[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-3087":3,"related-3087":56},{"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,"primary_topic":9,"secondary_topics":21,"classification":22,"search_phrases":39,"slug":40,"view_count":15,"doi":41,"paper":42,"created_at":55},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。",null,"INTERNATIONAL JOURNAL OF HEALTH AND PHARMACEUTICAL RESEARCH","2026-09-18T00:00:00Z","论文",10,false,0,{"impact":15,"substance":15,"depth":15,"authority":15,"freshness":15,"relevant":15,"comment":17},"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。",[19],{"name":10,"url":6},[12],[],{"status":23,"locked":14,"version":24,"method":25,"evidence":26},"needs_review","2026-10-22.2","rules",{"candidates":27,"agricultural_evidence":14,"reason":38},[28],{"code":29,"score":30,"supported":31,"title":32,"abstract":33,"tags":37},"I2",9,true,[],[34,35,36],"深度学习","deep learning","decision support",[],"农业相关性需要复核",[],"Loss-AwareResidualLearningforImbalancedM-3087","10.56201\u002Fijhpr.vol.11.no2.2026.pg1.41",{"doi":41,"openalex_id":43,"authors":44,"venue":10,"cited_by_count":15,"oa_url":47,"card":48,"direction":52,"ingested_from":54},"W7213544946",[45],{"name":46,"orcid":9},"Miracle Ugomma Anunobi","https:\u002F\u002Fiiardjournals.org\u002Fget\u002FIJHPR\u002FVOL. 11 NO. 2 2026\u002FLOSS AWARE RESIDUAL LEARNING 1-41.pdf",{"tldr":49,"method":50,"finding":51,"direction":52,"opportunity":53},"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","农业人工智能与决策模型","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。","openalex","2026-09-21T23:30:43.782875Z",{"total":15,"page":57,"page_size":57,"items":58},1,[]]