Computational Intelligence最新文献

Nowadays, most people have been admitted to emergencies with severe pain caused by kidney stones worldwide. In this case, diverse imaging approaches are aided in the detection process of stones in the kidney. Moreover, the specialist acquires better diagnosis and interpretation of this image. Here, computer-aided techniques are considered the practical techniques, which it is used as the auxiliary tool for the process of diagnosis. Most urologists have failed to train the type of kidney stone identification effectively and it is operator-dependent. Concerning the surgical operation, there is a requirement for accurate as well as adequate detection of stone position in the kidney. Thus, it has made the detection process even more difficult. To overcome the challenging issues, an effective detection model for kidney stones using classifiers. Initially, the input images are collected from the standard resources. Further, the input images are subjected to the adaptive multi-convolutional neural network with attention mechanism (AMC-AM) feature fusion model, in which, the pertinent features are extracted from the three networks: Visual Geometry Group16 (VGG16), Residual Network (ResNet), and Inception net. Thus, the three distinct features are obtained for the feature fusion procedure. Finally, the resultant features are subjected as input to the final layer of CNN. In the proposed network, the model is integrated with the attention mechanism and also the parameter tuning is done by proposing the modified social distance of coronavirus mask protection algorithm (MSD-CMPA). Therefore, the performance is examined using different metrics and compared with other baseline models. Hence, the proposed model overwhelms the outstanding results in detecting the kidney stones that aid the individual in getting rid of kidney disorders.

To meet the demands of modern technologies such as 5G, big data, edge computing, precision, and sustainable agriculture, the combination of Internet-of-Things (IoT) with software-defined networking (SDN) known as SD-IoT is suggested to automate the network by leveraging the programmable and centralized SDN interfaces. The previous literature has suggested quality-of-service (QoS) aware flow processing using manual strategies or heuristic algorithms, however, these schemes proposed with white-box approaches do not provide effective results as the network scales or dynamic changes are happening. This article proposes a novel QoS provision strategy using deep reinforcement learning (DRL) to calculate the optimal routes autonomously for SD-IoT traffic. To satisfy the different demands of flows in the SD-IoT network the flows are divided into two types. Hence, based on their service demand the routes are generated for them as per service request. The scenario is explained with precision agriculture based on SD-IoT and results are compared with benchmark strategies. A real internet topology is used for the evaluation of results. The results indicated that the proposed method gives improvements for QoS such as delay, throughput, packet loss rate, and jitter compared with benchmark models.

This paper aims to develop a new deep learning model (DeepPneuNet) and evaluate its performance in predicting Pneumonia infection diagnosis based on patients' chest x-ray images. We have collected 5856 chest x-ray images that are labeled as either “pneumonia” or “normal” from a public forum. Before applying the DeepPneuNet model, a necessary feature extraction and feature mapping have been done on the input images. Conv2D layers with a 1 × 1 kernel size are followed by ReLU activation functions to make up the model. These layers are in charge of recognizing important patterns and features in the images. A MaxPooling 2D procedure is applied to minimize the spatial size of the feature maps after every two Conv2D layers. The sparse categorical cross-entropy loss function trains the model, and the Adam optimizer with a learning rate of 0.001 is used to optimize it. The DeepPneuNet provides 90.12% accuracy for diagnosis of the Pneumonia infection for a set of real-life test images. With 9,445,586 parameters, the DeepPneuNet model exhibits excellent parameter efficiency. DeepPneuNet is a more lightweight and computationally efficient alternative when compared to the other pre-trained models. We have compared accuracies for predicting Pneumonia diagnosis of our proposed DeepPneuNet model with some state-of-the-art deep learning models. The proposed DeepPneuNet model is more advantageous than the existing state-of-the-art learning models for Pneumonia diagnosis with respect to accuracy, precision, recall, F-score, training parameters, and training execution time.

In recent years, the algorithm based on review text has been widely used in recommendation systems, which can help mitigate the effect of sparsity in rating data within recommender algorithms. Existing methods typically employ a uniform model for capturing user and item features, but they are limited to the shallow feature level, and the user's personalized preferences and deep features of the item have not been fully explored, which may affect the relationship between the two representations learned by the model. The deeper relationship between them will affect the prediction results. Consequently, we propose a personalized recommendation method based on the rating matrix and review text denoted PRM-RR, which is used to deeply mine user preferences and item characteristics. In the process of processing the comment text, we employ ALBERT to obtain vector representations for the words present in the review text firstly. Subsequently, taking into account that significant words and reviews bear relevance not solely to the review text but also to the user's individualized preferences, the proposed personalized attention module synergizes the user's personalized preference information with the review text vector, thereby engendering an enriched review-based user representation. The fusion of the user's review representation and rating representation is accomplished through the feature fusion module using cross-modal attention, yielding the final user representation. Lastly, we employ a factorization machine to predict the user's rating for the item, thereby facilitating the recommendation process. Experimental results on three benchmark datasets show that our method outperforms the baseline algorithm in all cases, demonstrating that our method effectively improves the performance of recommendations. The code is available at https://github.com/ZehuaChenLab/PRM-RR.

Power consumption management is vital in achieving sustainable and low-carbon green communication goals in 6G smart agriculture. This research aims to provide a low-power consumption measurement framework designed specifically for critical data handling in smart agriculture application networks. Deep Q-learning combined with game theory is proposed to allow network entities such as Internet of Things (IoT) devices, Intelligent Reflecting Surfaces (IRSs), and Base Stations (BS) to make intelligent decisions for optimal resource allocation and energy and power consumption. The learning capabilities of DQL with strategic reasoning of game theory, a hybrid framework, have been developed to realize an adaptive routing plan that emphasizes energy-conscious communication protocols and underestimates the environment. It further enables the investigation of multi-IRS performance through several key metrics assessments, such as reflected power consumption, energy efficiency, and Signal-to-Noise Ratio (SNR) improvement.

In the realm of power Internet of Things (IoT) networks, secure inspection detection (SID) is paramount for maintaining system integrity and security. This paper presents a novel framework that leverages deep learning-based semi-autoencoders in conjunction with a hybrid recommendation algorithm to enhance SID tasks. Our proposed method utilizes the deep learning-based semi-autoencoder to effectively capture and learn complex patterns from high-dimensional power IoT data, facilitating the identification of anomalies indicative of potential security threats. The hybrid recommendation algorithm, which combines collaborative filtering and content-based filtering, further refines the detection process by cross-verifying the identified anomalies with historical data and contextual information, thereby improving the accuracy and reliability of the SID tasks. Through extensive simulations and practical data evaluations, our proposed framework demonstrates superior performance over conventional methods, achieving higher detection accuracy. In particular, the detection accuracy of the proposed scheme is more than 20% higher than that of the competing schemes.

Depression, a prevalent mental disorder in modern society, significantly impacts people's daily lives. Recently, there have been advancements in developing automated diagnosis models for detecting depression. However, data scarcity, primarily due to privacy concerns, has posed a challenge. Traditional speech features have limitations in representing knowledge for depression diagnosis, and the complexity of deep learning algorithms necessitates substantial data support. Furthermore, existing multimodal methods based on neural networks overlook the heterogeneity gap between different modalities, potentially resulting in redundant information. To address these issues, we propose a multimodal depression detection model based on the Enhanced Cross-Attention (ECA) Mechanism. This model effectively explores text-speech interactions while considering modality heterogeneity. Data scarcity has been mitigated by fine-tuning pre-trained models. Additionally, we design a modal fusion module based on ECA, which emphasizes similarity responses and updates the weight of each modal feature based on the similarity information between modal features. Furthermore, for speech feature extraction, we have reduced the computational complexity of the model by integrating a multi-window self-attention mechanism with the Fourier transform. The proposed model is evaluated on the public dataset, DAIC-WOZ, achieving an accuracy of 80.0% and an average F1 value improvement of 4.3% compared with relevant methods.

Brain tumors are prevalent forms of malignant neoplasms that, depending on their type, location, and grade, can significantly reduce life expectancy due to their invasive nature and potential for rapid progression. Accordingly, brain tumors classification is an essential step that allows doctors to perform appropriate treatment. Many studies have been done in the sector of medical image processing by employing computational methods to effectively segment and classify tumors. However, the larger amount of information collected by healthcare images prohibits the manual segmentation process in a reasonable time frame, reducing error measures in healthcare settings. Therefore, automated and efficient techniques for segmentation are crucial. In addition, various visual information, noisy images, occlusion, uneven image textures, confused objects, and other features may impact the process. Therefore, the implementation of deep learning provides remarkable results in medicinal image processing, particularly in the segmentation and classification process. However, conventional deep learning-assisted methods struggle with complex structures and dimensional issues. Thus, this paper develops an effective technique for diagnosing brain tumors. The main aspect of the proposed system is to classify the brain tumor types by segmenting the affected regions of the raw images. This novel approach can be applied for various applications like diagnostic centers, decision-making tools, clinical trials, medical research institutes, disease prognosis, and so on. Initially, the requisite images are collected from standard datasets and further, it is subjected to the segmentation period. In this stage, the Multi-scale and Dilated TransUNet++ (MDTUNet++) model is employed to segment the abnormalities. Further, the segmented images are given into an Adaptive Dilated Dense Residual Attention Network (ADDRAN) to classify the brain tumor types. Here, to optimize the ADDRAN technique's parameters, an Improved Hermit Crab Optimizer (IHCO) is supported, which increases the accuracy rates of the overall network. Finally, the numerical examination is conducted to guarantee the robustness and usefulness of the designed model by contrasting it with other related techniques. For Dataset 1, the accuracy value attains 93.71 for the proposed work compared to 87.86 for CNN, 90.18 for DenseNet, and 89.56 and 90.96 for RAN and DRAN, respectively. Thus, supremacy has been achieved for the recommended system while detecting the brain tumor types.

Nail-biting (NB) or onychophagia is a compulsive disorder that affects millions of people in both children and adults. It has several health complications and negative social effects. Treatments include surgical interventions, pharmacological medications, or additionally, it can be treated using behavioral modification therapies that utilize positive reinforcement and periodical reminders. Although it is the least invasive, such therapies still depend on manual monitoring and tracking which limits their success. In this work, we propose a novel approach for automatic real-time NB detection and alert on a smartwatch that does not require surgical intervention, medications, or manual habit monitoring. It addresses two key challenges: First, NB actions generate subtle motion patterns at the wrist that lead to a high false-positives (FP) rate even when the hand is not on the face. Second, is the challenge to run power-intensive applications on a power-constrained edge device like a smartwatch. To overcome these challenges, our proposed approach implements a pipeline of three convolutional neural networks (CNN) models instead of a single model. The first two models are small and efficient, designed to detect face-touch (FT) actions and hand movement away (MA) from the face. The third model is a larger and deeper CNN model dedicated to classifying hand actions on the face and detecting NB actions. This separation of tasks addresses the key challenges: decreasing FPs by ensuring NB model is activated only when the hand on the face, and optimizing power usage by ensuring the larger NB model runs only for short periods while the efficient FT model runs most of the time. In addition, this separation of tasks gives more freedom to design, configure, and optimize the three models based on each model task. Lastly, for training the main NB model, this work presents further optimizations including developing NB dataset from start through a dedicated data collection application, applying data augmentation, and utilizing several CNN optimization techniques during training. Results show that the model pipeline approach minimizes FPs significantly compared with the single model for NB detection while improving the overall efficiency.

The sharing of patient information on an open network has drawn attention to the healthcare system. Security is the primary issue while sharing documents online. Thus, a dual watermarking technique has been developed to improve the security of shared data. The classical watermarking schemes are resilient to many attacks. Protecting the authenticity and copyrights of medical images is essential to prevent duplication, modification, or unauthorized distribution. This paper proposes a robust, novel dual watermarking system for securing healthcare data. Initially, watermarking is performed based on redundant lifting wavelet transform (LWT) and turbo code decomposition for COVID-19 patient images and patient text data. To achieve a high level of authenticity, watermarks in the form of encoded text data and decomposed watermark images are inserted together, and an inverse LWT is used to generate an initial watermarked image. Improve imperceptibility and robustness by incorporating the cover image into the watermarked image. Cross-guided bilateral filtering (CG_BF) improves cover image quality, while the integrated Walsh–Hadamard transform (IWHT) extracts features. The novel adaptive coati optimization (ACO) technique is used to identify the ideal location for the watermarked image in the cover image. To improve security, the watermarked image is dissected using discrete wavelet transform (DWT) and encrypted with a chaotic extended logistic system. Finally, the encrypted watermarked image is implanted in the desired place using a novel deep-learning model based on the Hybrid Convolutional Cascaded Capsule Network (HC³Net). Thus, the secured watermarked image is obtained, and the watermark and text data are extracted using the decryption and inverse DWT procedure. The performance of the proposed method is evaluated using accuracy, peak signal-to-noise ratio (PSNR), NC, and other metrics. The proposed method achieved an accuracy of 99.26%, which is greater than the existing methods.