[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-marginal-loss-wins-when-heart-scan-data-is-incomplete":10,"sections":41},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":30,"tags":31,"sources":36,"feedback":40,"feedback_at":22,"cost_usd":40,"total_tokens":40},4005,"marginal-loss-wins-when-heart-scan-data-is-incomplete","Marginal Loss Wins When Heart Scan Data Is Incomplete","A head-to-head study of three loss functions finds marginal loss most robust when AI models train on echocardiography data missing multiple label types.","Training cardiac AI on mismatched, incomplete datasets has a clear winner — but only under the hardest conditions.\n\nResearchers ran a systematic comparison of three loss functions — adaptive categorical cross entropy (aCCE), marginal loss, and adaptive binary cross entropy (aBCE) — across echocardiography segmentation tasks to determine which handles partially-labelled training data best. Echocardiography is the standard first-line tool for assessing heart function, and automating the segmentation of cardiac structures matters because it feeds the biomarkers clinicians actually use. The study tested each loss function across intra-domain and inter-domain scenarios, with both single and multiple missing labels, and at varying ratios of fully-labelled to partially-labelled examples. It is the first study to tackle the multi-domain, partial-label combination specifically in echocardiography segmentation.\n\nAll three loss functions held their own when training and test data came from the same domain — a relatively clean scenario. The gap opened in inter-domain tasks, where models face a distribution shift between training sources: aBCE and marginal loss both pulled ahead when one label was missing, but marginal loss was the only method that stayed robust when more than one label was absent simultaneously. That distinction matters because real-world clinical datasets are assembled from multiple hospitals and scanner types, and incomplete labelling is the norm, not the exception.\n\nThe study frames its finding carefully — no single loss function dominates every scenario — which is the right call, but it also means clinical ML teams deploying across heterogeneous data sources should treat marginal loss as a strong default rather than a universal answer.","[\"medical ai\",\"deep learning\",\"computer vision\",\"research\"]","2026-07-07T04:00:00.000Z","2026-07-07T14:36:54.715Z","2026-07-07T14:36:57.526Z","published",null,[24],{"id":25,"reviewer":26,"round":27,"reason":28,"status":29},"editor-r1","editor",1,"The headline and dek are vague and informal — 'Which Loss Function Wins' reads as a working placeholder rather than a finished publication-ready headline that states the actual finding (marginal loss outperforms rivals in the multi-missing-label scenario).","resolved","ai",[32,33,34,35],"medical ai","deep learning","computer vision","research",[37],{"name":38,"url":39},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.05008",0,{"sections":42},[43,47,52,57,62,67,72,77,82,86,91,95,100,105],{"name":44,"slug":30,"count":45,"latest_published_at":46},"AI",2590,"2026-07-16T04:00:00.000Z",{"name":48,"slug":49,"count":50,"latest_published_at":51},"Security","security",294,"2026-07-15T19:59:48.000Z",{"name":53,"slug":54,"count":55,"latest_published_at":56},"Deals","deals",179,"2026-06-29T20:02:07.000Z",{"name":58,"slug":59,"count":60,"latest_published_at":61},"Policy","policy",158,"2026-07-16T00:02:48.000Z",{"name":63,"slug":64,"count":65,"latest_published_at":66},"Hardware","hardware",122,"2026-07-14T19:46:26.000Z",{"name":68,"slug":69,"count":70,"latest_published_at":71},"Consumer Tech","consumer-tech",93,"2026-07-13T13:20:48.000Z",{"name":73,"slug":74,"count":75,"latest_published_at":76},"Software","software",70,"2026-07-13T19:52:25.000Z",{"name":78,"slug":79,"count":80,"latest_published_at":81},"Science","science",66,"2026-07-10T10:29:37.000Z",{"name":83,"slug":84,"count":85,"latest_published_at":18},"Dev Tools","dev-tools",59,{"name":87,"slug":88,"count":89,"latest_published_at":90},"Gaming","gaming",41,"2026-07-09T04:00:00.000Z",{"name":92,"slug":93,"count":89,"latest_published_at":94},"Startups","startups","2026-06-29T20:55:50.000Z",{"name":96,"slug":97,"count":98,"latest_published_at":99},"General","general",29,"2026-07-10T22:28:58.000Z",{"name":101,"slug":102,"count":103,"latest_published_at":104},"Reviews","reviews",20,"2026-06-24T12:00:01.000Z",{"name":106,"slug":107,"count":108,"latest_published_at":109},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]