[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-120942-en":3,"doc-seo-120942-105":29,"detail-sidebar-cat-0-en-105":90},{"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":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":21,"html_lang":23,"table_of_contents":24,"faqs":25,"seo_title":26,"seo_description":14,"update_tm":27,"read_time":28},120942,549758252649,"Ivy","https://ap-avatar.wpscdn.com/avatar/8000253669c5317157?_k=1778319167496531819",8,"Research & Report","Utilizing Modern Machine Learning Approaches for Image Cytometry","Image cytometry segmentation tools are traditionally insufficient in precision, reliability, and general-purpose capability, creating major bottlenecks for bacterial imaging. This dissertation addresses pixel-level requirements for thousands of cells across hundreds of time points while maintaining performance across diverse cellular morphologies within a single micrograph. It analyzes failures of prior segmentation methods, then introduces machine learning frameworks tailored to unilaminar cell imaging contexts, including new encoding strategies and tracking-oriented representations.","©c Copyright 2023 Kevin Cutler  \nUtilizing modern machine learning approaches for image cytometry  \nKevin Cutler  \nA dissertation  \nsubmitted in partial fulfillment of the requirements for the degree of  \nDoctor of Philosophy  \nUniversity of Washington  \n2023  \nReading Committee:  \nJoseph Mougous, Chair  \nPaul Wiggins, Chair  \nArmita Nourmohammad  \nProgram Authorized to Offer Degree: Physics  \nUniversity of Washington  \nAbstract  \nUtilizing modern machine learning approaches  \nfor image cytometry  \nKevin Cutler  \nCo-Chairs of the Supervisory Committee:  \nProfessor Joseph Mougous  \nMicrobiology  \nProfessor Paul Wiggins  \nPhysics  \nUntil recently, the scientific community has lacked image segmentation tools that are precise, reliable, and general-purpose. Such tools are especially needed in applications to bacterial image cytometry, wherein single-pixel precision is needed, perfect segmentation must be achieved over thousands of cells over hundreds of time points, and the approach must be applicable to a diversity of cellular morphologies present in a single micrograph. In this document, I detail the challenges of bacterial image segmentation, the failures of prior approaches, and the use of machine learning to solve this problem in virtually any unilaminar cell imaging context.  \nTABLE OF CONTENTS  \nPage  \nList of Figures ....................................... iii  \nGlossary ........................................... v  \nChapter 1: Introduction .................................. 1  \n1.1 Strategies for encoding image segmentation ................... 1  \n1.2 Strategies for achieving image segmentation .................. 4  \n1.3 The challenge of morphology ........................... 8  \nChapter 2: Prior work ................................... 10  \n2.1 Rationale behind the selection of segmentation algorithms .......... 10  \n2.2 Acquisition and annotation of ground-truth data ................ 11  \n2.3 Training and tuning segmentation algorithms .................. 12  \n2.4 Quantifying segmentation performance ..................... 15  \n2.5 Motivation for a new DNN-based segmentation algorithm ........... 17  \nChapter 3: Omnipose .................................... 22  \n3.1 Prediction classes ................................. 22  \n3.2 Loss functions ................................... 26  \n3.3 Omnipose demonstrates unprecedented segmentation accuracy ........ 27  \n3.4 Nematode segmentation ............................. 29  \n3.5 Benchmarking beyond bioimages ........................ 30  \n3.6 3D segmentation ................................. 31  \n3.7 Sensitive detection of cell intoxication ...................... 37  \nChapter 4: Affinity graphs and Self-contact ........................ 40  \n4.1 The hierarchy of segmentation encoding .................... 41  \n4.2 Boundaries are not encoded by standard instance labels ............ 42  \n4.3 Affinity graphs can be derived from net pixel displacement .......... 44  \n4.4 Euler integration error-corrects DNN flow predictions ............. 46  \n4.5 Sine squared loss alleviates interpolation at cell boundaries .......... 47  \n4.6 Self-contact requires a modified annotation technique ............. 49  \n4.7 Affinity segmentation enables the analysis of elongated, self-contacting cell morphologies ................................... 51  \n4.8 Affinity segmentation enables septum tracking in dividing cells ........ 53  \nChapter 5: Spacetime segmentation ............................ 55  \n5.1 The problem of cell tracking ........................... 55  \n5.2 The problem of cell division ........................... 56  \n5.3 The solution in the high-frame-rate regime ................... 58  \n5.4 Spacetime training data ............................. 59  \n5.5 Extracting tracked cells from lineage affinity graphs .............. 59  \n5.6 Preliminary results ................................ 60  \n5.7 Future work .................................... 62  \nChapter 6: Image annotation ............................... 65  \n6.","cbCaieBwk6C8whnl","https://ap.wps.com/l/cbCaieBwk6C8whnl","pdf",15464015,1,105,"English","en","# Introduction\n## Strategies for encoding image segmentation\n## Strategies for achieving image segmentation\n## The challenge of morphology\n# Prior work\n## Rationale behind the selection of segmentation algorithms\n## Acquisition and annotation of ground-truth data\n## Training and tuning segmentation algorithms\n## Quantifying segmentation performance\n## Motivation for a new DNN-based segmentation algorithm\n# Omnipose\n## Prediction classes\n## Loss functions\n## Omnipose demonstrates unprecedented segmentation accuracy\n## Nematode segmentation\n## Benchmarking beyond bioimages\n## 3D segmentation\n## Sensitive detection of cell intoxication\n# Affinity graphs and Self-contact\n## The hierarchy of segmentation encoding\n## Boundaries are not encoded by standard instance labels\n## Affinity graphs can be derived from net pixel displacement\n## Euler integration error-corrects DNN flow predictions\n## Sine squared loss alleviates interpolation at cell boundaries\n## Self-contact requires a modified annotation technique\n## Affinity segmentation enables elongated, self-contacting morphologies\n## Affinity segmentation enables septum tracking in dividing cells\n# Spacetime segmentation\n## The problem of cell tracking\n## The problem of cell division\n## The solution in the high-frame-rate regime\n## Spacetime training data\n## Extracting tracked cells from lineage affinity graphs\n## Preliminary results\n## Future work\n# Image annotation\n## Garbage in, garbage out\n## Dataset and image sizes\n## Instance labels require fluorescent labels\n## The four-color theorem\n## Human in the loop\n## Time lapse annotation\n# Miscellaneous methods\n## Growth, density, and substrate\n## Phase contrast and fluorescence microscopy\n## Exposure and outliers\n## Gamma adjustment\n## Semantic gamma normalization\n## Dataset processing\n## Defining the Omnipose prediction classes\n# Outlook\n## The future of image segmentation algorithms\n## The future of ground truth data\n## The future of Omnipose","[{\"question\":\"What specific limitations in image cytometry segmentation does the dissertation address?\",\"answer\":\"It targets segmentation tools that lack precision, reliability, and general-purpose performance. The work emphasizes pixel-level accuracy, robust results across thousands of cells and many time points, and applicability to diverse morphologies within a single micrograph.\"},{\"question\":\"How does the document approach bacterial image segmentation challenges?\",\"answer\":\"It details challenges and previous failures, then applies machine learning methods designed for unilaminar cell imaging contexts. The dissertation focuses on improved encoding strategies and segmentation frameworks that better handle difficult topology and boundaries.\"},{\"question\":\"Which methods are introduced for segmentation and cell tracking?\",\"answer\":\"The dissertation presents Omnipose for advanced segmentation and introduces affinity graphs and self-contact handling. It also proposes spacetime segmentation to address cell tracking and division, including extracting tracked cells from lineage affinity graphs.\"}]","Utilizing Modern Machine Learning Approaches for Image Cytometry | PDF",1785732934,265,{"code":4,"msg":30,"data":31},"ok",{"site_id":21,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":85,"head_meta":87,"extra_data":89,"updated_unix":27},"utilizing-modern-machine-learning-approaches-for-image-cytometry","",{"@graph":35,"@context":84},[36,53,67],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/utilizing-modern-machine-learning-approaches-for-image-cytometry/120942/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":61,"encodingFormat":60,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-08-03",true,{"@type":64,"interactionType":65,"userInteractionCount":4},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"What specific limitations in image cytometry segmentation does the dissertation address?","Question",{"text":74,"@type":75},"It targets segmentation tools that lack precision, reliability, and general-purpose performance. The work emphasizes pixel-level accuracy, robust results across thousands of cells and many time points, and applicability to diverse morphologies within a single micrograph.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"How does the document approach bacterial image segmentation challenges?",{"text":79,"@type":75},"It details challenges and previous failures, then applies machine learning methods designed for unilaminar cell imaging contexts. The dissertation focuses on improved encoding strategies and segmentation frameworks that better handle difficult topology and boundaries.",{"name":81,"@type":72,"acceptedAnswer":82},"Which methods are introduced for segmentation and cell tracking?",{"text":83,"@type":75},"The dissertation presents Omnipose for advanced segmentation and introduces affinity graphs and self-contact handling. It also proposes spacetime segmentation to address cell tracking and division, including extracting tracked cells from lineage affinity graphs.","https://schema.org",{"og:url":51,"og:type":86,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":88,"canonical":51},"index,follow",{"doc_id":7,"site_id":21},{"code":4,"msg":5,"data":91},[92,96,100,104,109,114,119,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":93,"show_sort_weight":94,"slug":95},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":97,"show_sort_weight":98,"slug":99},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":101,"show_sort_weight":102,"slug":103},"Exam",70,"exam",{"id":105,"doc_module":4,"doc_module_name":45,"category_name":106,"show_sort_weight":107,"slug":108},5,"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":45,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":45,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":105,"slug":137},19,"General","general"]