[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-328058-105":59,"doc-detail-328058-en":130},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":123,"head_meta":125,"extra_data":127,"updated_unix":129},105,"en","hierarchical-classification-via-cascading-feature-elimination-application-to-human-phenotype-ontology-aligned-facial-phenotyping-facemesh2hpo","Hierarchical Classification via Cascading Feature Elimination - Application to Human Phenotype Ontology-Aligned Facial Phenotyping (FaceMesh2HPO)","","Purpose: Many genetic disorders manifest with facial phenotypes, and clinicians trained to recognize specific traits can use structured morphological relationships to support diagnosis. Current image-based methods often produce syndrome-level “black-box” outputs and lack direct support for structured description of facial morphology. Here, FaceMesh2HPO classifies facial phenotypic descriptors aligned with the Human Phenotype Ontology (HPO).",{"@graph":69,"@context":122},[70,84,105],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/hierarchical-classification-via-cascading-feature-elimination-application-to-human-phenotype-ontology-aligned-facial-phenotyping-facemesh2hpo/328058/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/hierarchical-classification-via-cascading-feature-elimination-application-to-human-phenotype-ontology-aligned-facial-phenotyping-facemesh2hpo/328058.png","ImageObject",300,407,{"name":92,"@type":93},"Quinn","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-11-05","2026-09-21",true,{"@type":102,"interactionType":103,"userInteractionCount":24},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"What does FaceMesh2HPO aim to improve in facial phenotype-based diagnosis?","Question",{"text":112,"@type":113},"It provides a framework that classifies facial phenotypic descriptors aligned with the Human Phenotype Ontology, moving beyond black-box syndrome-level outputs toward structured facial morphology support for clinicians.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"How were training annotations and face data prepared for the model?",{"text":117,"@type":113},"A panel of 124 clinicians manually annotated a subset of GestaltMatcher Database images for 10 disorders using 107 HPO terms, refined with original curation. Non-syndromic reference faces were added from UTKFace, and 3D meshes were extracted from 2D images using 478 automatically detected points.",{"name":119,"@type":110,"acceptedAnswer":120},"Which model configuration produced the best performance, and what were the main findings?",{"text":121,"@type":113},"The strongest setup used 3D face meshes (including outline) plus age, sex, and ethnicity metadata, with iterative point elimination via a point-importance threshold of 0.01 and soft labels of 0.05 for negative samples. It achieved mean AUROC of 0.750 ± 0.042 in cross-validation, with 3D outperforming 2D and with hierarchical parent and compression nodes generally performing better than leaf nodes.","https://schema.org",{"og:url":83,"og:type":124,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":126,"canonical":83},"index,follow",{"doc_id":128,"site_id":62},328058,1792252188,{"code":4,"msg":5,"data":131},{"doc_id":128,"user_id":132,"nickname":92,"user_avatar":133,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":24,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":139,"language":140,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":67,"update_tm":144,"read_time":145,"preview_page_count":4},962075114765,"https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd","arXiv :2607 .05585v 1 [ cs .CV] 6 Jul 2026  \nHierarchical Classification via Cascading Feature Elimination: Application to Human Phenotype Ontology-Aligned Facial Phenotyping  \n(FaceMesh2HPO)  \nFabio Hellmann 1*, Alexander Hustinx2 , Benjamin D. Solomon3 , GestaltMatcher Database Consortium, Tzung-Chien Hsieh2 ,  \nPeter Krawitz2 , Elisabeth Andr´e 1  \n1* Chair for Human-centered Artificial Intelligence, University of Augsburg, Universit¨atsstr. 6a, Augsburg, 86159, Bavaria, Germany.  \n2 Institute for Genomic Statistics and Bioinformatics, University of Bonn, Venusberg-Campus 1, Bonn, 53127, North Rhine-Westphalia, Germany.  \n3 National Human Genome Research Institute, National Institutes of Health, 10 Center Dr, Bethesda, 20892, MD, United States of America.  \n*Corresponding author(s) . E-mail(s): [fabio.hellmann@informatik.uni-augsburg.de](fabio.hellmann@informatik.uni-augsburg.de) ; Contributing authors: alexander.hustinx@uni-bonn.de; [solomonb@mail.nih.gov](solomonb@mail.nih.gov) ; [thsieh@uni-bonn.de](thsieh@uni-bonn.de) ; [pkrawitz@uni-bonn.de](pkrawitz@uni-bonn.de) ;  \n[andre@informatik.uni-augsburg.de](andre@informatik.uni-augsburg.de) ;  \nAbstract  \nPurpose: Many genetic disorders manifest with facial phenotypes, and clinicians are trained to recognize specific traits or morphological relationships to support the diagnostic process. However, current image-based methods output syndromelevel predictions in a ”black-box” manner and do not directly support the structured description of facial morphology. Here, we introduce FaceMesh2HPO, a framework for classifying facial phenotypic descriptors aligned with the Human Phenotype Ontology (HPO) to support the diagnostic process.  \nMethods: A panel of 124 clinicians manually annotated a subset of GestaltMatcher Database images for 10 disorders with 107 total HPO terms (59 leaf and 48 parent terms), refining the original curation. We combined these annotations with non-syndromic reference faces from UTKFace and extracted 3D facial  \n1  \nmeshes with 478 automatically detected points from 2D images. We then trained a hierarchical, cascading classification pipeline of PointNet-based models organized along the HPO tree, using dynamically parameterized architectures and iterative point elimination that prunes mesh points according to term-specific importance-also called feature elimination.  \nResults: The best-performing configuration used 3D face meshes including the facial outline together with age, sex, and ethnicity metadata, a point-importance threshold of 0.01, and soft labels of 0.05 for negative samples. This configuration achieved a mean AUROC of 0.750 ± 0.042 across HPO models in cross-validation, with 3D outperforming 2D and both facial outline and metadata consistently improving performance. Per-node mean AUROCs ranged from ≈ 0.55 to ≈ 0.89, with parent and “compression” nodes near the root generally outperforming leaf nodes. The top five HPO models achieved AUROCs of ≈ 0.89 and F1-Scores of ≈ 0.89, whereas the weakest leaf models remained close to chance. On an external, independent test set, aggregated, normalized mean F1-score differences between test and validation varied by disorder: some unseen syndromes (e.g. , Seckel and Sotos syndromes) showed small differences, whereas others (e.g., Mowat–Wilson, Nicolaides–Baraitser, Floating–Harbor, FBXW7, and White–Sutton syndromes) showed larger deviations, indicating heterogeneous generalizability across disorders and ontology levels.  \nConclusion: Geometric representations of the face via 3D meshes, combined with a hierarchical PointNet architecture and cascaded point elimination along the HPO hierarchy, enable clinically meaningful classification of facial phenotypes. The model partially transfers to unseen disorders, especially at the level of parent HPO terms, while performance for specific leaf phenotypes remains constrained by underrepresented labels. These findings highlight the need for more diverse training cohorts","cbCailcgXfF5qaJH","https://ap.wps.com/l/cbCailcgXfF5qaJH","pdf",2522900,63,"English","# Abstract\n## Purpose\n## Methods\n## Results\n## Conclusion\n# Introduction","[{\"question\":\"What does FaceMesh2HPO aim to improve in facial phenotype-based diagnosis?\",\"answer\":\"It provides a framework that classifies facial phenotypic descriptors aligned with the Human Phenotype Ontology, moving beyond black-box syndrome-level outputs toward structured facial morphology support for clinicians.\"},{\"question\":\"How were training annotations and face data prepared for the model?\",\"answer\":\"A panel of 124 clinicians manually annotated a subset of GestaltMatcher Database images for 10 disorders using 107 HPO terms, refined with original curation. Non-syndromic reference faces were added from UTKFace, and 3D meshes were extracted from 2D images using 478 automatically detected points.\"},{\"question\":\"Which model configuration produced the best performance, and what were the main findings?\",\"answer\":\"The strongest setup used 3D face meshes (including outline) plus age, sex, and ethnicity metadata, with iterative point elimination via a point-importance threshold of 0.01 and soft labels of 0.05 for negative samples. It achieved mean AUROC of 0.750 ± 0.042 in cross-validation, with 3D outperforming 2D and with hierarchical parent and compression nodes generally performing better than leaf nodes.\"}]","Hierarchical Classification via Cascading Feature Elimination - Application to Human Phenotype Ontology-Aligned Facial Phenotyping (FaceMesh2HPO) | PDF",1789978585,159]