{"paper":{"title":"Uncovering Latent Pathological Signatures in Pulmonary CT via Cross-Window Knowledge Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Distilling knowledge from the best CT window transfers latent pathological signatures to students on other windows and raises per-window AUC by 10-16 points.","cross_cats":["cs.AI","cs.CV"],"primary_cat":"eess.IV","authors_text":"Bo Peng, Daqian Shi, Honghan Wu, Jing Gao, Johan Thygesen, Kun Wang, Na Wang, Tian Li, Wujian Xu, Ximing Liao, Yingqun Ji","submitted_at":"2026-05-12T03:40:38Z","abstract_excerpt":"Multi-window CT imaging captures complementary pathological information across anatomical structures of differing densities, yet existing deep learning methods fuse representations only at later stages, missing cross-density interactions. We propose a cross-window knowledge distillation framework in which student encoders learn latent clinical priors from a teacher trained on the most informative window. Evaluated retrospectively on three cohorts - COPD-CT-DF (n=719), RSNA PE (n=1,433), and an in-house CTEPD dataset (n=161) - distillation improved per-window AUC by 10.1-16.5 percentage points "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"distillation improved per-window AUC by 10.1-16.5 percentage points on COPD-CT-DF (0.75-0.81 to 0.90-0.94; all P<0.001), with ensemble AUC reaching 0.9960. Similar gains were observed on RSNA PE (0.80-0.83 to 0.90-0.92) and CTEPD (AUC 0.7481 vs. 0.6264).","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the teacher model trained on the single most informative window captures all clinically relevant cross-density pathological signatures that can be successfully distilled to student encoders on other windows without introducing bias or loss of information specific to those windows.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Cross-window knowledge distillation raises per-window AUC by 10-16 points in pulmonary CT by transferring latent pathological signatures from a teacher encoder on the most informative window to students on other windows.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Distilling knowledge from the best CT window transfers latent pathological signatures to students on other windows and raises per-window AUC by 10-16 points.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"f1e1547afae2bd6594e8a57ff8ba7c750c0ee9ab0c167fa3512dcdd35289883d"},"source":{"id":"2605.12562","kind":"arxiv","version":1},"verdict":{"id":"1e7f06f6-1c7a-4154-8011-889deec95f11","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T20:50:13.032150Z","strongest_claim":"distillation improved per-window AUC by 10.1-16.5 percentage points on COPD-CT-DF (0.75-0.81 to 0.90-0.94; all P<0.001), with ensemble AUC reaching 0.9960. Similar gains were observed on RSNA PE (0.80-0.83 to 0.90-0.92) and CTEPD (AUC 0.7481 vs. 0.6264).","one_line_summary":"Cross-window knowledge distillation raises per-window AUC by 10-16 points in pulmonary CT by transferring latent pathological signatures from a teacher encoder on the most informative window to students on other windows.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the teacher model trained on the single most informative window captures all clinically relevant cross-density pathological signatures that can be successfully distilled to student encoders on other windows without introducing bias or loss of information specific to those windows.","pith_extraction_headline":"Distilling knowledge from the best CT window transfers latent pathological signatures to students on other windows and raises per-window AUC by 10-16 points."},"references":{"count":22,"sample":[{"doi":"10.1186/s12931-024-02913-z","year":2024,"title":"Artificial intelligence in COPD CT images: identification, staging, and quantitation.Respir Res, 25(1):319, 2024","work_id":"ef27bcd5-42f1-4997-b30c-5c7725a6a783","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2015,"title":"Fundamentals of Radiology","work_id":"6279b252-480f-4c7a-99a4-29370f481a25","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"The CT pulmonary vascular parameters and disease severity in COPD patients on acute exacerbation: a correlation analysis.BMC Pulm Med, 21(1):34, 2021","work_id":"efa2f0fc-ef85-425c-ab5f-853b8a31d3a0","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2015,"title":"Bartolome R Celli, Marc Decramer, Jadwiga A Wedzicha, Kevin C Wilson, Alvar Agust´ ı, Gerard J Criner, et al. An official American Thoracic Society/European Respiratory Society statement: research que","work_id":"c8aa9c4f-8eab-4d0d-bcca-969621034667","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"High performance with fewer labels using semi-weakly supervised learning for pulmonary embolism diagnosis.NPJ Digit Med, 8(1):254, 2025","work_id":"ed849861-128c-4825-9649-e771d06f7cca","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":22,"snapshot_sha256":"f17ae9c9186620bd84a90b49e448480a386ac50a1ce303cfc2a03b2fa17b69df","internal_anchors":0},"formal_canon":{"evidence_count":1,"snapshot_sha256":"72deb389eb47e3227addd03deab41eb927dfb29e5997ee005e559990f2d1d45a"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}