Procedia Computer Science最新文献_第10页

Sign Language To Sign Language Translator 手语到手语翻译

Procedia Computer Science

Pub Date : 2025-01-01 DOI: 10.1016/j.procs.2025.03.213

Sharath S R , Suraj S , Abishek Kumar G M , P Siddharth , Nalinadevi K

Sign languages differ between countries, regions, and even communities within the same country, leading to communication barriers when interacting with deaf individuals from different linguistic backgrounds. This paper introduces a novel approach for sign language-to-sign language translation, enabling seamless communication across diverse deaf communities. The proposed model translates source sign language images to avatar sign images of the target language by utilizing separate key point estimation models for recognizing static sign elements that include handshape, orientation and position, achieving an accuracy of 88%. The research work uses HamNoSys as the intermediate representation to capture the essential elements of signs in the translation process. The HamNoSys sequence migration task is accomplished using the Seq2Seq model with a BLEU-1 score of 0.85. The target HamNoSys sequences are converted to machine-readable format (SiGML) to render the 3D avatar sign images. Experiments are done using static signs from three distinct sign languages.

手语在不同国家、地区甚至同一国家的社区之间都存在差异，这导致与不同语言背景的聋人交流时存在沟通障碍。本文介绍了一种新的手语到手语翻译方法，使不同聋人社区之间的无缝沟通成为可能。该模型利用独立的关键点估计模型识别手部形状、方向和位置等静态符号元素，将源手语图像转换为目标语言的化身手势图像，准确率达到88%。本研究使用HamNoSys作为中介表征来捕捉翻译过程中符号的基本要素。HamNoSys序列迁移任务使用BLEU-1分数为0.85的Seq2Seq模型完成。目标HamNoSys序列被转换为机器可读格式（SiGML），以呈现3D头像符号图像。实验使用了三种不同手语的静态符号。

引用次数: 0

Federated Learning in Detecting Fake News: A Survey 联邦学习在假新闻检测中的应用研究

Procedia Computer Science

Pub Date : 2025-01-01 DOI: 10.1016/j.procs.2025.03.223

Sri Vasavi Chandu , Uma Sankararao Varri , Vamshi A , Vinay Raj

Due to technological advancements, social media usage has increased a lot resulting in a huge spread of fake information and false news among users of different languages. To reduce the spread of fake information, there is a need to detect the fake/false information being posted on social media apps like Twitter, Facebook, Instagram, and many. In order to identify false news, researchers employ models based on machine learning, natural language processing, and deep learning. These models are to be trained initially by huge amounts of data so that the models can gain knowledge from the trained data and predict the output for the new data provided. This study performs a detailed systematic review on different recent federated learning models being proposed for detecting fake news. It provides a detailed comparison of recently published articles related to fake-news detection using federated learning in terms of models they used. This study also provides different datasets which can be used in detecting fake-news using federated learning.

由于技术的进步，社交媒体的使用增加了很多，导致虚假信息和虚假新闻在不同语言的用户中大量传播。为了减少虚假信息的传播，有必要检测Twitter、Facebook、Instagram等社交媒体应用程序上发布的虚假/虚假信息。为了识别假新闻，研究人员采用了基于机器学习、自然语言处理和深度学习的模型。这些模型最初将通过大量数据进行训练，以便模型可以从训练过的数据中获得知识，并预测所提供的新数据的输出。本研究对最近提出的用于检测假新闻的不同联邦学习模型进行了详细的系统回顾。它提供了最近发表的关于使用联邦学习检测假新闻的文章的详细比较。本研究还提供了不同的数据集，可用于使用联邦学习检测假新闻。

引用次数: 0

LakotaBERT: Transformer based model for Low Resource Lakota Language LakotaBERT：基于转换器的低资源Lakota语言模型

Procedia Computer Science

Pub Date : 2025-01-01 DOI: 10.1016/j.procs.2025.03.226

Kanishka Parankusham , Rodrigue Rizk , K C Santosh

Lakota, a critically endangered language of the Sioux people in North America, faces significant challenges due to declining fluency among younger generations. This paper presents the development of LakotaBERT, the first large language model (LLM) tailored for Lakota, aiming to support language revitalization efforts. Our research has two primary objectives: (1) to create a comprehensive Lakota language corpus and (2) to develop a customized LLM for Lakota. We compiled a diverse corpus of 105K sentences in Lakota, English, and parallel texts from various sources, such as books and websites, emphasizing the cultural significance and historical context of the Lakota language. Utilizing the RoBERTa architecture, we pre-trained our model and conducted comparative evaluations against established models such as RoBERTa, BERT, and multilingual BERT. Initial results demonstrate a masked language modeling accuracy of 51% with a single ground truth assumption, showcasing performance comparable to that of English-based models. We also evaluated the model using additional metrics, such as precision and F1 score, to provide a comprehensive assessment of its capabilities. By integrating AI and linguistic methodologies, we aspire to enhance linguistic diversity and cultural resilience, setting a valuable precedent for leveraging technology in the revitalization of other endangered indigenous languages.

拉科塔语是北美苏族的一种濒危语言，由于年轻一代的流利程度下降，它面临着巨大的挑战。本文介绍了LakotaBERT的开发，这是为Lakota量身定制的第一个大型语言模型（LLM），旨在支持语言振兴工作。我们的研究有两个主要目标：(1)创建一个全面的拉科塔语言语料库；(2)为拉科塔人开发一个定制的法学硕士。我们编制了一个包括拉科塔语、英语和各种来源（如书籍和网站）的平行文本的105K个句子的多样化语料库，强调拉科塔语的文化意义和历史背景。利用RoBERTa架构，我们预先训练了我们的模型，并与RoBERTa、BERT和多语言BERT等已建立的模型进行了比较评估。初步结果表明，在单一基础真值假设下，掩蔽语言建模准确率为51%，其性能与基于英语的模型相当。我们还使用其他指标（如精度和F1分数）对模型进行了评估，以提供对其功能的全面评估。通过整合人工智能和语言方法，我们希望加强语言多样性和文化复原力，为利用技术振兴其他濒危土著语言树立一个宝贵的先例。

{"title":"LakotaBERT: Transformer based model for Low Resource Lakota Language","authors":"Kanishka Parankusham , Rodrigue Rizk , K C Santosh","doi":"10.1016/j.procs.2025.03.226","DOIUrl":"10.1016/j.procs.2025.03.226","url":null,"abstract":"<div><div>Lakota, a critically endangered language of the Sioux people in North America, faces significant challenges due to declining fluency among younger generations. This paper presents the development of LakotaBERT, the first large language model (LLM) tailored for Lakota, aiming to support language revitalization efforts. Our research has two primary objectives: (1) to create a comprehensive Lakota language corpus and (2) to develop a customized LLM for Lakota. We compiled a diverse corpus of 105K sentences in Lakota, English, and parallel texts from various sources, such as books and websites, emphasizing the cultural significance and historical context of the Lakota language. Utilizing the RoBERTa architecture, we pre-trained our model and conducted comparative evaluations against established models such as RoBERTa, BERT, and multilingual BERT. Initial results demonstrate a masked language modeling accuracy of 51% with a single ground truth assumption, showcasing performance comparable to that of English-based models. We also evaluated the model using additional metrics, such as precision and F1 score, to provide a comprehensive assessment of its capabilities. By integrating AI and linguistic methodologies, we aspire to enhance linguistic diversity and cultural resilience, setting a valuable precedent for leveraging technology in the revitalization of other endangered indigenous languages.</div></div>","PeriodicalId":20465,"journal":{"name":"Procedia Computer Science","volume":"260 ","pages":"Pages 486-497"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144139276","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

IoT-Driven Smart Farming with Machine Learning for Sustainable Food Systems 物联网驱动的智能农业与可持续粮食系统的机器学习

Procedia Computer Science

Pub Date : 2025-01-01 DOI: 10.1016/j.procs.2025.03.233

Sanjana Murgod , Tanushree Kabbur , Bibijan Matte , Vaibhav Mujumdar , Meenaxi M Raikar

Integrating the Internet of Things (IoT) and machine learning (ML) technologies in agriculture, commonly called smart farming, is revolutionizing the sector by enhancing productivity, efficiency, and sustainability. This paper explores the application of IoT-driven smart farming using machine learning for sustainable agricultural practices. The system introduces an efficient Soil Moisture Detection System utilizing IoT Technology, revolutionizing modern farming practices. By continuously monitoring crucial parameters such as soil moisture, temperature, and humidity in real-time, the system ensures seamless data transmission to a centralized server. Additionally, integrating motion detection capabilities enhances security measures and promptly alerts farmers to environmental changes. The dataset consisting of 100,000 rows is generated to facilitate the development and training of five ML models to predict soil moisture trends. Decision Trees achieved an accuracy rate of 99.98%, while Random Forests achieved 99.99%. The integration of these predictive models empowers farmers with actionable insights for precise irrigation scheduling and optimal crop yield optimization. These models provide actionable insights for precise irrigation scheduling and optimal crop yield optimization. Field tests have confirmed the efficacy of this approach, demonstrating significant improvements in irrigation efficiency and subsequent crop yields. Thus, the proposed system represents a substantial advancement in leveraging the synergistic potential of IoT and ML technologies to foster sustainable agricultural practices.

将物联网（IoT）和机器学习（ML）技术整合到农业中，通常被称为智能农业，通过提高生产力、效率和可持续性，正在彻底改变该行业。本文探讨了物联网驱动的智能农业应用，利用机器学习实现可持续农业实践。该系统引入了利用物联网技术的高效土壤湿度检测系统，彻底改变了现代农业实践。通过持续实时监测土壤湿度、温度、湿度等关键参数，确保数据无缝传输到中央服务器。此外，集成运动检测功能增强了安全措施，并及时提醒农民注意环境变化。生成100000行数据集，以促进5个ML模型的开发和训练，以预测土壤湿度趋势。决策树的准确率为99.98%，而随机森林的准确率为99.99%。这些预测模型的整合为农民提供了精确灌溉调度和最佳作物产量优化的可行见解。这些模型为精确的灌溉调度和最佳作物产量优化提供了可行的见解。田间试验证实了这种方法的有效性，表明在灌溉效率和随后的作物产量方面有了显著改善。因此，拟议的系统在利用物联网和机器学习技术的协同潜力促进可持续农业实践方面取得了实质性进展。

{"title":"IoT-Driven Smart Farming with Machine Learning for Sustainable Food Systems","authors":"Sanjana Murgod , Tanushree Kabbur , Bibijan Matte , Vaibhav Mujumdar , Meenaxi M Raikar","doi":"10.1016/j.procs.2025.03.233","DOIUrl":"10.1016/j.procs.2025.03.233","url":null,"abstract":"<div><div>Integrating the Internet of Things (IoT) and machine learning (ML) technologies in agriculture, commonly called smart farming, is revolutionizing the sector by enhancing productivity, efficiency, and sustainability. This paper explores the application of IoT-driven smart farming using machine learning for sustainable agricultural practices. The system introduces an efficient Soil Moisture Detection System utilizing IoT Technology, revolutionizing modern farming practices. By continuously monitoring crucial parameters such as soil moisture, temperature, and humidity in real-time, the system ensures seamless data transmission to a centralized server. Additionally, integrating motion detection capabilities enhances security measures and promptly alerts farmers to environmental changes. The dataset consisting of 100,000 rows is generated to facilitate the development and training of five ML models to predict soil moisture trends. Decision Trees achieved an accuracy rate of 99.98%, while Random Forests achieved 99.99%. The integration of these predictive models empowers farmers with actionable insights for precise irrigation scheduling and optimal crop yield optimization. These models provide actionable insights for precise irrigation scheduling and optimal crop yield optimization. Field tests have confirmed the efficacy of this approach, demonstrating significant improvements in irrigation efficiency and subsequent crop yields. Thus, the proposed system represents a substantial advancement in leveraging the synergistic potential of IoT and ML technologies to foster sustainable agricultural practices.</div></div>","PeriodicalId":20465,"journal":{"name":"Procedia Computer Science","volume":"260 ","pages":"Pages 552-560"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144139087","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Fuzzy Set of Rules for Optimal Adaptive Selection of OFDM order And FFT Size for NB-IOT NB-IOT OFDM顺序和FFT大小最优自适应选择的模糊规则集

Procedia Computer Science

Pub Date : 2025-01-01 DOI: 10.1016/j.procs.2025.03.265

Amrita Khera , Susheel Kumar Gupta , Urwashi Sad , Sandeep Sahu , Hemant Choubey

The huge growth is being observed in case of Narrow-Band-IOT (NB-IOT) based futuristic communication technology. The sharing of spectrum is the primary benefit. The system makes use of the 5G mobile communication spectrum that is freely available. When using any kind of communication technology, the physical layer is mostly in charge of data communication. The objective of this work is to enhance the effectiveness of OFDM, which serves as the physical layer of the NB-IOT, a futuristic approach. In the end, fast Fourier transform (FFT) is used to generate orthogonal frequencies, which forms the foundation of OFDM. Increasing the modulation order also addresses the excess signal transition need. In order to construct an efficient OFDM system, the goal of this research is to define a fuzzy set of guidelines for the best selection of the modulation order and the FFT size. These fuzzily defined criteria might provide improved capacity and bandwidth efficiency. A is taken into account to deal with these issues. Selecting the lower order modulation and ideal FFT size in accordance with data need is suggested. The BER parameter is chosen for performance assessment, considering modulation parameters, FFT sizes, and signal duration’s impacts. The research suggests increasing FFT size and order modulation to accommodate increased capacity and demand, while comparing the BER efficiency of M-QAM and M-PSK modulation schemes.

以窄带物联网（NB-IOT）为基础的未来通信技术将出现巨大增长。频谱共享是主要的好处。该系统利用了免费提供的5G移动通信频谱。在使用任何一种通信技术时，物理层主要负责数据通信。这项工作的目标是提高OFDM的有效性，OFDM作为NB-IOT的物理层，是一种未来的方法。最后，利用快速傅里叶变换（FFT）产生正交频率，这是OFDM的基础。增加调制阶数也解决了多余的信号转换需求。为了构建一个高效的OFDM系统，本研究的目标是定义一组模糊准则来最佳选择调制阶数和FFT大小。这些模糊定义的标准可能提供改进的容量和带宽效率。在处理这些问题时考虑到A。建议根据数据需要选择低阶调制和理想FFT大小。考虑到调制参数、FFT大小和信号持续时间的影响，选择误码率参数进行性能评估。研究建议增加FFT的大小和顺序调制以适应增加的容量和需求，同时比较M-QAM和M-PSK调制方案的BER效率。

{"title":"Fuzzy Set of Rules for Optimal Adaptive Selection of OFDM order And FFT Size for NB-IOT","authors":"Amrita Khera , Susheel Kumar Gupta , Urwashi Sad , Sandeep Sahu , Hemant Choubey","doi":"10.1016/j.procs.2025.03.265","DOIUrl":"10.1016/j.procs.2025.03.265","url":null,"abstract":"<div><div>The huge growth is being observed in case of Narrow-Band-IOT (NB-IOT) based futuristic communication technology. The sharing of spectrum is the primary benefit. The system makes use of the 5G mobile communication spectrum that is freely available. When using any kind of communication technology, the physical layer is mostly in charge of data communication. The objective of this work is to enhance the effectiveness of OFDM, which serves as the physical layer of the NB-IOT, a futuristic approach. In the end, fast Fourier transform (FFT) is used to generate orthogonal frequencies, which forms the foundation of OFDM. Increasing the modulation order also addresses the excess signal transition need. In order to construct an efficient OFDM system, the goal of this research is to define a fuzzy set of guidelines for the best selection of the modulation order and the FFT size. These fuzzily defined criteria might provide improved capacity and bandwidth efficiency. A is taken into account to deal with these issues. Selecting the lower order modulation and ideal FFT size in accordance with data need is suggested. The BER parameter is chosen for performance assessment, considering modulation parameters, FFT sizes, and signal duration’s impacts. The research suggests increasing FFT size and order modulation to accommodate increased capacity and demand, while comparing the BER efficiency of M-QAM and M-PSK modulation schemes.</div></div>","PeriodicalId":20465,"journal":{"name":"Procedia Computer Science","volume":"260 ","pages":"Pages 839-846"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144139099","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Leveraging Ensembles of Pre-trained CNNs for Improved Lung Cancer Detection and Classification 利用预训练cnn集合改进肺癌检测和分类

Procedia Computer Science

Pub Date : 2025-01-01 DOI: 10.1016/j.procs.2025.03.188

Dasari Bhulakshmi , Dharmendra singh rajput

Lung cancer is a serious global health concern, highlighting the importance of early identification to improve patient survival rates. We explore the potential of deep learning(DL) models to improve lung cancer diagnosis through detection and classification models. The performance of pre-trained ResNet50, VGG19, and AlexNet models is evaluated on an augmented lung cancer image dataset to determine their suitability for lung cancer classification. The fine-tuned models are evaluated for their ability to identify and classify lung cancer, achieving high accuracy of 92.88%, 93.06%, and 95.23%. While promising, this approach has limitations. The efficacy of DL models is significantly influenced by both the quality and volume of the training data. Additionally, the ”black box” nature of DL models can make it challenging to understand their decision-making process. However, the results of this study suggest that DL ensembles hold significant potential for lung cancer diagnosis. Further research is necessary to address limitations and explore interpretability techniques for wider clinical acceptance.

肺癌是一个严重的全球健康问题，突出了早期识别对提高患者存活率的重要性。我们探索深度学习（DL）模型通过检测和分类模型来提高肺癌诊断的潜力。在增强的肺癌图像数据集上评估预训练的ResNet50、VGG19和AlexNet模型的性能，以确定它们对肺癌分类的适用性。对微调模型的肺癌识别和分类能力进行了评估，准确率分别为92.88%、93.06%和95.23%。这种方法虽然很有前途，但也有局限性。深度学习模型的有效性受到训练数据的质量和数量的显著影响。此外，深度学习模型的“黑箱”特性使理解其决策过程变得具有挑战性。然而，本研究的结果表明DL集合在肺癌诊断中具有重要的潜力。为了更广泛的临床接受，需要进一步的研究来解决局限性和探索可解释性技术。

引用次数: 0

Brain Tissue Segmentation from MRI Scans using Digital Image Processing 利用数字图像处理从MRI扫描中分割脑组织

Procedia Computer Science

Pub Date : 2025-01-01 DOI: 10.1016/j.procs.2025.03.174

Sushmita Chauhan , Poonam Saini , Sanjeev Sofat

The brain is one of the most unexplored parts of the human body and its complex and delicate structure has scientists worldwide looking for answers about its intricacies. Also, since the advent of deep learning techniques as well as imaging techniques like computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET), analysis of the brain has become the most intriguing and researched area in healthcare, as well as deep learning sectors of artificial intelligence. The extraction of the brain from the skull forms the basis and source of study for the prediction of age-related diseases like Alzheimer’s disease. Nowadays With the increase in life expectancy and the extravagant use of technology, it is evident that neurological diseases are on the rise. Therefore, it becomes essential that such diseases can be diagnosed at an early stage of their occurrence. The proposed work presents brain extraction from the skull with the help of three basic steps, data acquisition, pre-processing, and largest connected component extraction using contours. The data acquired is using the ADNI data repository. The preprocessing step involves contrast enhancement using CLAHE, binarization of the scan using Otsu thresholding, and de-blurring so that the noise that might be there in the scans can be removed and a clear image of the brain is available for further processing and classification of Alzheimer’s disease.

大脑是人体中最未被探索的部分之一，其复杂而微妙的结构使全世界的科学家都在寻找其复杂性的答案。此外，由于深度学习技术以及计算机断层扫描（CT），磁共振成像（MRI）和正电子发射断层扫描（PET）等成像技术的出现，大脑分析已成为医疗保健以及人工智能深度学习领域中最有趣和研究的领域。从头骨中提取大脑构成了预测老年痴呆症等与年龄有关疾病的研究基础和来源。如今，随着预期寿命的延长和技术的过度使用，神经系统疾病明显呈上升趋势。因此，在这些疾病发生的早期阶段进行诊断变得至关重要。提出的工作是在三个基本步骤的帮助下，从颅骨中提取大脑，数据采集，预处理和使用轮廓提取最大连接分量。获取的数据使用ADNI数据存储库。预处理步骤包括使用CLAHE增强对比度，使用Otsu阈值法对扫描进行二值化，以及去模糊，这样扫描中可能存在的噪音就可以被去除，清晰的大脑图像可以用于进一步处理和老年痴呆症的分类。

{"title":"Brain Tissue Segmentation from MRI Scans using Digital Image Processing","authors":"Sushmita Chauhan , Poonam Saini , Sanjeev Sofat","doi":"10.1016/j.procs.2025.03.174","DOIUrl":"10.1016/j.procs.2025.03.174","url":null,"abstract":"<div><div>The brain is one of the most unexplored parts of the human body and its complex and delicate structure has scientists worldwide looking for answers about its intricacies. Also, since the advent of deep learning techniques as well as imaging techniques like computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET), analysis of the brain has become the most intriguing and researched area in healthcare, as well as deep learning sectors of artificial intelligence. The extraction of the brain from the skull forms the basis and source of study for the prediction of age-related diseases like Alzheimer’s disease. Nowadays With the increase in life expectancy and the extravagant use of technology, it is evident that neurological diseases are on the rise. Therefore, it becomes essential that such diseases can be diagnosed at an early stage of their occurrence. The proposed work presents brain extraction from the skull with the help of three basic steps, data acquisition, pre-processing, and largest connected component extraction using contours. The data acquired is using the ADNI data repository. The preprocessing step involves contrast enhancement using CLAHE, binarization of the scan using Otsu thresholding, and de-blurring so that the noise that might be there in the scans can be removed and a clear image of the brain is available for further processing and classification of Alzheimer’s disease.</div></div>","PeriodicalId":20465,"journal":{"name":"Procedia Computer Science","volume":"260 ","pages":"Pages 32-39"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144139324","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Advances in AI for Medical Imaging: A Review of Machine and Deep Learning in Disease Detection 人工智能在医学成像中的进展：疾病检测中的机器和深度学习综述

Procedia Computer Science

Pub Date : 2025-01-01 DOI: 10.1016/j.procs.2025.03.201

Rnjai Lamba

This review provides an exhaustive overview of the impact of machine learning (ML) and deep learning (DL) methods on medical imaging. This paper focuses on how AI is revolutionizing the field of disease detection and diagnosis. These advances have enhanced the precision and ability to diagnose various medical conditions, including cancer, neurological diseases, and retinal disorders. Autonomous ML and DL techniques have enhanced the accuracy of diagnostic processes while simplifying them, eliminating potential human errors, and supporting better clinical judgment by automating intricate image processing functions. The paper presents a comprehensive analysis of the main techniques of ML and DL, such as Support Vector Machines (SVM), Random Forests, Convolutional Neural Networks (CNNs), and Generative Adversarial Networks (GANs). Numerous case studies have demonstrated the remarkable accuracy of these techniques when compared with traditional diagnostic methods. However, the broad adoption of these techniques in medical imaging faces obstacles because of low data quality, lack of interpretability of models, and the requirement of additional computational resources. These problems can be mitigated by creating interpretable AI systems, optimizing the efficiency of computational resources, and establishing ethical guidelines for utilizing these algorithms in healthcare. The review concludes by evaluating the potential of these technologies to transform individualized treatment and healthcare delivery. Ongoing collaboration among technologists, healthcare practitioners, and policy specialists is necessary to guarantee the responsible assimilation of AI into clinical practice.

本文综述了机器学习（ML）和深度学习（DL）方法对医学成像的影响。本文重点介绍了人工智能如何彻底改变疾病检测和诊断领域。这些进步提高了诊断各种疾病的精度和能力，包括癌症、神经系统疾病和视网膜疾病。自主机器学习和深度学习技术提高了诊断过程的准确性，同时简化了诊断过程，消除了潜在的人为错误，并通过自动化复杂的图像处理功能来支持更好的临床判断。本文全面分析了ML和DL的主要技术，如支持向量机（SVM）、随机森林、卷积神经网络（cnn）和生成对抗网络（gan）。许多案例研究表明，与传统诊断方法相比，这些技术具有显著的准确性。然而，由于数据质量低、模型缺乏可解释性以及需要额外的计算资源，这些技术在医学成像中的广泛采用面临障碍。这些问题可以通过创建可解释的人工智能系统、优化计算资源的效率以及为在医疗保健中使用这些算法建立道德准则来缓解。本综述最后评估了这些技术在改变个体化治疗和医疗保健服务方面的潜力。技术专家、医疗从业人员和政策专家之间的持续合作是必要的，以确保人工智能负责任地融入临床实践。

{"title":"Advances in AI for Medical Imaging: A Review of Machine and Deep Learning in Disease Detection","authors":"Rnjai Lamba","doi":"10.1016/j.procs.2025.03.201","DOIUrl":"10.1016/j.procs.2025.03.201","url":null,"abstract":"<div><div>This review provides an exhaustive overview of the impact of machine learning (ML) and deep learning (DL) methods on medical imaging. This paper focuses on how AI is revolutionizing the field of disease detection and diagnosis. These advances have enhanced the precision and ability to diagnose various medical conditions, including cancer, neurological diseases, and retinal disorders. Autonomous ML and DL techniques have enhanced the accuracy of diagnostic processes while simplifying them, eliminating potential human errors, and supporting better clinical judgment by automating intricate image processing functions. The paper presents a comprehensive analysis of the main techniques of ML and DL, such as Support Vector Machines (SVM), Random Forests, Convolutional Neural Networks (CNNs), and Generative Adversarial Networks (GANs). Numerous case studies have demonstrated the remarkable accuracy of these techniques when compared with traditional diagnostic methods. However, the broad adoption of these techniques in medical imaging faces obstacles because of low data quality, lack of interpretability of models, and the requirement of additional computational resources. These problems can be mitigated by creating interpretable AI systems, optimizing the efficiency of computational resources, and establishing ethical guidelines for utilizing these algorithms in healthcare. The review concludes by evaluating the potential of these technologies to transform individualized treatment and healthcare delivery. Ongoing collaboration among technologists, healthcare practitioners, and policy specialists is necessary to guarantee the responsible assimilation of AI into clinical practice.</div></div>","PeriodicalId":20465,"journal":{"name":"Procedia Computer Science","volume":"260 ","pages":"Pages 262-273"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144139365","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Deep CNN-Based Multi-Grade Brain Tumor Classification with Enhanced Data Augmentation 基于cnn的深度多级别脑肿瘤分类与增强数据增强

Procedia Computer Science

Pub Date : 2025-01-01 DOI: 10.1016/j.procs.2025.03.205

Immaculate Joy S , Sriram G , Sriram Venkatesan S

Innovations in the field of medical imaging techniques, especially in magnetic resonance imaging (MRI), have significantly enhanced diagnostic capabilities. However, the accurate classification of brain tumors from MRI scans remains a difficult task due to the subtle variations between different tumor types and the presence of non-tumorous regions. The primary challenges in automated MRI classification include the high variability in tumor appearance, similarities between benign and malignant tumor features, and the inherent imbalance in medical datasets. The proposed model architecture includes multiple convolutional layers with normalizing batches and removing outliers to enhance generalization and control for overfitting. The dataset was artificially expanded using data augmentation techniques like flipping, zooming, and rotating from 5,712 original images to 142,800 images, allowing the model to learn from a more diverse set of examples. The model demonstrated promising results, obtaining a training precision of 99% and a validation accuracy of 91.5% after 50 epochs, suggesting effective learning and generalization capabilities.

医学成像技术领域的创新，特别是磁共振成像（MRI），大大提高了诊断能力。然而，由于不同肿瘤类型和非肿瘤区域之间的微妙差异，从MRI扫描中准确分类脑肿瘤仍然是一项艰巨的任务。MRI自动分类的主要挑战包括肿瘤外观的高度可变性、良性和恶性肿瘤特征的相似性以及医疗数据集固有的不平衡。提出的模型架构包括多个卷积层，具有规范化批次和去除异常值，以增强泛化和控制过拟合。使用数据增强技术，如翻转、缩放和旋转，从5712张原始图像到142800张图像，人为地扩展了数据集，使模型能够从更多样化的示例集中学习。经过50次迭代，该模型的训练精度达到99%，验证精度达到91.5%，具有良好的学习和泛化能力。

{"title":"Deep CNN-Based Multi-Grade Brain Tumor Classification with Enhanced Data Augmentation","authors":"Immaculate Joy S , Sriram G , Sriram Venkatesan S","doi":"10.1016/j.procs.2025.03.205","DOIUrl":"10.1016/j.procs.2025.03.205","url":null,"abstract":"<div><div>Innovations in the field of medical imaging techniques, especially in magnetic resonance imaging (MRI), have significantly enhanced diagnostic capabilities. However, the accurate classification of brain tumors from MRI scans remains a difficult task due to the subtle variations between different tumor types and the presence of non-tumorous regions. The primary challenges in automated MRI classification include the high variability in tumor appearance, similarities between benign and malignant tumor features, and the inherent imbalance in medical datasets. The proposed model architecture includes multiple convolutional layers with normalizing batches and removing outliers to enhance generalization and control for overfitting. The dataset was artificially expanded using data augmentation techniques like flipping, zooming, and rotating from 5,712 original images to 142,800 images, allowing the model to learn from a more diverse set of examples. The model demonstrated promising results, obtaining a training precision of 99% and a validation accuracy of 91.5% after 50 epochs, suggesting effective learning and generalization capabilities.</div></div>","PeriodicalId":20465,"journal":{"name":"Procedia Computer Science","volume":"260 ","pages":"Pages 300-307"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144139369","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Collaborative Algorithm for User Trust and Data Security Based on Blockchain and Machine Learning 基于区块链和机器学习的用户信任与数据安全协同算法

Procedia Computer Science

Pub Date : 2025-01-01 DOI: 10.1016/j.procs.2025.05.108

Dishu Yang , Xingyu Liu

Machine learning has achieved remarkable results in numerous fields, demonstrating strong momentum and promising prospects for future development. However, machine learning is facing issues related to data security. User data contains a vast amount of sensitive personal information, and once privacy is breached, users may not only suffer from harassment but also face threats to their lives and property security. As a result, users’ willingness and trust in sharing local raw data are gradually decreasing. In response to this situation, federated learning technology has emerged, which enables efficient training of decentralized data through distributed machine learning methods while protecting users’ data privacy. Traditional federated learning systems suffer from issues such as single points of failure and lack of trust. Blockchain, as a decentralized, traceable, and tamper-resistant distributed ledger technology, provides a new solution for federated learning. It records every update of the global model, verifies and tracks local updates, and is equipped with a fair incentive mechanism. Based on these ideas, this paper proposes a federated learning framework combined with blockchain, aiming to address data security issues in federated learning.

机器学习在众多领域取得了显著成果，显示出强劲的发展势头和良好的发展前景。然而，机器学习面临着与数据安全相关的问题。用户数据中包含大量敏感的个人信息，一旦隐私被泄露，用户不仅会遭受骚扰，还会面临生命财产安全的威胁。因此，用户对本地原始数据共享的意愿和信任度逐渐下降。针对这种情况，联邦学习技术应运而生，在保护用户数据隐私的同时，通过分布式机器学习方法对分散的数据进行高效训练。传统的联邦学习系统存在单点故障和缺乏信任等问题。区块链作为一种去中心化、可追踪、防篡改的分布式账本技术，为联邦学习提供了一种新的解决方案。它记录全局模型的每一次更新，验证和跟踪局部更新，并配备公平的激励机制。基于这些思想，本文提出了一个结合区块链的联邦学习框架，旨在解决联邦学习中的数据安全问题。

{"title":"Collaborative Algorithm for User Trust and Data Security Based on Blockchain and Machine Learning","authors":"Dishu Yang , Xingyu Liu","doi":"10.1016/j.procs.2025.05.108","DOIUrl":"10.1016/j.procs.2025.05.108","url":null,"abstract":"<div><div>Machine learning has achieved remarkable results in numerous fields, demonstrating strong momentum and promising prospects for future development. However, machine learning is facing issues related to data security. User data contains a vast amount of sensitive personal information, and once privacy is breached, users may not only suffer from harassment but also face threats to their lives and property security. As a result, users’ willingness and trust in sharing local raw data are gradually decreasing. In response to this situation, federated learning technology has emerged, which enables efficient training of decentralized data through distributed machine learning methods while protecting users’ data privacy. Traditional federated learning systems suffer from issues such as single points of failure and lack of trust. Blockchain, as a decentralized, traceable, and tamper-resistant distributed ledger technology, provides a new solution for federated learning. It records every update of the global model, verifies and tracks local updates, and is equipped with a fair incentive mechanism. Based on these ideas, this paper proposes a federated learning framework combined with blockchain, aiming to address data security issues in federated learning.</div></div>","PeriodicalId":20465,"journal":{"name":"Procedia Computer Science","volume":"262 ","pages":"Pages 757-765"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144254030","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0