[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-3486":3,"related-3486":48},{"id":4,"title":5,"url":6,"summary":7,"summary_zh":8,"content":9,"source_name":10,"source_url":6,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":16,"sources":18,"tags":20,"search_phrases":21,"slug":22,"view_count":15,"doi":23,"paper":24,"created_at":47},3486,"starU‐Net: An enhanced U‐Net architecture with star operation and multi‐view fusion for improved vessel segmentation","https:\u002F\u002Fdoi.org\u002F10.1002\u002Fviw2.70197","Abstract Accurate vessel segmentation is crucial for diagnosing vascular diseases and supporting research in developmental biology. However, existing methods struggle to preserve fine capillaries and the topological continuity of vascular structures. In this study, we aim to develop a novel deep learning architecture that specializes in thin vessel segmentation for improved continuity. We propose starU‐Net, a four‐layer encoder‒decoder framework for vessel segmentation. It integrates three components in a problem‐driven design: (1) an enhanced feature extraction module leveraging a “star operation” to enable multiplicative feature interaction, (2) a shallow network design to minimize spatial degradation, and (3) a multi‐view feature fusion module using dynamic snake convolution to capture continuous tubular structures from multiple orientations. The model was trained and evaluated on three public retinal datasets and a novel, self‐constructed chick embryo dataset. Against CNN‐based, transformer‐based, and foundation‐model baselines, starU‐Net obtained the highest sensitivity, F 1 , area under the curve, and centreline Dice on all three public retinal datasets, at the cost of slightly lower specificity. The improvement in centreline Dice provides quantitative support for the reduction in thin vessel discontinuity. On the chick embryo dataset, whose vessels are markedly wider, the proposed starU‐Net was competitive but not leading. In summary, starU‐Net is an architecture that combines star operation‐based feature extraction with multi‐view tubular feature modeling to improve vascular segmentation. The proposed framework demonstrates strong performance in both clinical retinal imaging datasets and self‐constructed chick embryo angiogenesis data, serving as a promising computational tool for clinical decision support, biomedical research, and large‐scale vascular phenotyping in biomedical applications.","摘要　精确的血管分割对于诊断血管疾病和支持发育生物学研究至关重要。然而，现有方法难以保留细小毛细血管及血管结构的拓扑连续性。本研究旨在开发一种新型深度学习架构，专门针对细血管分割以改善其连续性。我们提出starU‐Net，一种用于血管分割的四层编码器‐解码器框架。该框架以问题驱动设计整合了三个组件：(1) 利用“星操作”实现乘性特征交互的增强特征提取模块，(2) 最小化空间退化的浅层网络设计，(3) 使用动态蛇形卷积从多方向捕获连续管状结构的多视角特征融合模块。该模型在三个公开视网膜数据集和一个新构建的鸡胚数据集上进行了训练和评估。与基于CNN、基于Transformer及基础模型的基线方法相比，starU‐Net在三个公开视网膜数据集上均获得了最高的灵敏度、F1值、曲线下面积和中心线Dice，代价是特异性略低。中心线Dice的提升为细血管不连续性的减少提供了定量支持。在血管明显更宽的鸡胚数据集上，所提出的starU‐Net具有竞争力但并非领先。总之，starU‐Net是一种结合基于星操作的特征提取与多视角管状特征建模以改进血管分割的架构。所提出的框架在临床视网膜成像数据集和自构建的鸡胚血管生成数据中均展现出强大性能，可作为临床决策支持、生物医学研究及生物医学应用中大规模血管表型分析的有前景的计算工具。",null,"Open Research (University of Surrey)","2026-09-22T00:00:00Z","论文",10,false,0,{"impact":15,"substance":15,"depth":15,"authority":15,"freshness":15,"relevant":15,"comment":17},"该论文聚焦医学血管分割的深度学习架构，与三农、农业信息化、智慧农业等主题无关，不予入选。",[19],{"name":10,"url":6},[12],[],"starU‐Net:AnenhancedU‐Netarchitecturewit-3486","10.1002\u002Fviw2.70197",{"doi":23,"openalex_id":25,"authors":26,"venue":10,"cited_by_count":15,"oa_url":38,"card":39,"direction":45,"ingested_from":46},"W7213971471",[27,29,32,35],{"name":28,"orcid":9},"Mengwei Bai",{"name":30,"orcid":31},"Tong Li","https:\u002F\u002Forcid.org\u002F0000-0003-1556-1241",{"name":33,"orcid":34},"Jing Lin","https:\u002F\u002Forcid.org\u002F0000-0001-9865-2098",{"name":36,"orcid":37},"Peng Fei Huang","https:\u002F\u002Forcid.org\u002F0000-0003-3651-7813","https:\u002F\u002Fonlinelibrary.wiley.com\u002Fdoi\u002Fpdfdirect\u002F10.1002\u002Fviw2.70197",{"tldr":40,"method":41,"finding":42,"direction":43,"opportunity":44},"提出starU-Net，用星操作和多视角融合提升血管分割连续性。","四层编解码器，星操作特征提取，动态蛇卷积多视角融合，视网膜和鸡胚数据。","在三个视网膜数据集上敏感度、F1、AUC和中心线Dice最高，细血管连续性改善。","农业人工智能与决策模型","可将该细管状结构分割架构迁移至农业场景，如作物根系、叶脉或灌溉管道提取。","农业遥感与作物表型","openalex","2026-09-25T23:30:22.913090Z",{"total":49,"page":50,"page_size":49,"items":51},6,1,[52,83,114,140,186,220],{"id":53,"title":54,"url":55,"summary":56,"summary_zh":57,"content":9,"source_name":58,"source_url":55,"published_at":59,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":60,"sources":62,"tags":64,"search_phrases":65,"slug":66,"view_count":15,"doi":67,"paper":68,"created_at":82},3476,"Explainability and Reliability in Large Language Model Systems: A Survey of SHAP, LIME and Attention Frameworks","https:\u002F\u002Fdoi.org\u002F10.67317\u002Fijmsrt26sep093","Increasing adoption of language models in sensitive sectors such as healthcare, finance, cybersecurity, education, etc., highlights the significance of developing transparent decision support systems. This paper reviews the use of explanation techniques for language and machine learning models, including SHAP, LIME, attention-based explanations and frameworks that transform outputs from models into natural language explanations. It turns out that the existing literature demonstrates that although classical language models demonstrate good predictive and semantic performance, they lack transparency. In this regard, SHAP and LIME help in the detection of feature importance and combination of frameworks lead to explanations being more understandable for users. The literature also outlines some limitations of the existing techniques such as explanation faithfulness, hallucinations, biases, privacy issues, computational complexity and lack of benchmarks for evaluation. The use of explainability in healthcare, finance, cybersecurity, industrial systems and IoT is discussed based on the evidence found in the literature.","语言模型在医疗、金融、网络安全、教育等敏感领域的应用日益广泛，这使得开发透明的决策支持系统变得尤为重要。本文综述了用于语言模型和机器学习模型的可解释性技术，包括SHAP、LIME、基于注意力的解释方法，以及将模型输出转化为自然语言解释的框架。现有文献表明，尽管经典语言模型在预测和语义方面表现良好，但缺乏透明度。在这方面，SHAP和LIME有助于识别特征重要性，而框架的组合则使解释更易于用户理解。文献还指出了现有技术的一些局限性，如解释忠实度、幻觉、偏见、隐私问题、计算复杂性以及缺乏评估基准。基于文献中发现的证据，本文讨论了可解释性在医疗、金融、网络安全、工业系统和物联网中的应用。","International Journal of Modern Science and Research Technology","2026-09-23T00:00:00Z",{"impact":15,"substance":15,"depth":15,"authority":15,"freshness":15,"relevant":15,"comment":61},"该文为通用大语言模型可解释性综述，未涉及三农或农业信息化场景，与平台主题无关。",[63],{"name":58,"url":55},[12],[],"ExplainabilityandReliabilityinLargeLangu-3476","10.67317\u002Fijmsrt26sep093",{"doi":67,"openalex_id":69,"authors":70,"venue":58,"cited_by_count":15,"oa_url":75,"card":76,"direction":81,"ingested_from":46},"W7214085362",[71,73],{"name":72,"orcid":9},"Mary Shakkina C",{"name":74,"orcid":9},"Miranda Lakshmi Travis","https:\u002F\u002Fwww.ijmsrt.com\u002Fstorages\u002Fdownload-paper\u002FIJMSRT26SEP093",{"tldr":77,"method":78,"finding":79,"direction":43,"opportunity":80},"综述大语言模型可解释性技术SHAP、LIME与注意力框架及其在敏感领域的应用。","文献综述，分析SHAP、LIME、注意力解释及自然语言解释框架。","经典模型预测好但缺乏透明；SHAP\u002FLIME可识别特征重要性，但存在忠实性、幻觉等局限。","农业大模型决策可解释性研究空白：构建面向农业场景的忠实性评估基准与可解释框架。","智慧农业 \u002F 农业物联网","2026-09-25T23:30:11.362189Z",{"id":84,"title":85,"url":86,"summary":87,"summary_zh":88,"content":9,"source_name":89,"source_url":86,"published_at":90,"category":12,"cover_url":9,"hotness":91,"is_selected":14,"score":15,"score_detail":92,"sources":94,"tags":98,"search_phrases":99,"slug":100,"view_count":15,"doi":101,"paper":102,"created_at":113},3346,"The Confluence of AI and Eco-Innovation: Shaping the Future of Multidisciplinary Research","https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.22912613","Abstract Artificial Intelligence (AI) and eco-innovation are increasingly converging to redefine how societies address environmental degradation, resource scarcity, and climate risk. This paper examines the intersection of these two domains as an emerging site of multidisciplinary inquiry, one that draws simultaneously on computer science, environmental economics, engineering, and policy studies. Rather than treating AI merely as a technical enabler bolted onto existing green technologies, the paper argues that AI is reshaping the very logic of eco-innovation—accelerating discovery cycles, enabling predictive resource management, and creating new business models built around circularity and efficiency. Drawing on a structured review of academic literature, industry reports, and policy documents published largely over the past decade, the study identifies recurring themes: AI-driven materials discovery, smart energy grids, precision agriculture, and algorithmic carbon accounting. It also surfaces tensions that are often underexplored, including the substantial energy footprint of AI systems themselves, questions of data governance in environmental monitoring, and the risk of techno-solutionism crowding out structural policy reform. Using a qualitative-interpretive methodology combined with thematic content analysis, the paper maps the current research landscape and proposes a framework for understanding AI-eco-innovation convergence across four layers: technological, organizational, institutional, and ecological. The paper concludes that meaningful progress in this space depends not on AI alone but on deliberate multidisciplinary collaboration among technologists, environmental scientists, economists, and regulators. It offers directions for future research, including longitudinal impact studies and frameworks for measuring the net environmental value of AI-enabled innovation.","摘要 人工智能（AI）与生态创新正日益融合，重新定义社会应对环境退化、资源稀缺和气候风险的方式。本文考察这两个领域的交汇点，将其视为一个新兴的多学科研究场域，同时涉及计算机科学、环境经济学、工程学和政策研究。本文并不将AI仅仅视为附加于现有绿色技术之上的技术赋能工具，而是认为AI正在重塑生态创新的内在逻辑——加速发现周期、实现预测性资源管理，并围绕循环性与效率创造新的商业模式。基于对过去十年间主要发表的学术文献、行业报告和政策文件的结构化综述，本研究识别出反复出现的主题：AI驱动的材料发现、智能电网、精准农业和算法碳核算。研究还揭示了常被忽视的张力，包括AI系统自身的巨大能源足迹、环境监测中的数据治理问题，以及技术解决方案主义排挤结构性政策改革的风险。本文采用定性-阐释方法并结合主题内容分析，描绘了当前的研究图景，并提出了一个理解AI-生态创新融合的框架，涵盖四个层面：技术、组织、制度和生态。本文的结论是，该领域的有意义进展不仅取决于AI本身，还取决于技术专家、环境科学家、经济学家和监管者之间自觉的多学科协作。本文提出了未来研究方向，包括纵向影响研究以及衡量AI赋能创新净环境价值的框架。","Zenodo (CERN European Organization for Nuclear Research)","2026-09-30T00:00:00Z",25,{"impact":15,"substance":15,"depth":15,"authority":15,"freshness":15,"relevant":15,"comment":93},"该论文讨论AI与生态创新的跨学科融合，未聚焦三农、农业信息化或智慧农业等本平台主题，相关性不足。",[95,96],{"name":89,"url":86},{"name":89,"url":97},"https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.22912614",[12],[],"TheConfluenceofAIandEco-Innovation:Shapi-3346","10.5281\u002Fzenodo.22912613",{"doi":101,"openalex_id":103,"authors":104,"venue":89,"cited_by_count":15,"oa_url":86,"card":107,"direction":81,"ingested_from":46},"W7214104704",[105],{"name":106,"orcid":9},"Saleha Javed Abbas Syed",{"tldr":108,"method":109,"finding":110,"direction":111,"opportunity":112},"综述AI与生态创新融合的多学科研究，提出四层分析框架并指出未来方向。","结构化文献综述与主题内容分析，涵盖学术、产业与政策文本。","AI正重塑生态创新逻辑，但需警惕能耗、数据治理与技术解决主义风险。","农业绿色发展与碳","可聚焦AI赋能农业碳核算与精准管理的净环境价值评估及跨学科治理框架。","2026-09-24T23:30:09.754901Z",{"id":115,"title":116,"url":117,"summary":118,"summary_zh":119,"content":9,"source_name":120,"source_url":117,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":121,"sources":123,"tags":125,"search_phrases":126,"slug":127,"view_count":15,"doi":128,"paper":129,"created_at":139},3281,"Economic Growth Nowcasting in Mauritania using Machine Learning and Satellite Data","https:\u002F\u002Fdoi.org\u002F10.63620\u002Fmkijbaft.2026.1027","This paper examines the role of satellite data and machine learning techniques in improving the nowcasting of real GDP in Mauritania, a context characterized by limited data availability and delays in official statistics. By combining traditional macroeconomic indicators with satellite-based variables such as Nighttime Lights and the Normalized Difference Vegetation Index (NDVI), the study develops a framework capable of capturing real-time economic dynamics. The results show that the inclusion of satellite data improves predictive accuracy, particularly when using non-linear models such as XGBoost. In particular, XGBoost records a reduction in RMSE from 0.019 to 0.017 and an increase in R² from 0.372 to 0.483 when satellite variables are included. Nighttime Lights are strongly correlated with economic activity, while NDVI exhibits more limited explanatory power. Overall, the findings highlight the potential of integrating alternative data sources and machine learning methods to enhance economic monitoring and support decision-making in data-constrained environments.","本文考察了卫星数据和机器学习技术在改进毛里塔尼亚实际GDP即时预测（nowcasting）中的作用，该国面临数据可得性有限和官方统计数据滞后的背景。通过将传统宏观经济指标与夜间灯光（Nighttime Lights）和归一化植被指数（NDVI）等卫星变量相结合，本研究构建了一个能够捕捉实时经济动态的框架。结果表明，纳入卫星数据提高了预测精度，尤其是在使用XGBoost等非线性模型时。具体而言，当纳入卫星变量后，XGBoost的均方根误差（RMSE）从0.019降至0.017，R²从0.372提升至0.483。夜间灯光与经济活动高度相关，而NDVI的解释力则较为有限。总体而言，研究结果凸显了整合替代数据源和机器学习方法以增强经济监测并支持数据受限环境下决策的潜力。","OpenAlex",{"impact":15,"substance":15,"depth":15,"authority":15,"freshness":15,"relevant":15,"comment":122},"研究毛里塔尼亚GDP即时预测，与三农、农业信息化、智慧农业无关，不予入选。",[124],{"name":120,"url":117},[12],[],"EconomicGrowthNowcastinginMauritaniausin-3281","10.63620\u002Fmkijbaft.2026.1027",{"doi":128,"openalex_id":130,"authors":131,"venue":9,"cited_by_count":15,"oa_url":117,"card":134,"direction":43,"ingested_from":46},"W7213984344",[132],{"name":133,"orcid":9},"Yahya Abou LY",{"tldr":135,"method":136,"finding":137,"direction":45,"opportunity":138},"结合卫星数据与机器学习改进毛里塔尼亚实时GDP预测。","使用夜间灯光和NDVI卫星变量，结合XGBoost等非线性模型。","加入卫星数据后XGBoost的RMSE降低、R²提升，夜间灯光解释力强。","可探索多源卫星数据融合与可解释AI，提升数据稀缺地区经济监测精度。","2026-09-23T23:30:35.177095Z",{"id":141,"title":142,"url":143,"summary":144,"summary_zh":145,"content":9,"source_name":146,"source_url":143,"published_at":147,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":148,"sources":150,"tags":152,"search_phrases":153,"slug":154,"view_count":15,"doi":155,"paper":156,"created_at":185},3275,"Reporting, verification, and accounting frameworks for carbon dioxide removal: Scientific integrity, permanence, and governance challenges across emerging pathways","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.rechem.2026.103860","Carbon dioxide removal (CDR) is becoming an important component of long-term climate mitigation. However, its climate value depends on whether removal claims reflect measurable, durable changes in atmospheric CO₂ rather than gross carbon capture or storage alone. This review evaluates monitoring, reporting, verification, and accounting (MRV\u002FA) across biological, biochar-based, geochemical, oceanic, and engineered CDR pathways using evidence published between 2015 and 2026. The synthesis is organized around four linked functions: measuring the physical intervention, validating the net atmospheric effect, recording the resulting carbon-accounting unit, and using it in markets, corporate claims, or national inventories. The evidence shows that credible accounting requires explicit treatment of counterfactual baselines, additionality, lifecycle emissions, leakage, timing, uncertainty, storage duration, and reversal risk. Remote sensing, field sensors, geochemical measurements, lifecycle assessment, computational models, and artificial intelligence (AI) can strengthen individual stages of MRV\u002FA, but their roles and validation requirements differ. Registries improve traceability but cannot compensate for weak measurement or accounting rules. Current policy and market frameworks also remain heterogeneous in how they treat durability, liability, uncertainty, and ownership. Future MRV\u002FA systems should therefore link pathway-specific measurements with transparent validation rules, time-dependent accounting, interoperable records, environmental performance, social legitimacy, and clearly assigned long-term responsibility.","二氧化碳去除（carbon dioxide removal, CDR）正成为长期气候减缓的重要组成部分。然而，其气候价值取决于去除声明是否反映了大气中CO₂可测量、持久的实际变化，而非仅反映碳捕集或封存的毛量。本综述基于2015年至2026年间发表的证据，评估了生物类、生物炭类、地球化学类、海洋类及工程类CDR路径的监测、报告、核查与核算（MRV\u002FA）。综述围绕四个相互关联的功能展开：测量物理干预、验证净大气效应、记录由此产生的碳核算单位，以及将其用于市场、企业声明或国家清单。证据表明，可信的核算需要明确处理反事实基线、额外性、生命周期排放、泄漏、时间尺度、不确定性、封存持续时间及逆转风险。遥感、实地传感器、地球化学测量、生命周期评估、计算模型和人工智能（AI）可强化MRV\u002FA的各个阶段，但其作用和验证要求各不相同。登记簿可提高可追溯性，但无法弥补测量或核算规则的薄弱。当前政策和市场框架在处理持久性、责任、不确定性和所有权方面也仍存在差异。因此，未来的MRV\u002FA系统应将路径特异性测量与透明的验证规则、时间依赖型核算、可互操作的记录、环境绩效、社会合法性及明确分配的长期责任相衔接。","Results in Chemistry","2026-09-20T00:00:00Z",{"impact":15,"substance":15,"depth":15,"authority":15,"freshness":15,"relevant":15,"comment":149},"该文为二氧化碳去除的MRV与核算框架综述，属气候环境科学领域，与三农、农业信息化、智慧农业等主题无直接关联，不建议进入每日精选。",[151],{"name":146,"url":143},[12],[],"Reporting,verification,andaccountingfram-3275","10.1016\u002Fj.rechem.2026.103860",{"doi":155,"openalex_id":157,"authors":158,"venue":146,"cited_by_count":15,"oa_url":179,"card":180,"direction":81,"ingested_from":46},"W7213742688",[159,162,164,166,168,170,172,175,177],{"name":160,"orcid":161},"Palanivendhan Murugadoss","https:\u002F\u002Forcid.org\u002F0000-0003-0388-1547",{"name":163,"orcid":9},"S. V. Niveditha",{"name":165,"orcid":9},"Sandeep Kumar Jain",{"name":167,"orcid":9},"Sujai Selvarajan",{"name":169,"orcid":9},"Sasmeeta Tripathy",{"name":171,"orcid":9},"Priya Parag Saxena",{"name":173,"orcid":174},"Ravikumar Jayabal","https:\u002F\u002Forcid.org\u002F0000-0001-7877-9913",{"name":176,"orcid":9},"Aseel Smerat",{"name":178,"orcid":9},"K. Kamakshi Priya","https:\u002F\u002Fwww.sciencedirect.com\u002Fscience\u002Farticle\u002Fpii\u002FS2211715626008349\u002Fpdf",{"tldr":181,"method":182,"finding":183,"direction":111,"opportunity":184},"综述2015-2026年各类二氧化碳去除路径的监测、报告、核查与核算框架及其科学完整性问题。","综述生物、生物炭、地球化学、海洋与工程化CDR的MRV\u002FA证据，涉及遥感、传感器","可信核算须显式处理基线、额外性、泄漏、时间与逆转风险；登记系统无法弥补测量与规则缺陷。","可研究农业CDR路径的路径特异性MRV指标与时间依赖核算规则，并探索AI验证与跨登记系统互操作。","2026-09-23T23:30:13.215315Z",{"id":187,"title":188,"url":189,"summary":190,"summary_zh":191,"content":9,"source_name":192,"source_url":189,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":193,"sources":195,"tags":197,"search_phrases":198,"slug":199,"view_count":15,"doi":200,"paper":201,"created_at":219},3170,"Cryospheric risks in the Himalayan region: impacts and community perceptions on snow avalanches and GLOFs in Chitral, Pakistan","https:\u002F\u002Fdoi.org\u002F10.1007\u002Fs11069-026-08414-0","Abstract Glacial Lake Outburst Floods (GLOFs) and snow avalanches pose recurrent threats to mountain communities across the Himalayas, Karakoram, and Hindukush regions. Recent glacier retreat, driven by rising temperatures, has intensified GLOF risk. Remote sensing-based inventories have mapped 3044 glacial lakes in Khyber Pakhtunkhwa (KPK) and Gilgit-Baltistan (GB). This study examines the impact of GLOFs and snow avalanches on livelihoods and infrastructure in Chitral District, as well as community perceptions of these hazards. Data were collected through household surveys in 10 villages, yielding 350 responses. Descriptive analysis was conducted to assess the extent of damage caused by these events. The findings reveal that while livestock losses were relatively minor, the effects on agriculture, crops, and infrastructure were substantial. Critical infrastructure, such as irrigation systems, roads, and bridges, required repairs that took 9 months or more to complete. A paired sample t-test further highlights a gap in risk awareness. While snow avalanches are widely recognized as a serious threat, local planning and preparedness efforts often underestimate GLOFs, despite their severe consequences. The study emphasizes the importance of involving local communities in decision-making and resilience planning to address these evolving hazards effectively.","摘要 冰川湖溃决洪水（GLOFs）和雪崩对喜马拉雅、喀喇昆仑和兴都库什地区的山区社区构成反复出现的威胁。近年来，气温上升导致的冰川退缩加剧了冰川湖溃决洪水的风险。基于遥感的编目已绘制了开伯尔-普什图省（KPK）和吉尔吉特-巴尔蒂斯坦（GB）的3044个冰川湖。本研究考察了冰川湖溃决洪水和雪崩对吉德拉尔县生计和基础设施的影响，以及社区对这些灾害的认知。数据通过在10个村庄开展入户调查收集，共获得350份回复。采用描述性分析评估这些事件造成的损失程度。研究发现，虽然牲畜损失相对较小，但对农业、作物和基础设施的影响十分严重。灌溉系统、道路和桥梁等关键基础设施所需的修复耗时9个月或更长。配对样本t检验进一步揭示了风险意识方面的差距。尽管雪崩被广泛认为是严重威胁，但地方规划和备灾工作往往低估了冰川湖溃决洪水，尽管其后果十分严重。本研究强调让当地社区参与决策和韧性规划的重要性，以有效应对这些不断演变的灾害。","Natural Hazards",{"impact":15,"substance":15,"depth":15,"authority":15,"freshness":15,"relevant":15,"comment":194},"该论文研究巴基斯坦喜马拉雅地区冰川湖溃决与雪崩风险，属自然灾害与山地社区研究，与三农、农业信息化、智慧农业等主题无关。",[196],{"name":192,"url":189},[12],[],"CryosphericrisksintheHimalayanregion:imp-3170","10.1007\u002Fs11069-026-08414-0",{"doi":200,"openalex_id":202,"authors":203,"venue":192,"cited_by_count":15,"oa_url":189,"card":214,"direction":45,"ingested_from":46},"W7213974324",[204,206,209,212],{"name":205,"orcid":9},"Syed ul Abrar",{"name":207,"orcid":208},"Irfan Ahmad Rana","https:\u002F\u002Forcid.org\u002F0000-0002-3157-1186",{"name":210,"orcid":211},"Shahbaz Altaf","https:\u002F\u002Forcid.org\u002F0000-0001-7846-4129",{"name":213,"orcid":9},"Muhammad Israr Siddiqui",{"tldr":215,"method":216,"finding":217,"direction":111,"opportunity":218},"研究巴基斯坦吉德拉尔地区雪崩与冰湖溃决洪水对生计和基础设施的影响及社区风险认知。","在10个村庄开展350份入户调查，用描述性统计和配对样本t检验分析。","农业、作物和基础设施受损严重，修复需9个月以上；社区低估冰湖溃决洪水风险。","可结合遥感冰湖编目与社区感知数据，构建山区农业基础设施脆弱性评估与预警模型。","2026-09-22T23:30:22.863654Z",{"id":221,"title":222,"url":223,"summary":224,"summary_zh":225,"content":9,"source_name":226,"source_url":223,"published_at":227,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":228,"sources":230,"tags":232,"search_phrases":233,"slug":234,"view_count":15,"doi":235,"paper":236,"created_at":247},3087,"Loss-Aware Residual Learning for Imbalanced Multi-Class Diabetic Retinopathy Diagnosis","https:\u002F\u002Fdoi.org\u002F10.56201\u002Fijhpr.vol.11.no2.2026.pg1.41","Diabetic Retinopathy (DR) is a severe complication of diabetes that can lead to vision impairment or blindness, making early detection crucial. Traditional deep learning models often struggle with class imbalance in medical datasets, leading to poor performance for minority classes. This study proposes a novel deep learning model based on Residual Networks (ResNet) to address the multi class classification of DR, with a focus on mitigating class imbalance. Standard Softmax activation functions tend to favor majority classes, thereby worsening the model's performance on minority classes. To address this, the study incorporates a custom loss function, Balanced Softmax Loss, which adjusts class weights to improve the recognition of minority classes. Additionally, the model integrates advanced techniques such as Squeeze-and-Excitation (SE) blocks, learnable wavelet transforms, and multi-head attention mechanisms to enhance feature extraction and model performance. The model was trained and evaluated on the APTOS 2019 Blindness Detection dataset, achieving a micro-average accuracy of 0.83 and a macro-average accuracy of 0.73 in the 5-class classification task. In a 4-class classification task, where severe and proliferative DR were merged, the model achieved a micro-average accuracy of 0.87 and a macro-average accuracy of 0.84. The model's interpretability was further enhanced through Explainable AI (XAI) techniques such as LIME, Grad-CAM, and SHAP. The trained model was deployed as a web-based application using Flask, enabling real-time classification of retinal images. The study highlights the model's effectiveness in addressing class imbalance and its potential for early DR diagnosis, thereby enhancing clinical decision support.","糖尿病视网膜病变（Diabetic Retinopathy, DR）是糖尿病的一种严重并发症，可导致视力损害甚至失明，因此早期检测至关重要。传统的深度学习模型在处理医学数据集中的类别不平衡问题时往往表现不佳，导致少数类别的识别性能较差。本研究提出了一种基于残差网络（Residual Networks, ResNet）的新型深度学习模型，用于DR的多类别分类，重点在于缓解类别不平衡问题。标准Softmax激活函数倾向于偏向多数类别，从而加剧了模型在少数类别上的性能不足。为解决这一问题，本研究引入了一种自定义损失函数——平衡Softmax损失（Balanced Softmax Loss），通过调整类别权重来提升少数类别的识别能力。此外，该模型还集成了挤压-激励（Squeeze-and-Excitation, SE）模块、可学习小波变换和多头注意力机制等先进技术，以增强特征提取和模型性能。该模型在APTOS 2019盲ness Detection数据集上进行了训练和评估，在5分类任务中取得了0.83的微平均准确率和0.73的宏平均准确率。在将严重型和增殖型DR合并的4分类任务中，模型取得了0.87的微平均准确率和0.84的宏平均准确率。通过LIME、Grad-CAM和SHAP等可解释人工智能（Explainable AI, XAI）技术，进一步增强了模型的可解释性。训练后的模型通过Flask部署为基于Web的应用程序，实现了视网膜图像的实时分类。本研究凸显了该模型在解决类别不平衡问题方面的有效性及其在DR早期诊断中的潜力，从而增强了临床决策支持。","INTERNATIONAL JOURNAL OF HEALTH AND PHARMACEUTICAL RESEARCH","2026-09-18T00:00:00Z",{"impact":15,"substance":15,"depth":15,"authority":15,"freshness":15,"relevant":15,"comment":229},"该论文研究糖尿病视网膜病变的深度学习诊断，属医学影像领域，与三农、农业信息化、智慧农业无关，不建议进入每日精选。",[231],{"name":226,"url":223},[12],[],"Loss-AwareResidualLearningforImbalancedM-3087","10.56201\u002Fijhpr.vol.11.no2.2026.pg1.41",{"doi":235,"openalex_id":237,"authors":238,"venue":226,"cited_by_count":15,"oa_url":241,"card":242,"direction":43,"ingested_from":46},"W7213544946",[239],{"name":240,"orcid":9},"Miracle Ugomma Anunobi","https:\u002F\u002Fiiardjournals.org\u002Fget\u002FIJHPR\u002FVOL. 11 NO. 2 2026\u002FLOSS AWARE RESIDUAL LEARNING 1-41.pdf",{"tldr":243,"method":244,"finding":245,"direction":43,"opportunity":246},"提出基于ResNet的损失感知残差学习模型，解决糖尿病视网膜病变多分类中的类别不平衡问题。","ResNet结合平衡Softmax损失、SE块、可学习小波变换和多头注意力，在A","5类分类微平均准确率0.83、宏平均0.73；合并严重与增殖期后4类分类微平均0.87、宏平均0.8","将损失感知残差学习与类别不平衡处理策略迁移至农业病害多分类诊断，提升少数类识别能力。","2026-09-21T23:30:43.782875Z"]