITR of SSVEP-BCIs.

In image fusion tasks, an ideal image decomposition method can bring better performance. MDLatLRR has done a great job in this aspect, but there is still exist some space for improvement. Considering that MDLatLRR focuses solely on the detailed parts (salient features) extracted from input images via latent low-rank representation (LatLRR), the basic parts (principal features) extracted by LatLRR are not fully utilized. Therefore, we introduced an enhanced multi-level decomposition method named dual-decomposed MDLatLRR (D2-LRR) which effectively analyzes and utilizes all image features extracted through LatLRR. Specifically, color images are converted into YUV color space and grayscale images, and the Y-channel and grayscale images are input into the trained parameters of LatLRR to obtain the detailed parts containing four rounds of decomposition and the basic parts. Subsequently, the basic parts are fused using an average strategy, while the detail part is fused using kernel norm operation. The fused image is ultimately transformed back into an RGB image, resulting in the final fusion output. We apply D2-LRR to medical image fusion tasks. The detailed parts are fused employing a nuclear-norm operation, while the basic parts are fused using an average strategy. Comparative analyses among existing methods showcase that our proposed approach attains cutting-edge fusion performance in both objective and subjective assessments.

Ontology user portraits describe the semantic structure of users’ interests. It is very important to study the similar relationship between user portraits to find the communities with overlapping interests. The hierarchical characteristics of user interest can generate multiple similarity relations, which is conducive to the formation of interest clusters. This paper proposed a method of overlapping community detection combining the hierarchical characteristics of user interest and the module distribution entropy of node. First, a hierarchical user interest model was constructed based on the ontology knowledge base to measure the multi-granularity topic similarity of users. Then, a heterogeneous hypergraph was established by using the multi-granularity topic similarity and the following similarity of users to represent the interest network. Based on the mechanism of module distribution entropy of nodes, the community detection algorithm was applied to identify the interested community. The real performance of the proposed algorithm on multiple networks was verified by experiments. The experimental results show that the proposed algorithm is better than the typical overlapping community detection algorithm in terms of accuracy and recall rate.

Anomaly detection, as one of the most critical tasks in hyperspectral image (HSI) processing, has been paid extensive attention in the past decades. The classical work of hyperspectral anomaly detection, sample correlation matrix R-based anomaly detection (R-AD), can achieve desirable detection accuracy, but its processing complexity remains a challenging problem. Real-time anomaly detection methods speed up the procedure through processing the data acquired by grating splitting or acousto-optic tunable filter (AOTF)-based imaging systems in the manner of pixel-by-pixel, line-by-line, or band-by-band. However, in practical industrial scenarios, the multiarray filter spectral imaging system is generally leveraged to avoid high construction and maintenance costs, which acquires multiline multiband (MLMB) data lacking a real-time processing version. In addition, the background of industrial assembly lines is chaotic, and the changes between batches are rapid, both increasing the difficulty of accurate anomaly detection online. To cope with these challenges, a real-time online processing version [real-time multiline multiband R anomaly detection (RTMLMB-RAD)] of R-AD based on MLMB data is proposed here for the first time. Specifically, a multiline multiband correlation matrix (MLMBCM) is designed to update the detector of R-AD recursively with MLMB data acquisition. Since MLMBCM relies only on the previous result and the current data, a large amount of data storage and complex matrix inversion calculation are avoided. At the same time, local anomaly detection mode is adopted to improve the sensitivity of anomalies in different batches and different types of products. The experimental results on six public hyperspectral datasets demonstrate that the proposed RTMLMB-RAD algorithm outperforms other state-of-the-art methods in terms of time cost and detection accuracy. Particularly, the proposed algorithm was tested on the tobacco primary processing line with a detection rate of 87 and a false alarm rate of 10 at a speed of 3 m/s, further verifying the algorithm's feasibility and effectiveness in practical industrial scenarios.

Global climate change is intensifying, and more measures to reduce carbon emissions should be taken. A method based on graph neural network can predict the carbon emission factors of power grid and that helps the users side to take more active measures to reduce the carbon emission behavior caused by electricity consumption. By learning the data of power grid loads and carbon emission factors through the graph neural network, which quickly predicts the carbon emission factors of nodes of the power grid, and provides an auxiliary basis for power grid dispatching. Finally, using a standard IEEE 39-bus system, the validity of the prediction model is verified by comparing the prediction data with the data obtained from traditional power flow calculations.

Due to the characteristics of online learning resource recommendation such as large scale, uneven quality and diversity of preferences, how to accurately obtain various personalized learning resource lists has become an urgent problem to be solved in the field of online learning resource recommendation. This paper proposes an online learning resource recommendation model based on the improved NSGA-Ⅱ algorithm, which integrates the Tabu search algorithm to improve the local search ability of NSGA-Ⅱ algorithm. It takes background fitness, cognitive fitness and diversity as the objective functions for optimization. The dynamic updating of crowding degree is used to avoid the risk that the individuals with low crowding degree in the same area are deleted at the same time. Meanwhile, an adaptive genetic algorithm is applied to assign the optimal crossover rate and the mutation rate according to individual adaptability level, which ensures the convergence of genetic algorithm and the diversity of population. The experimental results show that the proposed model is superior to the traditional recommendation algorithm in terms of accuracy index, mean fitness, recall rate, F1 mean, HV, GD and IGD, etc., thus verifying the feasibility and effectiveness of the algorithm.

Medical services play a pivotal role in people’s lives and in the national economy. Although the number of healthcare facilities is currently growing every year, there are still major problems in terms of access and pressure on the flow of people. Therefore, there is an urgent need for complementary medical services to alleviate the flow of patients and their psychological burden and to enable them to receive timely medical advice. This article designs and implements a medical Q&A system based on deep learning. We took a retrieval-based approach, using crawler technology that has been manually reviewed to build the Q&A database, and the Seq2Seq algorithm and the TF-IDF model to build the answer generation model. The medical question and answer system developed enable effective Q&A and relevant medical advice to be given. The algorithm proposed in this paper can quickly provide users with accurate answers compared to conventional search methods in real datasets.

article: The effect and mechansim of β-arrestins in the regulation of microglia polarization in the pathogenesis of Parkinson’s disease - Minerva Medica 2022 December;113(6):1036-8 - Minerva Medica - Journals