检索结果-内蒙古大学图书馆

A novel hybrid method for the registration of deformable medical images using common subregions

Multimedia Tools and Applications 2025年第10期84卷 7401-7424页

作者： Alam, Fakhre Rahman, Sami Ur Ahmad, Jamil Department of Computer Science and Information Technology University of Malakand Dir Pakistan

Deformable image registration is a fundamental technique in medical image analysis and provide physicians with a more complete understanding of patient anatomy and function. Deformable image registration has potentially improved the geometric precision of medical images with large and complex deformations. During the last decade, several types of deformable image registration techniques have been developed with high accuracy and efficiency. However, little work has been done regarding how image registration is performed based on interested common subregions which is still an open question in the field. In this paper, we propose a hybrid method that combines local rigid registration with global deformable registration to address the issue of interested common subregions registration. We first detected and segmented interested common subregions in source and target images, and then registered the obtained interested common subregions with local transformations. The obtained transformation parameters in local registration are then applied on the deformed source image which recovered it according to the coordinates of fixed target image. The fixed target image is then rigidly registered with the recovered source image. Finally, the obtained image from rigid image registration is again registered with fixed target image using deformable registration and a more informative registered image is obtained. The main objective of the proposed work is to address the issue of interested common subregions-based registration in deformable medical images. The proposed deformable image registration method is evaluated on 2D brain MRI images of glioma patients. The evaluation parameters are mean square error (MSE), sum of squared differences (SSD), peak signal to noise ratio (PSNR), cross-correlation CC) and computation time. The overall Analysis of variance (ANOVA) for the mentioned variables are significant (i.e., the value of P © The Author(s), under exclusive licence to Springer S

关键词： Mean square error

来源：评论

学校读者我要写书评

暂无评论

Examining the quality of learned representations in self-supervised medical image analysis: a comprehensive review and empirical study

引用

Multimedia Tools and Applications 2025年第9期84卷 6295-6325页

作者： Pani, Kaliprasad Chawla, Indu Department of Computer Science & IT Jaypee Institute of Information Technology Noida India

Medical Image Analysis (MIA) is integral to healthcare, demanding advanced computational techniques for precise diagnostics and treatment planning. The demand for accurate and interpretable models is imperative in the ever-evolving healthcare landscape. This paper explores the potential of Self-Supervised Learning (SSL), transfer learning and domain adaptation methods in MIA. The study comprehensively reviews SSL-based computational techniques in the context of medical imaging, highlighting their merits and limitations. In an empirical investigation, this study examines the lack of interpretable and explainable component selection in existing SSL approaches for MIA. Unlike prior studies that randomly select SSL components based on their performance on natural images, this paper focuses on identifying components based on the quality of learned representations through various clustering evaluation metrics. Various SSL techniques and backbone combinations were rigorously assessed on diverse medical image datasets. The results of this experiment provided insights into the performance and behavior of SSL methods, paving the way for an explainable and interpretable component selection mechanism for artificial intelligence models in medical imaging. The empirical study reveals the superior performance of BYOL (Bootstrap Your Own Latent) with resnet as the backbone, as indicated by various clustering evaluation metrics such as Silhouette Coefficient (0.6), Davies-Bouldin Index (0.67), and Calinski-Harabasz Index (36.9). The study also emphasizes the benefits of transferring weights from a model trained on a similar dataset instead of a dataset from a different domain. Results indicate that the proposed mechanism expedited convergence, achieving 98.66% training accuracy and 92.48% testing accuracy in 23 epochs, requiring almost half the number of epochs for similar results with ImageNet weights. This research contributes to advancing the understanding of SSL in MIA, providin

关键词： Supervised learning

来源：评论

学校读者我要写书评

暂无评论

Novel Framework for Generating Criminals Images Based on Textual Data Using Identity GANs

引用

computers, Materials & Continua 2023年第7期76卷 383-396页

作者： Mohamed Fathallah Mohamed Sakr Sherif Eletriby Department of Computer Science Faculty of Computers and InformationKafrelsheikh UniversityKafrelsheikhEgypt Department of Computer Science Faculty of Computers and InformationMenoufia UniversityMenoufiaEgypt

Text-to-image generation is a vital task in different fields,such as combating crime and terrorism and quickly arresting *** several years,due to a lack of deep learning and machine learning resources,police officials required artists to draw the face of a *** methods of identifying criminals are inefficient and *** paper presented a new proposed hybrid model for converting the text into the nearest images,then ranking the produced images according to the available *** framework contains two main steps:generation of the image using an Identity Generative Adversarial Network(IGAN)and ranking of the images according to the available data using multi-criteria decision-making based on neutrosophic *** IGAN has the same architecture as the classical Generative Adversarial Networks(GANs),but with different modifications,such as adding a non-linear identity block,smoothing the standard GAN loss function by using a modified loss function and label smoothing,and using mini-batch *** model achieves efficient results in Inception Distance(FID)and inception score(IS)when compared with other architectures of GANs for generating images from *** IGAN achieves 42.16 as FID and 14.96 as *** it comes to ranking the generated images using Neutrosophic,the framework also performs well in the case of missing information and missing data.

关键词： GAN deep learning text-to-image identity GAN

来源：评论

学校读者我要写书评

暂无评论

Split-n-Swap: A New Modification of the Twofish Block Cipher Algorithm

引用

computers, Materials & Continua 2023年第1期74卷 1723-1734页

作者： Awny Sayed Maha Mahrous Enas Elgeldawi Computer Science Department Faculty of ScienceMinia UniversityMinia61519Egypt Faculty of Computing and Information Technology Information Technology DepartmentKing Abdulaziz UniversityJeddah21589Saudi Arabia Computer Science Department Faculty of Computers and InformationMinia UniversityMinia61519Egypt

Securing digital data from unauthorized access throughout its entire lifecycle has been always a critical concern.A robust data security system should protect the information assets of any organization against cybercriminal *** Twofish algorithm is one of the well-known symmetric key block cipher cryptographic algorithms and has been known for its rapid *** when it comes to security,it is not the preferred cryptographic algorithm to use compared to other algorithms that have shown better *** applications and social platforms have adopted other symmetric key block cipher cryptographic algorithms such as the Advanced Encryption Standard(AES)algorithm to construct their main security *** this paper,a new modification for the original Twofish algorithm is proposed to strengthen its security and to take advantage of its fast *** new algorithm has been named Split-n-Swap(SnS).Performance analysis of the new modification algorithm has been performed using different measurement *** experimental results show that the complexity of the SnS algorithm exceeds that of the original Twofish algorithm while maintaining reasonable values for encryption and decryption times as well as memory utilization.A detailed analysis is given with the strength and limitation aspects of the proposed algorithm.

关键词： Twofish advanced encryption standard(AES) cryptography symmetric key block cipher

来源：评论

学校读者我要写书评

暂无评论

Workout Action Recognition in Video Streams Using an Attention Driven Residual DC-GRU Network

引用

computers, Materials & Continua 2024年第5期79卷 3067-3087页

作者： Arnab Dey Samit Biswas Dac-Nhuong Le Department of Computer Science and Technology Indian Institute of Engineering Science and TechnologyShibpurHowrah711103India Faculty of Information Technology Haiphong UniversityHaiphong180000Vietnam

Regular exercise is a crucial aspect of daily life, as it enables individuals to stay physically active, lowers thelikelihood of developing illnesses, and enhances life expectancy. The recognition of workout actions in videostreams holds significant importance in computer vision research, as it aims to enhance exercise adherence, enableinstant recognition, advance fitness tracking technologies, and optimize fitness routines. However, existing actiondatasets often lack diversity and specificity for workout actions, hindering the development of accurate recognitionmodels. To address this gap, the Workout Action Video dataset (WAVd) has been introduced as a significantcontribution. WAVd comprises a diverse collection of labeled workout action videos, meticulously curated toencompass various exercises performed by numerous individuals in different settings. This research proposes aninnovative framework based on the Attention driven Residual Deep Convolutional-Gated Recurrent Unit (ResDCGRU)network for workout action recognition in video streams. Unlike image-based action recognition, videoscontain spatio-temporal information, making the task more complex and challenging. While substantial progresshas been made in this area, challenges persist in detecting subtle and complex actions, handling occlusions,and managing the computational demands of deep learning approaches. The proposed ResDC-GRU Attentionmodel demonstrated exceptional classification performance with 95.81% accuracy in classifying workout actionvideos and also outperformed various state-of-the-art models. The method also yielded 81.6%, 97.2%, 95.6%, and93.2% accuracy on established benchmark datasets, namely HMDB51, Youtube Actions, UCF50, and UCF101,respectively, showcasing its superiority and robustness in action recognition. The findings suggest practicalimplications in real-world scenarios where precise video action recognition is paramount, addressing the persistingchallenges in the field. TheWAVd datas

关键词： Workout action recognition video stream action recognition residual network GRU attention

来源：评论

学校读者我要写书评

暂无评论

Central Tendency Feature Selection (CTFS): a novel approach for efficient and effective feature selection in intrusion detection systems

引用

Multimedia Tools and Applications 2025年 1-22页

作者： Bouke, Mohamed Aly Abdullah, Azizol Udzir, Nur Izura Samian, Normalia Department of Communication Technology and Network Faculty of Computer Science and Information Technology Universiti Putra Malaysia 43400 Serdang Malaysia

In the digital era, the escalation of data generation and cyber threats has heightened the importance of network security. Machine Learning-based Intrusion Detection Systems (IDS) play a crucial role in combating these threats, yet there is a significant gap in their feature selection methodologies. These traditional methods often fail to handle large datasets, leading to performance limitations efficiently. To advance this crucial aspect of IDS, we propose the novel Central Tendency-Based Feature Selection (CTFS) method. Our empirical investigation, utilizing the widely acknowledged UNSWNB15 and NSLKDD cybersecurity datasets, reveals that CTFS not only outperforms established feature selection methods like Univariate Selection, Recursive Feature Elimination (RFE), Principal Component Analysis (PCA), and Gini Index in terms of accuracy, achieving up to 95%, and recall, reaching up to 99%, but it also substantially enhances execution efficiency. Specifically, CTFS demonstrates a remarkable reduction in feature selection time to approximately 0.1 s for UNSWNB15 and 0.026 s for NSLKDD, outmatching the more time-intensive RFE, which recorded 569.7 s and 24.9 s, respectively. These findings showcase CTFS as a potent solution, balancing high performance with computational efficiency, a key consideration in time-sensitive security applications. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2025.

关键词： Cybersecurity

来源：评论

学校读者我要写书评

暂无评论

Feature Selection Techniques for Enhancing App User Review Analysis

引用

IEEE Access 2025年 13卷 85279-85293页

作者： Salleh, Amran Osman, Mohd Hafeez Hassan, Sa'Adah Said, Mar Yah Universiti Putra Malaysia Faculty of Computer Science and Information Technology Serdang43400 Malaysia

The rapid growth of user-generated content, particularly app user reviews, presents a significant challenge in analyzing and extracting useful insights. The unstructured nature, inconsistent quality, and large volume of these reviews make it difficult to identify relevant information for app maintenance and updates. This study addresses this challenge by evaluating the impact of different feature selection techniques on the performance of machine learning models in multi-label classification tasks for app user review analysis. Our findings indicate that the subset attributes derived from a combination of information Gain, GINI Index, and Correlation Matrix can improve the performance of machine learning models. Using a Support Vector Machine with proposed innovative score-based zero shot technique, promising results were achieved on average with 96.75% precision, 69.39% recall, and 80.81% F1-Score. Additionally, high-quality features, such as the percentage Difference can enhance performance in multi-label classification tasks, providing valuable insights for practitioners and researchers. The research implications and significance highlight the practical applications and strategic value of these findings, contributing to the advancement of knowledge and practice in the field of software engineering, particularly for multi-label classification tasks. © 2013 IEEE.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

Hyperparameter Optimization for Convolutional Neural Networks using the Salp Swarm Algorithm

引用

Informatica (Slovenia) 2023年第9期47卷 133-144页

作者： Abdulsaed, Entesar H. Alabbas, Maytham Khudeyer, Raidah S. Department of Computer Science College of Computer Science and Information Technology University of Basrah Basrah Iraq Department of Computer Information Systems College of Computer Science and Information Technology University of Basrah Basrah Iraq

Convolutional neural networks (CNNs) have exceptionally performed across various computer vision tasks. However, their effectiveness depends heavily on the careful selection of hyperparameters. Optimizing these hyperparameters can be challenging and time-consuming, especially when working with large datasets and complex network architectures. In response, we propose a novel approach for hyperparameter optimization in CNNs using the Salp Swarm Algorithm (SSA). Based on the natural behavior of mollusks, SSA mimics the collective intelligence that governs feeding and navigation. Taking advantage of SSA's unique properties, our research thoroughly explores the hyperparameter space. This exploration aims to identify the algorithm that maximizes CNNs performance. This paper presents the architecture of the SSA-based framework for hyperparameter optimization and compares it to other established optimization techniques, such as Particle Swarm Optimization (PSO) and Genetic Algorithm (GA). We also present experimental results using the MNIST and fashion MNIST datasets, achieving an impressive classification accuracy of 99.46% for MNIST and 94.53% for fashion-MNIST. This case study not only contributes to the fields of deep learning and hyperparameter optimization by demonstrating the effectiveness of SSA in optimizing CNNs, but it also provides benefits to researchers and practitioners who are looking for optimal hyperparameter configurations for CNNs in a variety of computer vision applications. We also evaluate the scalability and robustness of our proposed method in the context of different CNNs structures. The insights we gained highlight SSA's potential for addressing challenges related to hyperparameter optimization. © 2023 Slovene Society Informatika. All rights reserved.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Audio verification in forensic investigation using light deep neural network

引用

International Journal of information technology (Singapore) 2024年第5期16卷 2813-2821页

作者： AL-Shakarchy, Noor D. Abdullah, Zahraa Najm Alameen, Zainab M. Harjan, Zahraa A. Department Of Computer Science Faculty of Computer Science and Information Technology Kerbala University Kerbala Iraq Kerbala University Kerbala Iraq Faculty of Computer Science and Information Technology Kerbala University Kerbala Iraq

Recently people have difficulties distinguishing real speech from computer-generated speech so that the synthetic voice is getting closer to a natural-sounding voice, due to the advancements in deep learning and voice-generation techniques. This high level of naturalness allows for synthesizing a person target voice and reproduce someone's voice with high accuracy. Audio forensics is care with improving the speech intelligibility, enhancement, and analysis in order to prove the authentication of recorded audio files it is authentic and no manipulations were done to them. This work proposes the Audio verification model in Forensic Investigation using one Dimensional Convolutional Neural networks with new light architecture. The proposed model build a natural voice characteristics pattern by analysis the information available in an audio file and extract salient features in order to build a feature maps, and then tracked these feature maps over time is the same model to make the decision about whether the audio is Synthetic or not. The proposed model performed in two main stages: the pre-processing stage then verification predictor stage, each stage contained several steps performed different functions. The experiments were conducted on a last version benchmark dataset called the Fake-or-Real Dataset from the APTLY lab and ASVspoof 2019 dataset. The proposed model achieve successful results, so that the accuracy realizes 99.89% (± 1.78%), and the value of the loss function is 0.007 on the Fake-or-Real Dataset, and achieved a 0.006% Equal Error Rate (EER) on ASVSpoof 2019 dataset. © Bharati Vidyapeeth's Institute of computer Applications and Management 2024.

关键词： Audio Batch normalization Dropout One dimensional convolutional neural network (1DCNN) Pooling Real voice Synthetic voice

来源：评论

学校读者我要写书评

暂无评论

Enhanced Federated Learning Framework based on Deep Learning and Neutrosophic Set for Android Malware Classification

引用

Neutrosophic Sets and Systems 2025年 82卷 781-799页

作者： Abdellah, Mohamed Refaat Oudah, Hasan H. Badawy, Ahmed Mohamed Ahmed Mohamed, Abdelfattah Abdelfattah M. Bedier, Shady Ahmed Metwaly, Ahmed A. Eassa, Mohamed Abdelhafeez, Ahmed The Department of Computer Science College of Information Technology Misr University for Science and Technology Cairo Egypt College of Arts Department of Journalism Ahl Al Bayt University Karbala Iraq Cybersecurity Department Faculty of Computers and Artificial Intelligence Helwan University Egypt Computer Science Department Faculty of Information System and Computer Science October 6 University Giza12585 Egypt Department of Computer Science Faculty of Computers and Informatics Zagazig University Zagazig44519 Egypt Applied Science Research Center Applied Science Private University Amman Jordan

Malware detection is one of the critical tasks of cybersecurity, especially considering the growing popularity of mobile devices. The integrity and security of mobile ecosystems rely on the capacity to identify malware quickly and precisely, as security hacks against these devices become increasingly frequent. Due to the requirement for massive volumes of data aggregation, traditional centralized machine learning (ML) techniques for malware detection face challenges with data sharing, computational complexity, and privacy. This research addresses these challenges by proposing a novel model, called "CNN-Fed", which is based on Convolutional Neural Networks (CNN) with Federated Learning (Fed). The main goal of this work is to create a global classifier for Android malware detection that is highly accurate and does not require centralised data aggregation. Using four benchmark datasets—Drebin, Malgenome, Kronodroid, and Tuandromd—the CNN-Fed model is trained across many clients in a federated setting under the suggested architecture. Then we use multi-criteria decision-making (MCDM) methodology to evaluate the number of clients and number of rounds. We use two MCDM methods such as Entropy methodology to compute the criteria weights and the CoCoSo methodology to rank the alternatives. We use the interval-valued neutrosophic sets (IVNSs) to deal with uncertainty and vague information. We use five criteria such as accuracy, F1-score, precision, recall, and FP and six of alternatives refer to the number of clients and number of rounds. The results show the best alternative has a larger number of clients and rounds. © 2025, University of New Mexico. All rights reserved.

关键词： Android malware

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：