Artificial Intelligence Review最新文献

{"title":"Generative anomaly detection: a comprehensive review of modeling principles, advances, and future opportunities","authors":"Jiaqi Zhu,&nbsp;Yunfeng Fan,&nbsp;Geng Han,&nbsp;Xiang Shi,&nbsp;Fang Deng,&nbsp;Jie Chen","doi":"10.1007/s10462-026-11591-w","DOIUrl":"10.1007/s10462-026-11591-w","url":null,"abstract":"<div><p>The unprecedented evolution of generative artificial intelligence (GAI) has reshaped the landscape of anomaly detection, enabling models to capture complex data distributions and expose subtle deviations that are previously elusive to traditional methods. However, the rapid proliferation of generative techniques, ranging from autoencoders and generative adversarial networks to diffusion models and large pre-trained foundation models, has resulted in a fragmented body of knowledge lacking a unified understanding. This survey addresses this critical gap by presenting the first comprehensive synthesis of GAI-empowered anomaly detection. We propose a holistic taxonomy that unifies classical generative paradigms with the transformative role of foundation models, positioning them within the broader context of detection and representation learning. Beyond cataloging methods, we examine how complementary generative tasks such as anomaly synthesis and restoration enhance anomaly detection capabilities, and analyze their practical deployment across key application domains. We underscore the reciprocal relationship between advances in GAI and the evolving challenges of anomaly detection, further delineating unresolved issues and outlining promising research directions to stimulate future progress in the field. By consolidating disparate developments into a coherent framework and offering forward-looking insights, this survey seeks to catalyze deeper integration between the GAI and anomaly detection communities, advancing both theoretical foundations and practical impact in this rapidly evolving field. An online project for this survey is available at: https://github.com/zjiaqi725/Awesome-Generative-Anomaly-Detection.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 9","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11591-w.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148750627","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Retrieval-augmented generation for natural language processing: a survey","authors":"Shangyu Wu,&nbsp;Ying Xiong,&nbsp;Yufei Cui,&nbsp;Haolun Wu,&nbsp;Can Chen,&nbsp;Ye Yuan,&nbsp;Lianming Huang,&nbsp;Xue Liu,&nbsp;Tei-Wei Kuo,&nbsp;Nan Guan,&nbsp;Chun Jason Xue","doi":"10.1007/s10462-026-11605-7","DOIUrl":"10.1007/s10462-026-11605-7","url":null,"abstract":"<div><p>Large language models (LLMs) have achieved strong empirical performance in various fields, benefiting from their huge amount of parameters that store knowledge. However, LLMs still suffer from several key issues, such as hallucination problems, knowledge update issues, and lacking domain-specific expertise. The appearance of retrieval-augmented generation (RAG), which leverages an external knowledge base to augment LLMs, mitigates these limitations. This paper presents a systematic review of RAG techniques for natural language processing (NLP), with a focus on retrievers and retrieval fusions. We introduce a novel taxonomy of retrieval fusions, such as query-based, logits-based, latent, and parametric fusion, and provide structured comparisons across accessibility, efficiency, and use cases. The paper further examines RAG applications across diverse NLP tasks, discusses evaluation methodologies and benchmark limitations, and analyzes training paradigms with and without knowledge base updates. Finally, we explore industrial deployment considerations and identify emerging challenges and future directions, including security, efficiency, and graph-based retrieval.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 9","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11605-7.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148750643","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"AI-driven household electricity load forecasting: challenges, methods, and future directions","authors":"Elmar Ibrahimov,&nbsp;Xuefei Yin,&nbsp;Lee Weng,&nbsp;Yong Zhu,&nbsp;Yanming Zhu,&nbsp;Alan Wee-Chung Liew","doi":"10.1007/s10462-026-11583-w","DOIUrl":"10.1007/s10462-026-11583-w","url":null,"abstract":"<div><p>Accurate household electricity load forecasting is becoming increasingly vital with the continued growth of smart grids, household renewable energy systems, and smart meter deployment. Unlike regional or grid load forecasting, household-level forecasting presents unique challenges due to highly irregular consumption patterns, data scarcity, privacy concerns, and behavioral variability. In recent years, artificial intelligence (AI) methods have demonstrated strong potential to address these complexities, enabling more accurate, robust, and adaptive forecasting systems. This survey presents a comprehensive and up-to-date review, with a focus on AI-based techniques specifically tailored to household-level forecasting. A key contribution of this work is the development of a challenge-centered taxonomy that categorizes methods based on four critical problem domains: methodological limitations, data-related constraints, behavioral complexity, and privacy and security concerns. By aligning representative AI approaches with these core challenges, the survey offers a structured and insightful understanding of the current research landscape. It also provides a comparative analysis with prior surveys, identifies gaps in the literature, and highlights promising research directions, including multimodal learning, adaptive modeling, integration of large language models, and privacy-preserving forecasting. This work could serve as a valuable resource for researchers and practitioners aiming to advance intelligent and trustworthy forecasting solutions in household energy systems.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 9","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-05-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11583-w.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148614845","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Correction: Imbalanced data oversampling through subspace optimization with Bayesian reinforcement","authors":"Mahesh Kumbhar,&nbsp;Sunith Bandaru,&nbsp;Alexander Karlsson","doi":"10.1007/s10462-026-11561-2","DOIUrl":"10.1007/s10462-026-11561-2","url":null,"abstract":"","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 7","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-05-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11561-2.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148011833","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An auditable pipeline for fuzzy full-text screening in systematic reviews: integrating contrastive semantic highlighting and LLM judgment","authors":"Arya Rahgozar,&nbsp;Pouria Mortezaagha","doi":"10.1007/s10462-026-11599-2","DOIUrl":"10.1007/s10462-026-11599-2","url":null,"abstract":"<div><p>Full-text screening is the major bottleneck of systematic reviews (SRs), particularly in domains such as population health modelling of noncommunicable diseases (NCDs), where decisive eligibility information is scattered across long and heterogeneous full texts. In this methodological study, we introduce a scalable and auditable pipeline that reframes inclusion and exclusion as a fuzzy decision process. We evaluate the approach within the Population Health Modelling Consensus Reporting Network (POPCORN) and benchmark it against statistical and crisp baselines. Articles are parsed into overlapping chunks and embedded with a domain-adapted model; for each criterion (Population, Intervention, Outcome, Study Approach), we compute contrastive similarity (inclusion–exclusion cosine) and a <i>vagueness margin</i>, which a Mamdani fuzzy controller maps into graded inclusion degrees with dynamic thresholds in a multi-label classification setting. A large language model (LLM) judge adjudicates highlighted spans with tertiary labels, confidence scores, and criterion-referenced rationales; when evidence is insufficient, fuzzy membership is attenuated rather than hard-excluded. In a pilot on an all-positive gold set (<span>(N{=}16)</span> full texts; <span>(M{=}3{,}208)</span> chunks), the fuzzy system achieved document-level recall of 81.25% (Population), 87.50% (Intervention), 87.50% (Outcome), and 75.00% (Study Approach) with 95% Wilson confidence intervals, exceeding statistical baselines (recall 56.25–75.00%) and crisp baselines (recall 43.75–81.25%). Strict “All Criteria” inclusion was reached for 50.00% of articles, compared to 25.00% and 12.50% under the baselines. Cross-model agreement on justifications was 98.27%, and human–machine agreement was 96.07%. A pilot review showed 91% inter-rater agreement (<span>(kappa {=}0.82)</span>) with screening time reduced from <span>(sim)</span>20 minutes to under 1 minute per article at a significantly lower cost. These results suggest that fuzzy logic, combined with contrastive highlighting and explainable LLM adjudication, delivers high recall, stable rationales, and end-to-end traceability. We emphasize that the reported recall figures were computed on an all-positive gold set (no excluded documents), so specificity, ROC/PR curves, and AUC are not defined here; the complete operating-characteristic analysis on a mixed-label corpus will be reported in a forthcoming study.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 10","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13529844/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148878908","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A comprehensive comparison of constrained multi-objective evolutionary algorithms","authors":"Mengjun Ming,&nbsp;Rui Wang,&nbsp;Xin Zhou,&nbsp;Tao Zhang,&nbsp;Hisao Ishibuchi,&nbsp;Carlos A. Coello Coello","doi":"10.1007/s10462-026-11576-9","DOIUrl":"10.1007/s10462-026-11576-9","url":null,"abstract":"<div><p>In order to solve constrained multi-objective optimization problems (CMOPs), many algorithms have been proposed in the evolutionary computation research community in the past two decades. In general, the effectiveness of an algorithm on CMOPs is evaluated by artificial test problems. At present, comprehensive performance comparisons of constraint-handling techniques on various test problems remain scarce. In this study, we selected 12 representative algorithms that integrate with different constraint-handling techniques and compared their performance on existing test suites. Performance comparisons identify the strengths and weaknesses of different constraint-handling techniques on different types of CMOPs, and provide guidance on how to select/design constraint-handling techniques in specific scenarios. In addition, after a careful review of current artificial test problems, we found out that their design is inadequate to reflect some characteristics of real-world applications (e.g., the change in dimensionality of the Pareto front due to the inclusion of constraints, a large number of constraints, and diverse geometries of unconstrained Pareto fronts). Therefore, we point out the need to further develop test problems with these real-world characteristics. Furthermore, we present a fully reproducible experimental platform with multi-metric evaluation and statistical validation to enhance the scientific rigor of the study.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 9","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-05-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11576-9.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148614356","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Empowering healthcare 5.0 with deep learning: techniques, trends, and future directions","authors":"Paramita Kundu Maji,&nbsp;Sanjay Chakraborty,&nbsp;Afifa Sadiq,&nbsp;Saikat Basu,&nbsp;Krishnendu Ghosh","doi":"10.1007/s10462-026-11593-8","DOIUrl":"10.1007/s10462-026-11593-8","url":null,"abstract":"<div><p>The rapid advancement of intelligent technologies has driven a transformative shift in healthcare, giving rise to Healthcare 5.0, a new paradigm centered on patient-focused, digitally empowered medical services. This study presents a comprehensive review of cutting-edge Deep Learning (DL) techniques that are redefining Healthcare 5.0 through applications in disease prediction and early diagnosis, medical image analysis and radiology, and multimodal deep learning (MMDL). We analyze a broad spectrum of DL architectures, including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTMs), Gated Recurrent Units (GRUs), Generative Adversarial Network (GANs), transformers, autoencoders, transfer learning, and MMDL, highlighting the unique features that make them particularly suited for healthcare tasks. Using bibliometric analysis of 1,342 Scopus-indexed publications (2018-2025) via VOSviewer, we uncover key research trends, thematic clusters, and emerging focus areas. Furthermore, we assess DL models across diverse real-world healthcare datasets and discuss critical challenges, including data privacy, model interpretability, and system integration. Following the PRISMA methodology, this review also synthesizes recent research on advanced DL architectures for disease diagnosis and Healthcare 5.0, alongside widely used medical imaging and multimodal datasets. The manuscript has been substantially strengthened by incorporating a comparative analysis, quantitative aggregation of reported results, an explicit novelty and contribution statement, and a systematic comparison of DL survey studies across Healthcare 4.0 and Healthcare 5.0, thereby providing a more critical, data-driven, and clearly differentiated perspective beyond prior reviews. Finally, we address significant issues, including ethical considerations, privacy concerns, and model limitations, while outlining promising directions for future research. This work serves as a valuable resource for researchers and practitioners, offering both a broad overview and deep insights into the latest developments in DL for Healthcare 5.0 systems.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 9","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-05-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11593-8.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148614358","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Ethical explainable artificial intelligence for green innovation: a systematic review and meta-analysis of transparency, fairness, and accountability in sustainable technologies","authors":"Kashif Mazhar,&nbsp;Syed Shahid Mazhar,&nbsp;Farhina Sardar Khan,&nbsp;Dhoha AlSaleh,&nbsp;Amir Mazhar,&nbsp;Sharmeen Imam,&nbsp;Aseel Smerat","doi":"10.1007/s10462-026-11566-x","DOIUrl":"10.1007/s10462-026-11566-x","url":null,"abstract":"<div><p>Artificial Intelligence (AI) is increasingly deployed across sustainability-oriented domains such as smart energy systems, precision agriculture, water management, and sustainable manufacturing to support the United Nations Sustainable Development Goals (SDGs). However, the growing use of complex and opaque models has raised concerns regarding transparency, accountability, fairness, and ethical governance in high-impact sociotechnical decision-making contexts. This study presents a PRISMA-guided systematic literature review and meta-analysis of 140 peer-reviewed empirical studies (2020–2025) that integrate Explainable Artificial Intelligence (XAI) within sustainability-focused AI applications. Following the structured identification, screening, and eligibility assessment of 550 records retrieved from Scopus, Web of Science, IEEE Xplore, ScienceDirect, and SpringerLink, quantitative synthesis was conducted using random-effects models to explicitly accommodate cross-study heterogeneity. The meta-analysis indicates that XAI-integrated AI systems reported consistently high descriptive performance levels within the analysed literature (pooled classification accuracy ≈ 95.84%, aggregated regression fit R² ≈ 0.964). These pooled estimates do not represent transferable or generalisable performance benchmarks, as substantial heterogeneity exists across studies in terms of algorithmic paradigms, dataset characteristics, feature-engineering strategies, and experimental protocols (I² ≈ 60–70%). Rather, the aggregated values summarise literature-level reporting tendencies under diverse experimental conditions. Importantly, the correlation analysis revealed no systematic evidence of performance degradation associated with the incorporation of explainability mechanisms (Spearman’s <i>ρ</i> =  − 0.08, <i>p</i> &gt; 0.05). This finding should be interpreted as a context-dependent tendency observed within the reviewed sustainability literature, rather than as definitive proof that no interpretability–performance trade-off exists under all modelling conditions. Across domains, post hoc, model-agnostic explainability techniques—particularly SHAP and LIME—dominate current practice, whereas model-specific approaches such as Grad-CAM are primarily applied in image- and sensor-based contexts. Despite technical advances, the explicit operationalisation of fairness, accountability, and governance mechanisms remains limited and uneven. The findings suggest that explainability functions as a performance-preserving design tendency in the existing literature, supporting transparency, epistemic justification, and calibrated stakeholder trust without systematically compromising predictive performance.</p><h3>Graphical abstract</h3><div><figure><div><div><picture><source><img></source></picture></div></div></figure></div></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 9","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11566-x.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148466433","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-agent task and motion planning trends analysis: a survey","authors":"Xi Tao,&nbsp;José-Fernán Martínez-Ortega,&nbsp;Néstor Lucas-Martínez,&nbsp;Mario San-Emeterio-de-la-Parte","doi":"10.1007/s10462-026-11588-5","DOIUrl":"10.1007/s10462-026-11588-5","url":null,"abstract":"<div><p>Multi-agent Task and Motion Planning (TAMP) has become a key research frontier in autonomous agents, supported by the rapid evolution of Embodied AI and the increasing deployment of robot teams in real-world environments. Unlike single-agent planning, multi-agent TAMP must jointly reason about discrete task allocation, continuous motion feasibility, communication constraints, and execution uncertainties. This survey provides a systematic analysis of recent advances in multi-agent TAMP from 2021 to 2025. We first integrate and refine existing taxonomies, and propose a unified classification framework based on three dimensions. A total of 60 representative works are reviewed under this taxonomy, revealing dominant solution strategies. Furthermore, this survey summarizes emerging research trends, including LLM-based TAMP, decentralized adaptive planning frameworks, and reinforcement learning based TAMP solutions. Finally, we highlight key future directions, such as VLM-based planning, Theory-based planning, data privacy and long-horizon task planning.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 9","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11588-5.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148752128","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Deep learning techniques for otologic imaging: a systematic review of segmentation methods","authors":"Zainab Ghazanfar,&nbsp;Haill An,&nbsp;Saba Ghazanfar Ali,&nbsp;Bin Sheng,&nbsp;Younhyun Jung","doi":"10.1007/s10462-026-11586-7","DOIUrl":"10.1007/s10462-026-11586-7","url":null,"abstract":"<div><p>Developing segmentation models for otologic structures in medical images is fundamental for automated analysis and supporting clinical decision-making in ear-related patient care and surgical management. However, this task remains highly challenging because the ear contains small and morphologically complex substructures that often exhibit low visual contrast against surrounding tissues, making manual annotation labor-intensive and prone to inconsistency. Recent advances in deep learning (DL) have transformed medical image analysis across multiple clinical specialties, presenting new architectures to overcome these challenges. Beyond conventional convolutional neural networks (CNN), state-of-the-art approaches include vision transformers, diffusion models, and large-scale foundation models capable of capturing long-range dependencies, modelling sub-millimeter anatomy, and generalizing across variable imaging conditions. This systematic review provides the most comprehensive evaluation to date of how these DL models have been applied to otologic segmentation. Although DL has been widely reviewed in broader medical imaging, dedicated coverage of otologic applications is limited. Existing otologic reviews are typically narrow in scope, often confined to the middle ear, and overlook many clinically relevant anatomical domains. They also primarily emphasize CNN-based architectures and do not incorporate recent advances from 2023 to 2025, including transformer-based, diffusion-based, and hybrid architectures that integrate domain-prior strategies for enhanced contextual modelling. To address this gap, our review synthesizes research across all major otologic substructures and traces the evolution of DL architectures from CNN-based to hybrid and emerging models. We identify key performance trends and highlight persistent barriers to clinical translation, including severe class imbalance, limited multi-center datasets, variability in imaging, heterogeneous annotation standards, and high computational cost that restrict deployment in hospital settings. Based on these insights, we propose forward-looking research priorities to accelerate translation into clinical workflows. These include explainable and trustworthy DL models to support clinician confidence and regulatory approval, lightweight architectures suitable for diverse resource environments, domain adaptation across scanners and institutions, and scalable deployment frameworks compatible with diagnostic and surgical navigation systems. By integrating current advances, limitations, and future opportunities, this review presents a structured roadmap for developing clinically deployable DL models for otologic segmentation and outlines how these technologies can enhance future diagnostic and surgical practice.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 9","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11586-7.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148466452","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}