{"paper":{"title":"Attention-Based Multimodal Survival Prediction with Cross-Modal Bilinear Fusion","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A multimodal model fuses histology, RNA-seq, and clinical data with low-rank bilinear pooling to predict survival more accurately than concatenation baselines.","cross_cats":["cs.LG"],"primary_cat":"q-bio.QM","authors_text":"Hassan Keshvarikhojasteh, Josien P.W. Pluim, Mitko Veta","submitted_at":"2026-05-12T13:09:25Z","abstract_excerpt":"We propose a novel multimodal deep learning framework for patient-level survival prediction, which integrates whole-slide histology features, RNA-seq expression profiles, and clinical variables. Our architecture combines an ABMIL module~\\cite{ilse2018attention} for slide-level representation with feedforward encoders for RNA and clinical data. These embeddings are then integrated through low-rank bilinear cross-modal fusion~\\cite{liu2018efficient} to model conditional interactions across modalities while controlling parameter growth. The model outputs continuous risk scores that are subsequent"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experiments on the CHIMERA challenge dataset demonstrate improved predictive performance over concatenation-based baselines and competitive generalization on hidden evaluation cohorts.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the low-rank bilinear fusion captures the clinically relevant conditional interactions across histology, RNA-seq, and clinical modalities without discarding important information, and that the Kaplan-Meier calibration step produces well-calibrated survival estimates on the target population.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"A multimodal survival model using attention-based histology features, RNA-seq encoders, and low-rank bilinear fusion shows improved performance over concatenation baselines on the CHIMERA dataset for HR-NMIBC.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A multimodal model fuses histology, RNA-seq, and clinical data with low-rank bilinear pooling to predict survival more accurately than concatenation baselines.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"b3993b958ac51b2b730172afc7503b864ea29c5bd1d1eef7b8a2dc2be817fa35"},"source":{"id":"2605.13897","kind":"arxiv","version":1},"verdict":{"id":"becee2bf-f037-44b2-ad6a-914b883069a7","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T05:16:51.392474Z","strongest_claim":"Experiments on the CHIMERA challenge dataset demonstrate improved predictive performance over concatenation-based baselines and competitive generalization on hidden evaluation cohorts.","one_line_summary":"A multimodal survival model using attention-based histology features, RNA-seq encoders, and low-rank bilinear fusion shows improved performance over concatenation baselines on the CHIMERA dataset for HR-NMIBC.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the low-rank bilinear fusion captures the clinically relevant conditional interactions across histology, RNA-seq, and clinical modalities without discarding important information, and that the Kaplan-Meier calibration step produces well-calibrated survival estimates on the target population.","pith_extraction_headline":"A multimodal model fuses histology, RNA-seq, and clinical data with low-rank bilinear pooling to predict survival more accurately than concatenation baselines."},"references":{"count":14,"sample":[{"doi":"","year":2025,"title":"Chimera challenge – combining histology, medical imaging and molecular data for medical prognosis and diagnosis.https://chimera.grand-challenge.org(2025), accessed: 2026-02-04","work_id":"8d5fd582-7abc-4be0-8dfa-8029f492cb82","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2019,"title":"IEEE Transactions on Pattern Analysis and Machine Intelligence 41(2), 423–443 (2019)","work_id":"3010b033-ce63-4355-bd4c-cbf7c1e976c5","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2019,"title":"Cheerla, A., Gevaert, O.: Deep learning with multimodal representation for pan- cancer prognosis prediction. Bioinformatics35(14), i446–i454 (2019)","work_id":"28a84a5e-c1bf-41e3-a05b-f0c54f773f51","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Nature medicine30(3), 850–862 (2024)","work_id":"40e0f487-5304-4a88-8814-b740cf6962ed","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"IEEE Transactions on Medical Imaging41(4), 757–770 (2020)","work_id":"b87cdb17-b3a2-4249-8473-0174b928cf43","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":14,"snapshot_sha256":"9cace80f7eb35c9a383e9cfb5cdb6f5ce2703f914139051ae3a9d41f54275251","internal_anchors":2},"formal_canon":{"evidence_count":2,"snapshot_sha256":"3f82d941553091023a2e56361bda911d684b707b46f82911436d3393757d28b1"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}