检索结果-内蒙古大学图书馆

Alexandria engineering Journal 2025年 110卷 376-376页

作者： Adnan Hasan Bdair AIghuraibawi Selvakumar Manickam Zaid Abdi Alkareem Alyasseri Rosni Abdullah Ayman Khallel Riyadh Rahef Nuiaa Al Ogaili Fahd N. Al-Wesabi Abdulsamad Ebrahim Yahya National Advanced IPv6 Centre (NAv6) Universiti Sains Malaysia Penang 11800 Malaysia Baghdad College of Economic Sciences University Baghdad Iraq Information Technology Research and Development Center University of Kufa Najaf Iraq College of Engineering University of Warith Al-Anbiyaa Karbala Iraq School of Computer Sciences Universiti Sains Malays Penang 11800 Malaysia Faculty of Computing and Informatics Universiti Malaysia Sabah Sabah Malaysia Department of Cybersecurity Engineering Technology Al Hikma University College Baghdad Iraq Department of Computer Science/ College of Computer Science and Information Technology Wasit University Kut Wasit 52001 Iraq Department of Computer Science College of Science & Art at Mahayil King Khalid University Saudi Arabia Department of Information Technology College of Computing and Information Technology Northern Border University Arar Saudi Arabia

CyberSentry: Enhancing SCADA Security through Advanced Deep Learning and Optimization Strategies

学校读者我要写书评

暂无评论

International Journal of Critical Infrastructure Protection 2025年

作者： Alaa O. Khadidos Adil O. Khadidos Shitharth Selvarajan Taher Al-Shehari Nasser A Alsadhan Subhav Singh Department of Information Systems Faculty of Computing and Information Technology King Abdulaziz University Jeddah Saudi Arabia Center of Research Excellence in Artificial Intelligence and Data Science King Abdulaziz University Jeddah Saudi Arabia Department of Information Technology Faculty of Computing and Information Technology King Abdulaziz University Jeddah Saudi Arabia School of Built Environment Engineering and Computing Leeds Beckett University LS1 3HE Leeds U.K Department of Computer Science and Engineering Chennai Institute of Technology Chennai India Centre for Research Impact & Outcome Chitkara University Institute of Engineering and Technology Chitkara University Punjab Rajpura 140401 India Computer Skills Department of Self-Development Skill Common First Year Deanship King Saud University 11362 Riyadh Saudi Arabia Computer Science Department College of Computer and Information Sciences King Saud University Riyadh 12372 Saudi Arabia Chitkara Centre for Research and Development Chitkara University Himachal Pradesh-174103 India Division of research and development Lovely Professional University Phagwara Punjab India

SCADA systems form the core of infrastructural facilities, including power grids, water treatment facilities, and industrial processes. Changing cyber threats present increasingly sophisticated attacks against which traditional security models inadequately protect SCADA systems. These traditional models usually have drawbacks in the way of inadequate feature selection, inefficiency in detecting most attacks, and suboptimal parameter tuning, which cause vulnerabilities and reduce resilience in systems. This paper presents CyberSentry, a new security framework designed to overcome limitations so as to provide robust protection for SCADA systems. These three modules makeup CyberSentry: the RMIG feature selection model, tri-fusion net for attack detection, and Parrot-Levy Blend Optimization (PLBO) for parameter tuning. The Recursive Multi-Correlation-based Information Gain (RMIG) feature selection model enhances accuracy in detection by optimizing the set of fatal features through recursive multi-correlation analysis by Information Gain prioritization. The Tri-Fusion Net combines anomaly detection, signature-based detection, and machine learning classifiers to enhance the detection versatility and robustness. The PLBO module ensures efficient and dynamic tuning for the parameters through undocumented Parrot and Levy optimization techniques. The proposed CyberSentry framework integrates, within a unified architecture, anomaly detection, signature-based detection, and machine learning classifiers to enhance the security of SCADA systems against diverse cyber threats. Features extracted in this manner are analyzed using machine learning classifiers that exploit their predictive capabilities for robust threat classification. The proposed approaches are fused within the Tri-Fusion Net to complement each other in areas where the separate methods lack certain strengths. This, therefore, ensures broad threat detection, as is validated by extensive testing with various datasets

关键词： SCADA Systems Security Cyber-Attack Intrusion Detection Deep Learning Feature Selection and Optimization

Artificial Intelligence-Based Suicide Prevention and Prediction: A Systematic Review (2019-2023)

学校读者我要写书评

暂无评论

SSRN

SSRN 2024年

作者： Atmakuru, Anirudh Shahini, Alen Chakraborty, Subrata Seoni, Silivia Salvi, Massimo Hafeez-Baig, Abdul Rashid, Sadaf Tan, Ru San Barua, Prabal Datta Molinari, Filippo Acharya, U. Rajendra Manning College of Information and Computer Sciences University of Massachusetts AmherstMA United States Biolab PolitoBIOMedLab Department of Electronics and Telecommunications Politecnico di Torino Corso Duca degli Abruzzi 24 Turin10129 Italy Faculty of Science Agriculture Business and Law University of New England ArmidaleNSW2531 Australia Faculty of Engineering and IT University of Technology Sydney SydneyNSW2007 Australia Biolab Department of Electronics and Telecommunications Politecnico di Torino Corso Duca degli Abruzzi 24 Turin10129 Italy Department of Electronics and Telecommunications Politecnico di Torino Corso Duca degli Abruzzi 24 Turin10129 Italy School of Management and Enterprise University of Southern Queensland ToowoombaQLD Australia Sunshine Coast Hospital and Health Service Queensland Australia Department of Cardiology National Heart Centre Singapore Singapore Duke-NUS Medical School Singapore University of Southern Queensland Australia School of Mathematics Physics and Computing University of Southern Queensland Springfield Australia

Suicide is a major global public health concern, and the application of artificial intelligence (AI) methods, such as natural language processing, machine learning, and deep learning, has shown promise in advancing suicide prediction and prevention efforts. In this work, we conduct a systematic literature review to identify relevant studies published between 2019 and 2023, resulting in the inclusion of 156 studies. We provide a comprehensive overview of the current state of research conducted on AI-driven suicide prevention and prediction, focusing on different data types and AI techniques employed. We discuss the benefits and challenges of these approaches, and propose future research directions to improve the practical application of AI in suicide research. AI is highly capable of improving the accuracy and efficiency of risk assessment, enabling personalized interventions, and enhancing our understanding of risk and protective factors. Multidisciplinary approaches combining diverse data sources and AI methods can help identify individuals at risk by analyzing social media content, patient histories, and data from mobile devices, enabling timely intervention. However, challenges related to data privacy, algorithmic bias, model interpretability, and real-world implementation must be addressed to realize the full potential of these technologies. Future research should focus on integrating prediction and prevention strategies, harnessing multimodal data, and expanding the scope to include diverse populations. Collaboration across disciplines and stakeholders is essential to ensure that AI-driven suicide prevention and prediction efforts are ethical, culturally sensitive, and person-centered. © 2024, The Authors. All rights reserved.

关键词： Forecasting

Improving unsupervised domain adaptation by reducing bi-level feature redundancy

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Wang, Mengzhu Zhang, Xiang Lan, Long Wang, Wei Tan, Huibin Luo, Zhigang Science and Technology on Parallel and Distributed Laboratory College of Computer National University of Defense Technology Changsha China Institute for Quantum State Key Laboratory of High Performance Computing National University of Defense Technology Changsha China DUT-RU International School of Information Science & Engineering Dalian University of Technology DalianLiaoning116000 China Department of Science and Technology on Parallel and Distributed Processing National University of Defense Technology Changsha China

Reducing feature redundancy has shown beneficial effects for improving the accuracy of deep learning models, thus it is also indispensable for the models of unsupervised domain adaptation (UDA). Nevertheless, most recent efforts in the field of UDA ignore this point. Moreover, main schemes realizing this in general independent of UDA purely involve a single domain, thus might not be effective for cross-domain tasks. In this paper, we emphasize the significance of reducing feature redundancy for improving UDA in a bi-level way. For the first level, we try to ensure compact domain-specific features with a transferable decorrelated normalization module, which preserves specific domain information whilst easing the side effect of feature redundancy on the sequel domain-invariance. In the second level, domain-invariant feature redundancy caused by domain-shared representation is further mitigated via an alternative brand orthogonality for better generalization. These two novel aspects can be easily plugged into any BN-based backbone neural networks. Specifically, simply applying them to ResNet50 has achieved competitive performance to the state-of-the-arts on five popular benchmarks. Our code will be available at https://***/dreamkily/gUDA. © 2020, CC BY.

关键词： Deep learning

Interpolation-Split: a data-centric deep learning approach with big interpolated data to boost airway segmentation performance

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Cheung, Wing Keung Pakzad, Ashkan Mogulkoc, Nesrin Needleman, Sarah Rangelov, Bojidar Gudmundsson, Eyjolfur Zhao, An Abbas, Mariam McLaverty, Davina Asimakopoulos, Dimitrios Chapman, Robert Savas, Recep Janes, Sam M. Hu, Yipeng Alexander, Daniel C. Hurst, John R. Jacob, Joseph Satsuma Lab Centre for Medical Image Computing University College London London United Kingdom Department of Computer Science University College London London United Kingdom Department of Medical Physics and Biomedical Engineering University College London London United Kingdom Department of Respiratory Medicine Ege University Hospital Izmir Turkey Medical School University College London London United Kingdom School of Clinical Medicine University of Cambridge Cambridge United Kingdom Interstitial Lung Disease Service Department of Respiratory Medicine University College London Hospitals NHS Foundation Trust London United Kingdom Department of Radiology Ege University Hospital Izmir Turkey Lungs for Living Research Centre UCL London United Kingdom UCL Respiratory University College London London United Kingdom Respiratory Medicine Royal Free London NHS Foundation Trust London United Kingdom

The morphology and distribution of airway tree abnormalities enables diagnosis and disease characterisation across a variety of chronic respiratory conditions. In this regard, airway segmentation plays a critical role in the production of the outline of the entire airway tree to enable estimation of disease extent and severity. In this study, we propose a data-centric deep learning technique with big interpolated data, Interpolation-Split, to boost segmentation performance of the airway tree. The proposed technique utilises interpolation and image split to improve data usefulness and quality. Then, an ensemble learning strategy is implemented to aggregate the segmented airway segments at different scales. In terms of average segmentation performance (dice similarity coefficient, DSC), our method (A) achieves 90.55%, 89.52% and 85.80%, (B) outperforms the baseline models by 2.89%, 3.86% and 3.87% on average and (C) produces maximum segmentation performance gain by 14.11%, 9.28% and 12.70% for individual case when (1) nnU-Net with instant normalisation and leaky ReLU, (2) nnU-Net with batch normalisation and ReLU and (3) modified dilated U-Net are used respectively. Furthermore, our proposed technique has a low RAM/GPU memory usage;and it is GPU memory efficient and highly flexible enabling it to be deployed on any 2D deep learning model. © 2023, CC BY-NC-ND.

关键词： Deep learning

On the Philosophical, Cognitive and Mathematical Foundations of Symbiotic Autonomous Systems (SAS)

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Wang, Yingxu Karray, Fakhri Kwong, Sam Plataniotis, Konstantinos N. Leung, Henry Hou, Ming Tunstel, Edward Rudas, Imre J. Trajkovic, Ljiljana Kaynak, Okyay Kacprzyk, Janusz Zhou, Mengchu Smith, Michael H. Chen, Philip Patel, Shushma University of Calgary Canada FIEEE Dept. of Electrical & Computer Engineering University of Waterloo ON Canada FIEEE Department of Computer Science City University of Hong Kong Hong Kong FIEEE Dept. of Electrical & Computer Engineering University of Toronto ON Canada SMIEEE Toronto Research Centre DRDC Canada FIEEE Autonomous & Intelligent Systems Dept. Raytheon Technologies Research Center United States Óbuda University Budapest Hungary FIEEE School of Engineering Science Simon Fraser University BurnabyBC Canada FIEEE Bogazici University Bebek Istanbul Turkey FIEEE Systems Research Institute Polish Academy of Sciences Warsaw Poland FIEEE Dept. of Computer Science New Jersey Institute of Technology NJ United States SMIEEE Furaxa Inc. OrindaCA United States FIEEE School of Computer Science and Engineering South China University of Technology Guangzhou China FBCS Faculty of Computing Engineering and Media De Montfort University LeicesterLE1 9BH United Kingdom

Symbiotic Autonomous Systems (SAS) are advanced intelligent and cognitive systems exhibiting autonomous collective intelligence enabled by coherent symbiosis of human-machine interactions in hybrid societies. Basic research in the emerging field of SAS has triggered advanced general AI technologies functioning without human intervention or hybrid symbiotic systems synergizing humans and intelligent machines into coherent cognitive systems. This work presents a theoretical framework of SAS underpinned by the latest advances in intelligence, cognition, computer, and system sciences. SAS are characterized by the composition of autonomous and symbiotic systems that adopt bio-brain-social-inspired and heterogeneously synergized structures and autonomous behaviors. This paper explores their cognitive and mathematical foundations. The challenge to seamless human-machine interactions in a hybrid environment is addressed. SAS-based collective intelligence is explored in order to augment human capability by autonomous machine intelligence towards the next generation of general AI, autonomous computers, and trustworthy mission-critical intelligent systems. Emerging paradigms and engineering applications of SAS are elaborated via an autonomous knowledge learning system that symbiotically works between humans and cognitive robots. © 2021, CC BY-NC-ND.

关键词： Cognitive systems

DeePMD-kit v3: A Multiple-Backend Framework for Machine Learning Potentials

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Zeng, Jinzhe Zhang, Duo Peng, Anyang Zhang, Xiangyu He, Sensen Wang, Yan Liu, Xinzijian Bi, Hangrui Li, Yifan Cai, Chun Zhang, Chengqian Du, Yiming Zhu, Jia-Xin Mo, Pinghui Huang, Zhengtao Zeng, Qiyu Shi, Shaochen Qin, Xuejian Yu, Zhaoxi Luo, Chenxing Ding, Ye Liu, Yun-Pei Shi, Ruosong Wang, Zhenyu Bore, Sigbjørn Løland Chang, Junhan Deng, Zhe Ding, Zhaohan Han, Siyuan Jiang, Wanrun Ke, Guolin Liu, Zhaoqing Lu, Denghui Muraoka, Koki Oliaei, Hananeh Singh, Anurag Kumar Que, Haohui Xu, Weihong Xu, Zhangmancang Zhuang, Yong-Bin Dai, Jiayu Giese, Timothy J. Jia, Weile Xu, Ben York, Darrin M. Zhang, Linfeng Wang, Han School of Artificial Intelligence and Data Science Unversity of Science and Technology of China Hefei China AI for Science Institute Beijing100080 China DP Technology Beijing100080 China Academy for Advanced Interdisciplinary Studies Peking University Beijing100871 China State Key Lab of Processors Institute of Computing Technology Chinese Academy of Sciences Beijing100871 China University of Chinese Academy of Sciences Beijing China Baidu Inc. Beijing China Department of Computer Science University of Toronto TorontoON Canada Department of Chemistry Princeton University PrincetonNJ08540 United States University of Chinese Academy of Sciences Beijing100871 China State Key Laboratory of Physical Chemistry of Solid Surfaces iChEM College of Chemistry and Chemical Engineering Xiamen University Xiamen361005 China College of Integrated Circuits Hunan University Changsha410082 China State Key Laboratory of Advanced Technology for Materials Synthesis and Processing Center for Smart Materials and Device Integration School of Material Science and Engineering Wuhan University of Technology Wuhan430070 China College of Science National University of Defense Technology Changsha410073 China Hunan Key Laboratory of Extreme Matter and Applications National University of Defense Technology Changsha410073 China ByteDance Research Beijing100098 China Ningbo Institute of Materials Technology and Engineering Chinese Academy of Sciences Ningbo315201 China College of Materials Science and Opto-Electronic Technology University of Chinese Academy of Sciences Beijing100049 China Key Laboratory of Theoretical and Computational Photochemistry of Ministry of Education College of Chemistry Beijing Normal University Beijing100875 China Department of Geosciences Princeton University PrincetonNJ08544 United States Department of Applied Physics and Applied Mathematics Columbia University New YorkNY10027 United States IKKEM Fujian Xiamen361005 China Graduate

In recent years, machine learning potentials (MLPs) have become indispensable tools in physics, chemistry, and materials science, driving the development of software packages for molecular dynamics (MD) simulations and related applications. These packages, typically built on specific machine learning frameworks such as TensorFlow, PyTorch, or JAX, face integration challenges when advanced applications demand communication across different frameworks. The previous TensorFlow-based implementation of DeePMD-kit exemplified these limitations. In this work, we introduce DeePMD-kit version 3, a significant update featuring a multi-backend framework that supports TensorFlow, PyTorch, JAX, and PaddlePaddle backends, and demonstrate the versatility of this architecture through the integration of other MLPs packages and of Differentiable Molecular Force Field. This architecture allows seamless backend switching with minimal modifications, enabling users and developers to integrate DeePMD-kit with other packages using different machine learning frameworks. This innovation facilitates the development of more complex and interoperable workflows, paving the way for broader applications of MLPs in scientific research. Copyright © 2025, The Authors. All rights reserved.

关键词： Molecular dynamics

The Multi-modality Cell Segmentation Challenge: Towards Universal Solutions

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Ma, Jun Xie, Ronald Ayyadhury, Shamini Ge, Cheng Gupta, Anubha Gupta, Ritu Gu, Song Zhang, Yao Lee, Gihun Kim, Joonkee Lou, Wei Li, Haofeng Upschulte, Eric Dickscheid, Timo de Almeida, José Guilherme Wang, Yixin Han, Lin Yang, Xin Labagnara, Marco Gligorovski, Vojislav Scheder, Maxime Rahi, Sahand Jamal Kempster, Carly Pollitt, Alice Espinosa, Leon Mignot, Tâm Middeke, Jan Moritz Eckardt, Jan-Niklas Li, Wangkai Li, Zhaoyang Cai, Xiaochen Bai, Bizhe Greenwald, Noah F. Van Valen, David Weisbart, Erin Cimini, Beth A. Cheung, Trevor Brück, Oscar Bader, Gary D. Wang, Bo Peter Munk Cardiac Centre University Health Network TorontoON Canada Department of Laboratory Medicine and Pathobiology University of Toronto TorontoON Canada Vector Institute TorontoON Canada Department of Molecular Genetics University of Toronto TorontoON Canada Donnelly Centre University of Toronto TorontoON Canada Princess Margaret Cancer Centre University Health Network TorontoON Canada School of Medicine and Pharmacy Ocean University of China Qingdao China New Delhi India Laboratory Oncology Dr. BRA-IRCH All India Institute of Medical Sciences New Delhi India Department of Image Reconstruction Nanjing Anke Medical Technology Co. Ltd. Nanjing China Shanghai Artificial Intelligence Laboratory Shanghai China Graduate School of AI KAIST Seoul Korea Republic of Shenzhen Research Institute of Big Data Shenzhen China Shenzhen China Helmholtz AI Research Center Jülich Jülich Germany Faculty of Mathematics and Natural Sciences Institute of Computer Science Heinrich Heine University Düsseldorf Düsseldorf Germany Hinxton United Kingdom Champalimaud Foundation - Centre for the Unknown Lisbon Portugal Department of Bioengineering Stanford University Palo AltoCA United States Tandon School of Engineering New York University New YorkNY United States School of Biomedical Engineering Health Science Center Shenzhen University Shenzhen China Lausanne Switzerland School of Biological Sciences University of Reading Reading United Kingdom Laboratoire de Chimie Bactérienne CNRS Université Aix Marseille UMR Institut de Microbiologie de la Méditerranée Marseille France Department of Internal Medicine I University Hospital Dresden Technical University Dresden Dresden Germany Else Kroener Fresenius Center for Digital Health Technical University Dresden Dresden Germany Department of Automation University of Science and Technology of China Hefei China Institute of Advanced Technology University of Science and Technology of China Hefei Chi

Cell segmentation is a critical step for quantitative single-cell analysis in microscopy images. Existing cell segmentation methods are often tailored to specific modalities or require manual interventions to specify hyper-parameters in different experimental settings. Here, we present a multi-modality cell segmentation benchmark, comprising over 1500 labeled images derived from more than 50 diverse biological experiments. The top participants developed a Transformer-based deep-learning algorithm that not only exceeds existing methods but can also be applied to diverse microscopy images across imaging platforms and tissue types without manual parameter adjustments. This benchmark and the improved algorithm offer promising avenues for more accurate and versatile cell analysis in microscopy imaging. © 2023, CC BY-NC-ND.

关键词： Deep learning

Active interactions between animals and technology: biohybrid approaches for animal behaviour research

学校读者我要写书评

暂无评论

Animal Behaviour 2025年 224卷

作者： Papadopoulou, M. Ball, M. Bartashevich, P. Burns, A.L.J. Chiara, V. Clark, M.A. Costelloe, B.R. Fele, M. French, F. Hauert, S. Heinrich, M.K. Herbert-Read, J.E. Hoitt, J. Ioannou, C.C. Landgraf, T. Matchette, S.R. Polverino, G. Sankey, D.W.E. Scott, D.M. Sridhar, V.H. Strömbom, D. Trianni, V. Vo-Doan, T.T. King, A.J. Department of Biosciences Faculty of Science and Engineering Swansea University Swansea United Kingdom Department of Biology Lafayette College Easton PA United States Institute for Theoretical Biology Department of Biology Humboldt-Universität zu Berlin Berlin Germany Cluster of Excellence ‘Science of Intelligence’ Berlin Germany Department Fish Biology Fisheries and Aquaculture Leibniz Institute of Freshwater Ecology and Inland Fisheries Berlin Germany Faculty of Life Sciences Albrecht Daniel Thaer-Institute of Agricultural and Horticultural Sciences Humboldt-Universität zu Berlin Berlin Germany Museum and Institute of Zoology Polish Academy of Science Warszawa Poland School of Biological Sciences University of Bristol Bristol United Kingdom School of Natural Sciences Macquarie University Sydney Australia Centre for the Advanced Study of Collective Behaviour University of Konstanz Konstanz Germany Department of Biology University of Konstanz Konstanz Germany Department of Collective Behavior Max Planck Institute of Animal Behavior Konstanz Germany School of Computing and Digital Media London Metropolitan University London United Kingdom Bristol Robotics Laboratory University of Bristol Bristol United Kingdom IRIDIA Université Libre de Bruxelles Brussels Belgium Department of Zoology University of Cambridge Cambridge United Kingdom Department of Mathematics and Computer Science Freie Universität Berlin Berlin Germany Department of Ecological and Biological Sciences University of Tuscia Viterbo Italy School of Natural and Environmental Science Newcastle University Newcastle upon Tyne United Kingdom Centre for Ecology and Conservation Faculty of Environment Science and Economy University of Exeter Penryn Campus Cornwall United Kingdom School of Animal Rural and Environmental Sciences Nottingham Trent University Nottingham United Kingdom Department for the Ecology of Animal Societies Max Planck Institute of Animal Behavior Konstanz Germany Ins

Biohybrid approaches (where living and engineered components are combined) provide new opportunities for advancing animal behaviour research and its applications. This review article and accompanying special issue explores how different types of novel technologies can be used in the field of animal behaviour from three perspectives: (1) comprehension, (2) application and (3) integration. Under the perspective of ‘comprehension,’ we present examples of how technologies like virtual animals or robots can be used in experimental settings to interact with living animals in a standardized manner. Such interactions can advance our understanding of fundamental topics such as mate choice, social learning and collective behaviour. Under ‘application,’ we investigate the potential for technologies to monitor, react and interact with animals in a variety of scenarios. For example, we discuss how drones can be used to keep large herbivores away from valuable crops and robotic predators to deter invasive species. Under ‘integration,’ we discuss possibilities for the coexistence of engineered and biological systems, augmenting the capacity or resilience of either or both components. Integration can be physical, for example, livestock can have sensors sit in their inner body for temperature monitoring, or within the environment, where sensors or robots monitor and interact with animals, such as a short-term earthquake forecasting method. Based upon these three themes, we discuss and classify existing biohybrid animal behaviour research, including the four articles included in our special issue. We also consider the ethics of this emerging field, highlight the advantages and potential issues associated with using technologies to create biohybrid systems and emphasize how such technologies can support the advancement of animal behaviour research. © 2025 The Association for the Study of Animal Behaviour

关键词： behavioural control behavioural monitoring biomimetic robotics ethorobotics interspecies interaction sensory integration

Active learning using adaptable task-based prioritisation

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Saeed, Shaheer U. Ramalhinho, João Pinnock, Mark Shen, Ziyi Fu, Yunguan Montaña-Brown, Nina Bonmati, Ester Barratt, Dean C. Pereira, Stephen P. Davidson, Brian Clarkson, Matthew J. Hu, Yipeng The Centre for Medical Image Computing Wellcome EPSRC Centre for Interventional and Surgical Sciences Department of Medical Physics and Biomedical Engineering University College London LondonWC1E 6BT United Kingdom InstaDeep LondonW2 1AY United Kingdom The School of Computer Science and Engineering University of Westminster LondonW1W 6UW United Kingdom The Institute for Liver and Digestive Health University College London LondonNW3 2QG United Kingdom The Division of Surgery and Interventional Sciences University College London LondonWC1E 6BT United Kingdom

Supervised machine learning-based medical image computing applications necessitate expert label cu-ration, while unlabelled image data might be relatively abundant. Active learning methods aim to prioritise a subset of available image data for expert annotation, for label-efficient model training. We develop a controller neural network that measures priority of images in a sequence of batches, as in batch-mode active learning, for multiclass segmentation tasks. The controller is optimised by rewarding positive task-specific performance gain, within a Markov decision process (MDP) environment that also optimises the task predictor. In this work, the task predictor is a segmentation network. A meta-reinforcement learning algorithm is proposed with multiple MDPs, such that the pre-trained controller can be adapted to a new MDP that contains data from different institutes and/or requires segmentation of different organs or structures within the abdomen. We present experimental results using multiple CT datasets from more than one thousand patients, with segmentation tasks of nine different abdominal organs, to demonstrate the efficacy of the learnt prioritisation controller function and its cross-institute and cross-organ adaptability. We show that the proposed adaptable prioritisation metric yields converging segmentation accuracy for the novel class of kidney, unseen in training, using between approximately 40% to 60% of labels otherwise required with other heuristic or random prioritisation metrics. For clinical datasets of limited size, the proposed adaptable prioritisation offers a performance improvement of 22.6% and 10.2% in Dice score, for tasks of kidney and liver vessel segmentation, respectively, compared to random prioritisation and alternative active sampling strategies. © 2022, CC BY.

关键词： Reinforcement learning