[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84401-en":3,"doc-seo-84401-105":30,"detail-sidebar-cat-0-en-105":92},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},84401,7971461741311,"Ophelia","https://ap-avatar.wpscdn.com/avatar/74000253aff267980c6?x-image-process=image/resize,m_fixed,w_180,h_180&k=1779345379180704826",8,"Research & Report","Multi-Resolution Feature Stem for Diabetic Retinopathy Lesion Segmentation","Diabetic retinopathy (DR) is a leading cause of preventable blindness that depends on accurate lesion segmentation for early detection and monitoring. Lesions vary greatly in size, from tiny microaneurysms to larger hemorrhages and exudates, creating conflicting demands on model architecture and input resolution. Experiments across U-Net, UNet++, Vision Transformers, and DeepLabV3+ at 512×512 and 1024×1024 reveal opposing resolution effects by lesion type. The work proposes a Multi-Resolution Feature Stem that integrates an input pyramid into a UNet++ backbone to retain fine detail and context while resolving this trade-off efficiently.","Multi-Resolution Feature Stem for Diabetic Retinopathy lesion segmentation  \nIndranil Dutta, Taehee Jeong  \nSan Jose State University  \nindranil.dutta, [taehee.jeong @sjsu.edu](taehee.jeong @sjsu.edu)  \narXiv :2607 .08679v 1 [ cs .CV] 9 Jul 2026  \nAbstract—Diabetic Retinopathy (DR) is a leading cause of preventable blindness worldwide, requiring automated lesion segmentation using deep learning models for early detection and monitoring. However, DR lesions vary dramatically in size from tiny microaneurysms to large hemorrhages and exudates. This variability creates conflicting demands on the model architecture and input resolution, posing a challenge for effective design. This work investigates the impact of input resolution on different lesion types. Through systematic experimentation with multiple architectures (U-Net, UNet++, Vision Transformers, DeepLabV3+) at 512 × 512 and 1024 × 1024 resolutions, we identify a critical, counter-intuitive phenomenon where increasing input resolution has opposing effects on different lesion types. We demonstrate that while higher resolution is essential for resolving fine-grained microaneurysms, it can unexpectedly degrade performance on larger hemorrhages. This finding challenges the common assumption that higher resolution is uniformly beneficial. To address this, we propose a novel Multi-Resolution Feature Stem, an input-level pyramid integrated with a UNet++ backbone. This architecture processes multiple scales in parallel, capturing fine-grained details without sacrificing contextual information. This work contributes crucial empirical evidence of this complex, resolution-dependent behavior and a practical, parameterefficient architecture that successfully resolves this trade-off. Our code is available at [https://github.com/taeheej/Multi-Resolution](https://github.com/taeheej/Multi-Resolution)Feature-Stem-for-Diabetic-Retinopathy-lesion-segmentation.  \nIndex Terms—Diabetic retinopathy, Lesion segmentation, Multi-Resolution, Pyramid integration  \nI. INTRODUCTION  \nDiabetic retinopathy (DR) is a microvascular complication of diabetes mellitus and remains one of the leading causes of preventable blindness among working-age adults worldwide [1] . 589 million adults are currently living with diabetes globally. The number is projected to reach 853 million by 2050 [2] . This increase is particularly pronounced in Asian populations, where traditional diets rich in carbohydrates combined with rapid urbanization have contributed to alarming rates of type 2 diabetes. Early detection and timely monitoring of disease progression are required to potentially prevent vision loss in millions of patients [3] .  \nCurrent clinical practice for the screening of diabetic retinopathy involves manual examination and annotation of  \nThis work was supported in part by a Mobilint Grant awarded to San Jose State University. (Corresponding author: Taehee Jeong)  \n2026 International Conference on Advances in Artificial Intelligence and Machine Learning (AAIML), 20-22 March 2026, IEEE Copyright 2026  \n(a) Healthy Retina (b) Early DR with mi- (c) Moderate DR with  \ncroaneurysms hemorrhages and hard  \nexudates  \n(d) Severe DR with soft exudates and cotton wool spots  \n(e) Proliferative DR  \n(f) Varying lesion types and severity  \nFig. 1: Fundus photographs illustrating various DR lesions  \nfundus images by trained ophthalmologists. This process is time-consuming, expensive, and creates significant bottlenecksin patient care. Patients often experience delays of two weeks or longer between image acquisition and receiving diagnostic results [4] . This delays might result in missed opportunities for early intervention during critical disease stages [5] . Furthermore, manual annotation is subject to inter-observer variability, with studies showing disagreement rates of about 15% among expert retina specialists [6] .  \nAutomated lesion segmentation using deep learning models offers the potential for scalable, consistent, ","cbCairzOqBCItaQM","https://ap.wps.com/l/cbCairzOqBCItaQM","pdf",13799733,5,1,6,"English","en",105,"# Introduction\n# Related Work","[{\"question\":\"What problem does the paper address in diabetic retinopathy lesion segmentation?\",\"answer\":\"It addresses how drastically different lesion sizes (microaneurysms vs. hemorrhages/exudates) create competing requirements for input resolution and model design. The goal is more effective segmentation for diverse lesion types.\"},{\"question\":\"What main finding does the paper report about input resolution?\",\"answer\":\"Increasing input resolution does not help uniformly: it benefits fine-grained microaneurysms but can degrade performance for larger hemorrhages. The effects are counter-intuitive and depend on lesion type.\"},{\"question\":\"How does the proposed Multi-Resolution Feature Stem work?\",\"answer\":\"It builds an input pyramid at multiple scales (e.g., 1024×1024, 512×512, 256×256), processes each scale through shared convolutional layers, fuses multi-scale features, and integrates the fused representation into the first layer of a UNet++ encoder.\"}]",1784195328,15,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":87,"head_meta":89,"extra_data":91,"updated_unix":28},"multi-resolution-feature-stem-for-diabetic-retinopathy-lesion-segmentation","",{"@graph":36,"@context":86},[37,54,69],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/multi-resolution-feature-stem-for-diabetic-retinopathy-lesion-segmentation/84401/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":24,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-07-28","2026-07-16",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82],{"name":73,"@type":74,"acceptedAnswer":75},"What problem does the paper address in diabetic retinopathy lesion segmentation?","Question",{"text":76,"@type":77},"It addresses how drastically different lesion sizes (microaneurysms vs. hemorrhages/exudates) create competing requirements for input resolution and model design. The goal is more effective segmentation for diverse lesion types.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"What main finding does the paper report about input resolution?",{"text":81,"@type":77},"Increasing input resolution does not help uniformly: it benefits fine-grained microaneurysms but can degrade performance for larger hemorrhages. The effects are counter-intuitive and depend on lesion type.",{"name":83,"@type":74,"acceptedAnswer":84},"How does the proposed Multi-Resolution Feature Stem work?",{"text":85,"@type":77},"It builds an input pyramid at multiple scales (e.g., 1024×1024, 512×512, 256×256), processes each scale through shared convolutional layers, fuses multi-scale features, and integrates the fused representation into the first layer of a UNet++ encoder.","https://schema.org",{"og:url":52,"og:type":88,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":90,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":93},[94,98,102,106,110,114,119,122,127,130,134],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},"Comic",60,"comic",{"id":22,"doc_module":4,"doc_module_name":46,"category_name":111,"show_sort_weight":112,"slug":113},"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":46,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":20,"slug":137},19,"General","general"]