检索结果-内蒙古大学图书馆

ieee International symposium on Biomedical Imaging

作者： Shiva Mirzaeian Kyle M. Jensen Adithya Ram Ballem Vince D. Calhoun Armin Iraji Tri-institutional Center for Translational Research in Neuroimaging and Data Science(TReNDS) Atlanta GA Department of Mathematics and Statistics Georgia State University Atlanta GA Department of Computer Science Georgia State University Atlanta GA Department of Psychology Georgia State University Atlanta GA

ISBN: (数字)9798331520526

ISBN: (纸本)9798331520533

Spatial group independent component analysis (sgr-ICA) has become a crucial method to understand brain function in functional magnetic resonance imaging (fMRI) research, especially in resting-state fMRI (rs-fMRI) studies. Early studies using low-order sgr-ICA identified large-scale brain networks, but higher-order sgr-ICA (75–150 components) revealed more refined intrinsic connectivity networks (ICNs). While previous studies employing high-order sgr-ICA models (300 or even 1000 components) were often limited by small datasets, we applied an ultra-high-order ICA model with 500 components to a large rs-fMRI dataset of over 100,000 individuals' selecting approximately 58,000 after quality control. This comprehensive dataset enabled us to delineate the brain into fine-grained intrinsic connectivity networks (ICNs), offering a robust and detailed representation of these networks. We also evaluate functional connectivity differences between healthy people and individuals with schizophrenia. Our results demonstrate that ultra-high-order sgr-ICA provides reliable, precise estimations of ICNs and captures significant group differences between healthy people and schizophrenia patients. These findings highlight the potential of ultra-high-order sgr-ICA for improving rs-fMRI's clinical applications and single-subject analyses.

关键词： Stability criteria Independent component analysis Quality control Organizations Functional magnetic resonance imaging Schizophrenia Brain modeling Reliability Indexes Biomedical imaging

来源：评论

学校读者我要写书评

暂无评论

Driving the Future of Connected Autonomous Mobility with Advanced AI and Multi-Modal data Fusion: Plenary Talk

Driving the Future of Connected Autonomous Mobility with Adv...

引用

International symposium on Applied Machine Intelligence and Informatics (SAMI)

作者： Juraj Gazda Technical University of Košice Košice Slovakia

ISBN: (数字)9798350379365

ISBN: (纸本)9798350379372

In this talk, we will explore the intersection of connected autonomous vehicles (CAVs), edge computing, multi-agent cooperation, computer vision, and large language models (LLMs) to tackle emerging challenges in intelligent transportation. Edge computing enables decentralized, low-latency processing, essential for real-time decision-making in tasks such as obstacle detection and navigation. Multi-agent cooperation, in this context, focuses on efficient sharing of radio and computational resources, ensuring that vehicles can dynamically allocate bandwidth and processing power to maintain high performance in communication and computation-heavy environments. Computer vision technologies enable vehicles to interpret complex surroundings through advanced scene analysis, enhancing safety and decision-making. Additionally, LLMs contribute to processing large amounts of contextual data, enabling more human-like reasoning and complex decision-making. Together, these technologies offer a holistcapproach to advancing the capabilities of autonomous vehicles, creating more robust, intelligent, and sustainable transportation systems.

关键词： Computer vision Decision making Transportation Radio navigation Real-time systems Computational efficiency Safety Autonomous vehicles Machine intelligence Edge computing

来源：评论

学校读者我要写书评

暂无评论

VSS-SAM: Visual State Space-Enhanced SAM for 3D Medical Image Segmentation

VSS-SAM: Visual State Space-Enhanced SAM for 3D Medical Imag...

引用

ieee International symposium on Biomedical Imaging

作者： Jinxuan Lyu Xuhao Dong Bin Zhang Shengping Liu Haifeng Wang Dong Liang Yihang Zhou Research Center for Medical AI Chinese Academy of Sciences Shenzhen Institute of Advanced Technology Shenzhen China School of Pharmacy and Bioengineering Chongqing University of Technology Chongqing China Paul C. Lauterbur Research Center for Biomedical Imaging Chinese Academy of Sciences Shenzhen Institute of Advanced Technology Shenzhen China

ISBN: (数字)9798331520526

ISBN: (纸本)9798331520533

The Segment Anything Model (SAM) is a large general segmentation model proposed by Meta, which has shown amazing performance in many natural image segmentation tasks. Recently, MA-SAM has been proposed to transplant SAM model to medical applications. However, such models still has shortcomings in utilizing the three-dimensional information of medical images. In this paper, we proposed a new structure to fully utilizing 3D information on medical images to enhance the performance of SAM, called VSS-SAM. Particularly, the proposed method applied two branches: 1) SAM as an encoder to learn the basic topology relations among slices, and 2) parallel Mamba as second branch to effectively capture long-range spatial dependencies. Finally, a new decoder was proposed to integrate the multi-view feature representations extracted from the two branches and output the final prediction, aiming to achieve optimal segmentation performance. We validated our method on three publicly available datasets. Experimental results show that the segmentation performance of VSS-SAM is significantly better than that of existing methods in multiple data sets.

关键词： Image segmentation visualization Solid modeling Biomedical equipment Three-dimensional displays Feature extraction Topology Organ transplantation Decoding Biomedical imaging

来源：评论

学校读者我要写书评

暂无评论

A data Perspective on Enhanced Identity Preservation for Diffusion Personalization

A Data Perspective on Enhanced Identity Preservation for Dif...

引用

ieee Workshop on Applications of Computer Vision (WACV)

作者： Xingzhe He Zhiwen Cao Nicholas Kolkin Lantao Yu Kun Wan Helge Rhodin Ratheesh Kalarot University of British Columbia Adobe

ISBN: (数字)9798331510831

ISBN: (纸本)9798331510848

large text-to-image models have revolutionized the ability to generate imagery using natural language. However, particularly unique or personal visual concepts, such as pets and furniture, will not be captured by the original model. This has led to interest in how to personalize a text-to-image model. Despite significant progress, this task remains a formidable challenge, particularly in preserving the subject's identity. Most researchers attempt to address this issue by modifying model architectures. These methods are capable of keeping the subject structure and color but fail to preserve identity details. Towards this issue, our approach takes a data-centric perspective. We introduce a novel regularization dataset generation strategy on both the text and image level. This strategy enables the model to preserve fine details of the desired subjects, such as text and logos. Our method is architecture-agnostic and can be flexibly applied on various text-to-image models. We show on established benchmarks that our data-centric approach forms the new state of the art in terms of identity preservation and text alignment.

关键词： Training Image quality visualization Image color analysis Natural languages Memory management Text to image Diffusion models Videos Overfitting

来源：评论

学校读者我要写书评

暂无评论

Cross-D Conv: Cross-Dimensional Transferable Knowledge Base via Fourier Shifting Operation

Cross-D Conv: Cross-Dimensional Transferable Knowledge Base ...

引用

ieee International symposium on Biomedical Imaging

作者： Mehmet Can Yavuz Yang Yang Department of Radiology and Biomedical Imaging University of California San Francisco USA

ISBN: (数字)9798331520526

ISBN: (纸本)9798331520533

In biomedical imaging analysis, the dichotomy between 2D and 3D data presents a significant challenge. While 3D volumes offer superior real-world applicability, they are less available for each modality and not easy to train in large scale, whereas 2D samples are abundant but less comprehensive. This paper introduces Cross-D Conv operation, a novel approach that bridges the dimensional gap by learning the phase shifting in the Fourier domain. Our method enables seamless weight transfer between 2D and 3D convolution operations, effectively facilitating cross-dimensional learning. The proposed architecture leverages the abundance of 2D training data to enhance 3D model performance, offering a practical solution to the multimodal data scarcity challenge in 3D medical model pretraining. Experimental validation on the RadImagenet (2D) and multimodal volumetric sets demonstrates that our approach achieves comparable or superior performance in feature quality assessment. The enhanced convolution operation presents new opportunities for developing efficient classification and segmentation models in medical imaging. This work represents an advancement in cross-dimensional and multimodal medical image analysis, offering a robust framework for utilizing 2D priors in 3D model pretraining while maintaining computational efficiency of 2D training. Codes & Weights

关键词： Training Solid modeling Image segmentation Three-dimensional displays Convolution Computational modeling Transfer learning Knowledge based systems data models Biomedical imaging

来源：评论

学校读者我要写书评

暂无评论

Revisiting Clip: Efficient Alignment of 3D MRI and Tabular data Using Domain-Specific Foundation Models

Revisiting Clip: Efficient Alignment of 3D MRI and Tabular D...

引用

ieee International symposium on Biomedical Imaging

作者： Jakob Krogh Petersen Valdemar Licht Mads Nielsen Asbjørn Munk Department of Computer Science Pioneer Center for AI Denmark University of Copenhagen Denmark

ISBN: (数字)9798331520526

ISBN: (纸本)9798331520533

Multi-modal models require aligned, shared embedding spaces. However, common CLIP-based approaches need large amounts of samples and do not natively support 3D or tabular data, both of which are crucial in the medical domain. To address these issues, we revisit CLIP-style alignment by training a domain-specific 3D foundation model as an image encoder and demonstrate that modality alignment is feasible with only 62 MRI scans. Our approach is enabled by a simple embedding accumulation strategy required for training in 3D, which scales the amount of negative pairs across batches in order to stabilize training. We perform a thorough evaluation of various design choices, including the choice of backbone and loss functions, and evaluate the proposed methodology on zero-shot classification and image-retrieval tasks. While zero-shot image-retrieval remains challenging, zero-shot classification results demonstrate that the proposed approach can meaningfully align the representations of 3D MRI with tabular data. Code and model checkpoints are available here.

关键词： Training Three-dimensional displays Codes Foundation models Magnetic resonance imaging Natural languages Contrastive learning Stability analysis data models Biomedical imaging

来源：评论

学校读者我要写书评

暂无评论

Comparative analysis and Evaluation of Well-Being Activity-Infused Fine-Tuned Language Models with Benchmark Models

Comparative Analysis and Evaluation of Well-Being Activity-I...

引用

Computational Intelligence in Health and Medicine (CIHM), ieee symposium on

作者： Hamza Haruna Mohammed Gabriel Kiss J. Artur Serrano Frank Lindseth Department of Computer Science Faculty of Information Technology and Electrical Engineering Norwegian University of Science and Technology Trondheim Norway Department of Neuromedicine and Movement Science Faculty of Medicine and Health Sciences Norwegian University of Science and Technology Trondheim Norway

ISBN: (数字)9798331508333

ISBN: (纸本)9798331508340

large language models (LLMs) have significantly impacted natural language processing (NLP), healthcare, and beyond fields. LLMs have demonstrated unprecedented abilities to understand and generate human-like text, making them highly valuable for various tasks, including information retrieval, question-answering, and conversational AI. However, as these models are integrated into real-world applications, especially in sensitive domains such as healthcare, it becomes essential to evaluate their performance comprehensively. We present a process for fine-tuning LLMs to incorporate cardiac exercise and corpus of well-being activity, aiming to promote activity health support through NLP-driven applications. Using QLoRA, you can ensure that the model size remains small and reduce overhead while maintaining the high performance of the fine-tuned model. Subsequently, this paper presents a comparative analysis benchmark pipeline to evaluate fine-tuned models with base models on standard benchmark datasets from the healthcare domain. The aim is to evaluate the performance, robustness and accuracy of these models, showing that the fine-tuning process does not degrade the model performance, but improves the performance to some extent in some benchmarks.

关键词： Analytical models Semantic search large language models Computational modeling Medical services Benchmark testing Natural language processing data models Robustness Standards

来源：评论

学校读者我要写书评

暂无评论

Refined Urban Informal Settlements Mapping at Agglomeration Scale With the Guidance of Background Knowledge From Easy-Accessed Crowdsourced Geospatial data

引用

ieee TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 2025年 63卷

作者： Fan, Runyu Niu, Hongyang Xu, Zijian Chen, Jiajun Feng, Ruyi Wang, Lizhe China Univ Geosci Sch Comp Sci Wuhan 430074 Peoples R China China Univ Geosci Hubei Key Lab Intelligent Geoinformat Proc Wuhan 430074 Peoples R China

Urban informal settlements (UISs) denote densely populated locales characterized by inadequate urban infrastructure standards, often exhibiting an amalgamation of rural and urban attributes, primarily situated within the confines of major cities or metropolitan regions. UIS mapping is a typical task that aims to identify pixels corresponding to UISs in remote sensing images. The extremely similar visual characteristics, sample uncertainty, and highly manual costs bring large-scale UIS mapping noteworthy challenges. In this article, we propose a sample uncertainty-aware semisupervised learning method guided by background knowledge from crowdsourced geospatial data (namely, SemiUIS) for UIS mapping and produces a 1-m resolution UIS map in the urban agglomeration in the middle reaches of the Yangtze River (UAMRYR), China. Our proposed SemiUIS method integrates semisupervised learning with the guidance of background knowledge from crowdsourcing data to reduce the uncertainty of the generated pseudosemantic segmentation samples to improve the quality of the constructed dataset by jointly optimizing the background (Not UIS areas) and the foreground (UIS areas). In contrast, traditional semisupervised learning methods only focus on explicitly optimizing the foreground (UIS areas) in the pseudosample generation process. Experiments were conducted in the 31 prefecture-level cities of UAMRYR. Visual interpretation of labeled remote sensing samples, street view images, and two novel knowledge validation indicators, global knowledge precision (GKP) and local knowledge precision (LKP), are proposed to verify the UIS mapping results. The proposed method reached an overall accuracy (OA) of 90.58%, mean intersection over union (mIoU) of 75.88%, accuracy (Acc) of 78.77%, Intersection over Union (IoU) of 62.97%, GKP of 90.33%, and LKP of 85.01%, respectively, with only 36 (5%) training samples. This work will be available at https://***/RunyuFan/UisYangtze.

关键词： Remote sensing Semantic segmentation Accuracy Semisupervised learning Feature extraction visualization Geospatial analysis Costs Uncertainty Crowdsourcing Crowdsourced geospatial data deep learning (DL) remote sensing semisupervised learning urban informal settlement (UIS) mapping

来源：评论

学校读者我要写书评

暂无评论

Leveraging large-Scale TLC 3D NAND Flash Characterization to Investigate Early Lifetime Reliability of SSDs through Extreme Value Statistics

Leveraging Large-Scale TLC 3D NAND Flash Characterization to...

引用

Annual International symposium on Reliability Physics

作者： Rino Micheloni Luca Crippa Alessia Marelli Lorenzo Zuolo Piero Olivo Cristian Zambelli Avaneidi SpA Saronno VA Italy Dipartimento di Ingegneria Università degli Studi di Ferrara Ferrara Italy

ISBN: (数字)9798331504779

ISBN: (纸本)9798331504786

Solid State Drives based on the Triple Level Cell 3D NAND Flash technology are becoming the state-of-the-art in storage architecture. Their reliability analysis is a must. This work presents: i) a large-scale statistical characterization of the errors in the codewords of a TLC 3D NAND Flash Solid State Drive and ii) provides an extreme value statistics model based on the Points Over Threshold approach to capture the early lifetime variations. The model provides a valuable tool for storage designers to implement reliability management strategies since the early drive deployment in the field.

关键词： Solid modeling Three-dimensional displays Solid state drives Triples (data structure) Computer architecture Predictive models Reliability engineering Reliability Weibull distribution Flash memories

来源：评论

学校读者我要写书评

暂无评论

Personalized Content-based Movie Recommendation System: A Comparative analysis

Personalized Content-based Movie Recommendation System: A Co...

引用

2025 International Conference on Visual Analytics and data visualization, ICVADV 2025

作者： Nithya, B. Asha, V. Sreeja, S.P. Sinha, Abhay Aditya Akash, P. Akshay Bharadwaj, S. New Horizon College of Engineering Department of Master of Computer Applications Bengaluru India

ISBN: (纸本)9798331521394

A recommendation system is an analytical tool designed to provide personalized suggestions to users for specific resources such as literature, films, or music, derived from a comprehensive dataset. Movie recommendation systems aim to predict movies a user might enjoy based on attributes of previously liked movies. These systems are particularly advantageous for organizations that manage data from extensive customer bases and seek to deliver optimal suggestions. Factors such as genre, cast, and filmmakers significantly influence recommendations, which may be based on a single attribute or a combination of features. This paper focuses on a genre-based recommendation system employing content-based filtering to cater to user preferences. Using Python, the design tools algorithms like Cosine Similarity, Jaccard Similarity, and Pearson Correlation on a large dataset to identify patterns and preferences eIectively. Through a detailed evaluation of these approaches, the system provides robust and applicable movie suggestions. By smoothing user-friendly navigation and aligning recommendations with user interests, the system enhances user engagement and satisfaction. Also, this study examines the strengths and limitations of the applied algorithms, emphasizing the implicit for combining multiple similarity measures to enhance accuracy. The findings emphasize the significance of substantiated content delivery in optimizing user experience. © 2025 ieee.

关键词： Recommender systems

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：