检索结果-内蒙古大学图书馆

WALINET: A water and lipid identification convolutional neural network for nuisance signal removal in 1H MR spectroscopic imaging

引用

MAGNETIC RESONANCE IN MEDICINE 2024年第4期93卷 1430-1442页

作者： Weiser, Paul J. Langs, Georg Motyka, Stanislav Bogner, Wolfgang Courvoisier, Sebastien Hoffmann, Malte Klauser, Antoine Andronesi, Ovidiu C. Massachusetts Gen Hosp Athinoula A Martinos Ctr Biomed Imaging Boston MA 02114 USA Harvard Med Sch Massachusetts Gen Hosp Dept Radiol Boston MA USA Med Univ Vienna Dept Biomed Imaging & Image Guided Therapy Computat Imaging Res Lab Vienna Austria Med Univ Vienna High Field MR Ctr Dept Biomed Imaging & Image Guided Therapy Vienna Austria Ctr Biomed Imaging CIBM Geneva Switzerland Univ Geneva Dept Radiol & Med Informat Geneva Switzerland Siemens Healthineers Int AG Adv Clin Imaging Technol Lausanne Switzerland

Purpose: Proton magnetic resonance spectroscopic imaging (1H-MRSI) provides non-invasive spectral-spatial mapping of metabolism. However, long-standing problems in whole-brain 1H-MRSI are spectral overlap of metabolite peaks with large lipid signal from scalp, and overwhelmingwater signal that distorts spectra. Fast and effective methods are needed for high-resolution 1H-MRSI to accurately remove lipid and water signals while preserving the metabolite signal. The potential of supervised neural networks for this task remains unexplored, despite their success for other MRSI processing. methods: We introduce a deep learning method based on a modified Y-NET network for water and lipid removal in whole-brain 1H-MRSI. TheWALINET (WAter and LIpid neural NETwork) was compared with conventional methods such as the state-of-the-art lipid L2 regularization and Hankel-Lanczos singular value decomposition (HLSVD) water suppression. methods were evaluated on simulated models and in vivo whole-brain MRSI using NMRSE, SNR, CRLB, and FWHM metrics. Results: WALINET is significantly faster and needs 8s for high-resolution whole-brain MRSI, compared with 42min for conventional HLSVD+L2. WALINET suppresses lipid and water in the brain by 25-45 and 34-53-fold, respectively. WALINET has better performance than HLSVD+L2, providing: (1) more lipid removal with 41% lower NRMSE;(2) better metabolite signal preservation with 71% lower NRMSE in simulated data;155% higher SNR and 50% lower CRLB in in vivo data. Metabolic maps obtained by WALINET in healthy subjects and patients show better gray-/white-matter contrast with more visible structural details. Conclusions: WALINET has superior performance for nuisance signal removal and metabolite quantification on whole-brain 1H-MRSI compared with conventional state-of-the-art techniques. This represents a new application of deep learning for MRSI processing, with potential for automated high-throughput workflow.

关键词： brain metabolite quantification mr spectroscopic imaging ultrahigh-field mr water and lipid removal

来源：评论

学校读者我要写书评

暂无评论

Evaluating the Adaptability of Deep Learning-Based Multi-feature Sonar image Detection Algorithms

引用

TRAITEMENT DU signal 2024年第3期41卷 1223-1230页

作者： Wang, Zhangyun Huang, Feipeng Guizhou Transportat Planning Survey & Design Acade Guiyang 550081 Peoples R China Chinese Acad Geol Sci Inst Geomech Beijing 100081 Peoples R China

In the underwater environment, characterized by inherent complexities and prevalent noise, analyzing sonar images and detecting objects is challenging. This study employs a deep learning approach utilizing a convolutional neural network (CNN) to enhance sonar image analysis through multi -feature detection and fusion. The Visual Geometry Group Network 16 (VGG-16), a model of CNN, is utilized initially for feature extraction from sonar images. Subsequently, a weighted feature fusion technique is applied to amalgamate the feature vectors extracted by the CNN, thus forming an integrated multi -feature detection and classification model for sonar imagery. The adaptability of the proposed model is rigorously assessed through cross -validation methods. To ascertain the effectiveness of the model, the sonar images are first denoised, followed by evaluating the accuracy of sonar image classification and the prediction of sonar signals across various algorithmic models. The findings indicate that the multi -feature fusion approach yields a classification accuracy ranging between 86% and 91%, surpassing other evaluated algorithms. Furthermore, the sonar signal curve predicted by the CNN algorithm more closely approximates the actual sonar signal curve compared to alternative methods. These results affirm that the deep learning -based CNN algorithm significantly enhances the accuracy of sonar image detection and classification.

关键词： sonar image multi feature detection, convolutional neural network (CNN), weighted feature fusion image processing

来源：评论

学校读者我要写书评

暂无评论

Quaternion Vector Quantized Variational Autoencoder

引用

IEEE signal processing LETTERS 2025年 32卷 151-155页

作者： Luo, Hui Liu, Xin Sun, Jian Zhang, Yang Macau Univ Sci & Technol Fac Innovat Engn Macau 999078 Peoples R China Univ Manitoba Dept Math Winnipeg MB R3T 2N2 Canada

Vector quantized variational autoencoders, as variants of variational autoencoders, effectively capture discrete representations by quantizing continuous latent spaces and are widely used in generative tasks. However, these models still face limitations in handling complex image reconstruction, particularly in preserving high-quality details. Moreover, quaternion neural networks have shown unique advantages in handling multi-dimensional data, indicating that integrating quaternion approaches could potentially improve the performance of these autoencoders. To this end, we propose QVQ-VAE, a lightweight network in the quaternion domain that introduces a quaternion-based quantization layer and training strategy to improve reconstruction precision. By fully leveraging quaternion operations, QVQ-VAE reduces the number of model parameters, thereby lowering computational resource demands. Extensive evaluations on face and general object reconstruction tasks show that QVQ-VAE consistently outperforms existing methods while using significantly fewer parameters.

关键词： Quaternions Vectors image reconstruction Convolution neural networks Face recognition Quantization (signal) Decoding Training Indexes quaternion generative models quaternion neural networks vector quantized variational autoencoder

来源：评论

学校读者我要写书评

暂无评论

Multi-stage image deblurring method based on deep convolutional neural networks

引用

JOURNAL OF ELECTRONIC IMAGING 2025年第2期34卷

作者： Zhao, Liquan Lin, Yongpei Jia, Yanfei Northeast Elect Power Univ Minist Educ Key Lab Modern Power Syst Simulat & Control & Rene Jilin Peoples R China Beihua Univ Coll Elect & Informat Engn Jilin Peoples R China

Due to factors such as relative motion between the imaging device and the target object or out-of-focus optical system, some images may become blurred, severely affecting subsequent image processing tasks. We propose an image deblurring method based on deep convolutional neural networks to recover clear images from blurred ones. The network consists of three structurally similar sub-networks. Each sub-network comprises the multi-scale feature extraction module, single-scale feature extraction module, fast encoding-decoding module, encoding-decoding feature fusion module, and fusion strategy we proposed. The network can extract multi-scale features from input images of different scales, fusing coarse and fine-scale features and ultimately enhancing the clarity of the blurred images. In addition, we introduce the content loss function and frequency reconstruction loss function into the conventional loss function to better measure the quality of deblurred images. Experiments were conducted on the GoPro and HIDE datasets. The HIDE dataset is divided into distant and close-up scenes, allowing for a better evaluation of the generalization ability of deblurring methods in different scenarios. Compared with the existing methods, the experimental results demonstrate that the proposed approach achieves the highest peak signal-to-noise ratio and structural similarity index measure values, resulting in clearer and more natural deblurred images. Furthermore, the proposed algorithm exhibits faster image processing speed and smaller model size. The model performs better in terms of real-time capability.

关键词： image deblurring multi-stage neural network image processing

来源：评论

学校读者我要写书评

暂无评论

4KSecure: A Universal Method for Active Manipulation Detection in images of Any Resolution

引用

APPLIED SCIENCES-BASEL 2025年第8期15卷 4469-4469页

作者： Duszejko, Pawel Piotrowski, Zbigniew Mil Univ Technol Fac Elect PL-00908 Warsaw Poland

This paper presents a method for actively safeguarding image integrity based on embedding a hidden signature generated by a neural network. The proposed solution utilizes two cooperating networks: the first one is responsible for generating a signature pattern that is difficult to detect, while the second one embeds this signature into the original image in a way that ensures high visual transparency. This automatically generated signature enables the detection of manipulations in the image and allows for localizing areas where integrity violations have occurred. The main innovation of the proposed approach lies in its ability to handle images of virtually any resolution, including widely used standards in modern communication and publications, such as HD (1280 x 720), Full HD (1920 x 1080) and even 4K (3840 x 2160). This represents a significant improvement over most existing methods, which are typically limited to the small, square images (e.g., 256 x 256 or 512 x 512 pixels) commonly found in popular object classification datasets, such as imageNet (approximately 224 x 224 pixels). As a result, the proposed method opens new possibilities for actively securing the integrity of large and non-standard image formats while maintaining reasonable computational requirements. It surpasses previous limitations in terms of scalability and image proportions.

关键词： image forensics deep learning active detection passive detection tampering detection watermarking neural networks

来源：评论

学校读者我要写书评

暂无评论

VOICE: Variance of Induced Contrastive Explanations to Quantify Uncertainty in neural Network Interpretability

引用

IEEE JOURNAL OF SELECTED TOPICS IN signal processing 2025年第1期19卷 19-31页

作者： Prabhushankar, Mohit Alregib, Ghassan Georgia Inst Technol Dept Elect & Comp Engn Atlanta GA 30332 USA

In this paper, we visualize and quantify the predictive uncertainty of gradient-based post hoc visual explanations for neural networks. Predictive uncertainty refers to the variability in the network predictions under perturbations to the input. Visual post hoc explainability techniques highlight features within an image to justify a network's prediction. We theoretically show that existing evaluation strategies of visual explanatory techniques partially reduce the predictive uncertainty of neural networks. This analysis allows us to construct a plug in approach to visualize and quantify the remaining predictive uncertainty of any gradient-based explanatory technique. We show that every image, network, prediction, and explanatory technique has a unique uncertainty. The proposed uncertainty visualization and quantification yields two key observations. Firstly, oftentimes under incorrect predictions, explanatory techniques are uncertain about the same features that they are attributing the predictions to, thereby reducing the trustworthiness of the explanation. Secondly, objective metrics of an explanation's uncertainty, empirically behave similarly to epistemic uncertainty. We support these observations on two datasets, four explanatory techniques, and six neural network architectures.

关键词： Uncertainty Visualization neural networks Transformers Perturbation methods Training Task analysis Predictive uncertainty gradients contrastive explanations counterfactual explanations neural networks deep learning

来源：评论

学校读者我要写书评

暂无评论

Identification for the Low-Contrast image signal with Regularized Variational Term and Dynamical Saturating Nonlinearity

引用

Journal of Systems Science & Complexity 2023年第3期36卷 1089-1102页

作者： ZHANG Ning MA Yumei PAN Zhenkuan HUANG Baoxiang WANG Dongcheng College of Computer Science&Technology Qingdao UniversityQingdao 266071China

In recent years,image processing based on stochastic resonance(SR)has received more and more *** this paper,a new model combining dynamical saturating nonlinearity with regularized variational term for enhancement of low contrast image is *** regularized variational term can be setting to total variation(TV),second order total generalized variation(TGV)and non-local means(NLM)in order to gradually suppress noise in the process of solving the *** addition,the new model is tested on a mass of gray-scale images from standard test image and low contrast indoor color images from Low-Light dataset(LOL).By comparing the new model and other traditional image enhancement models,the results demonstrate the enhanced image not only obtain good perceptual quality but also get more excellent value of evaluation index compared with some previous methods.

关键词： Dynamical saturating nonlinearity image enhancement low contrast regularized variational term stochastic resonance(SR)

来源：评论

学校读者我要写书评

暂无评论

MULTI-head self-attention-based recurrent neural network with dwarf mongoose optimization algorithm-espoused QRS detector design

引用

signal image AND VIDEO processing 2024年第5期18卷 4935-4944页

作者： Malathi, S. R. Kumar, P. Vijay Sri Venkateswara Coll Engn Dept Elect & Commun Engn Chennai Tamil Nadu India Indian Inst Management Calcutta Dept Business Analyt C7VX VVVDiamond Harbour Rd Kolkata 700104 W Bengal India

QRS detection is the essential electrocardiogram (ECG) analysis procedure. A reliable QRS recognition system that achieves high accuracy despite a typical QRS morphologies and significant noise is required for ECG data gathered by wearable devices. Most contemporary systems seek fast execution durations and minimal energy consumption while attaining high prediction rates. To minimize these issues, this research article presents an approach of multi-head self-attention-based recurrent neural network with dwarf mongoose optimization algorithm-espoused QRS detector design (MHSARNN-QRS) is proposed. Initially, ECG data are taken from MIT/BIH arrhythmia dataset (MIT-AD). Every disintegrated signal is transmitted to multi-head self-attention-based recurrent neural network (MHSARNN) to examine morphologies and predict QRS like correct and incorrect. Then, the QRS wave is located through dwarf mongoose optimization algorithm by reducing probability of neglected identification improves the detection performance. The performance of proposed MHSARNN-QRS method is evaluated using accuracy, sensitivity, specificity, detection error rate, computation time, f1-score, positive prediction, and time for processing a single record (s) and single beat (ms) are analyzed. Performance of the MHSARNN-QRS approach attains high sensitivity, lower single record, lower single beat, and greater accuracy compared with existing methods.

关键词： Dwarf mongoose optimization algorithm Master-slave adaptive notch filter MIT/BIH arrhythmia dataset Multi-head self-attention-based recurrent neural network

来源：评论

学校读者我要写书评

暂无评论

IoT-based nano wireless sensor approach for detection of ships using mixed convolutional neural network approach

引用

signal image AND VIDEO processing 2024年第11期18卷 8185-8194页

作者： Gupta, Vishal Rahmani, Mohammad Khalid Imam NIT Delhi Dept Comp Sci Engn Delhi India Saudi Elect Univ Coll Comp & Informat Riyadh 11673 Saudi Arabia

Ships and other maritime objects are often unable to endure the harsh and dynamic sea environment. Collecting real-time data and detecting these objects using various sensors such as RADARs, Synthetic Aperture RADARs, and mounted RADARs present significant challenges due to numerous influencing factors. To address this issue, our research aims to develop an Internet of Things (IoT)-based multi-scale and multi-scene ship identification system. This system leverages a multi-scale neural network integrated with a high-response convolutional neural network (CNN)-based Kalman filter architecture. To construct this model, we selected various ship categories and initially employed a base CNN model to develop a new model with different convolutional layers. Our approach utilizes mixed methods for tracking and detecting objects, with a focus on small ships. The dataset is processed through multiple neural network layers, and we implemented the Kalman filter to estimate and predict the ships' positions. Additionally, using the YOLOv3 model, we achieved improved accuracy and reduced error rates through mathematical optimization. Our method utilizes a dataset of 5,604 samples and incorporates a hybrid approach with YOLOv3. Our model demonstrates significant improvements for both medium-sized and small ships. The proposed work provides both qualitative and quantitative advancements. Our model exceeded the best results from parallel experiments by 3.9% and 1.2% in terms of Average Precision (AP). Furthermore, YOLOv3 achieved a performance score of 97.34% across various metric parameters, while our proposed approach attained the highest scores of 97.8% and 94.87%, respectively.

关键词： Convolutional neural network Yolo-3 image processing Kalman filter IoT

来源：评论

学校读者我要写书评

暂无评论

NeFF: neural Feature Fields for Few-Shot View Synthesis Addressing the Shape-Radiance Ambiguity

引用

IEEE signal processing LETTERS 2024年 31卷 551-555页

作者： Chang, Yuan Ding, Peng Shen, Yun Liang, Wei Yang, Mingchuan China Telecom Res Inst Beijing 102209 Peoples R China

We present NeFF, a 3D neural scene representation estimated from captured images. neural radiance fields(NeRF) have demonstrated their excellent performance for image based photo-realistic free-viewpoint rendering. However, one limitation of current NeRF based methods is the shape-radiance ambiguity, which means that without any regularization, there may be an incorrect shape that explains the training set very well but that generalizes poorly to novel views. This degeneration becomes particularly evident when fewer input views are provided. We propose an explicit regularization to avoid the ambiguity by introducing the neural Feature Fields which map spatial locations to view-independent features. We synthesize feature maps by projecting the feature fields into images using volume rendering techniques as NeRF does and get an auxiliary loss that encourages the correct view-independent geometry. Experimental results demonstrate that our method has better robustness when dealing with sparse input views.

关键词： image color analysis Rendering (computer graphics) Training Geometry Cameras Mathematical models Feature extraction neural radiance fields rendering view synthesis

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：