{"doi":"10.1145/3765612.3767199","title":"Fine-Grained Interaction and Task-Driven Fusion of Histology Images and Genomics for Multimodal Cancer Survival Prediction","abstract":"Cancer survival analysis aims to estimate patients' risk of death to aid in prognosis, benefiting from multimodal learning, especially through the integration of gigapixel histology whole-slide images (WSIs) and genomic profiles. Typically, WSIs are divided into smaller patches (often exceeding 104 patches), serving as histology tokens, while genomic data is categorized into distinct gene groups (e.g., established biological pathways) as gene tokens. Recent studies have employed a gene-queried cross-attention mechanism to select and aggregate histology tokens, effectively reducing the space complexity of WSIs. However, without proper regularization and constraints, different gene-queried cross-attention maps often exhibit similar distributions, leading to redundant updates in histology tokens. To address this, we propose a novel regularization technique for gene-queried cross-attention maps, ensuring that the updated histology tokens capture diverse and meaningful patterns. Building on this, we design a task-driven informativeness learning module that operates on the final token features within each modality to enable trustworthy survival prediction. Specifically, the informativeness of token features is assessed by considering the consistency between each token's informativeness and its survival-prediction performance. Extensive experiments on five datasets from The Cancer Genome Atlas (TCGA) demonstrate the effectiveness of our method. The code is available at https://github.com/XulinChen/ACM-BCB2025_FITF.","journal":null,"year":2025,"id":586259,"datarank":0.0,"base_score":0.0,"endowment":0.0,"self_citation_contribution":0.0,"citation_network_contribution":0.0,"self_endowment_contribution":0.0,"citer_contribution":0.0,"corpus_percentile":null,"corpus_rank":null,"citation_count":0,"citer_count":0,"citers_with_citation_signal":0,"citers_with_endowment":0,"datacite_reuse_total":0,"is_dataset":false,"is_dataset_confidence":0.814,"is_data_producer":false,"deposit_databanks":null,"is_oa":true,"file_count":0,"downloads":0,"has_version_chain":false,"published_date":"2025-01-01","fair_score":null,"fair_percentile":null,"algorithm_id":"datarank_citation_only_1hop_v6","ranking_scope":"data_only","authors":[{"id":259336,"name":"Junzhou Huang","orcid":"0000-0002-9548-1227","position":1,"is_corresponding":false},{"id":1500274,"name":"Xulin Chen","orcid":"0009-0000-6682-0472","position":0,"is_corresponding":true}],"reference_count":22,"raw_metadata":null,"created_at":"2026-07-19T02:59:28.666390Z","pmid":null,"pmcid":null,"fwci":null,"citation_percentile":null,"influential_citations":0,"oa_status":null,"license":null,"views":0,"total_file_size_bytes":0,"version_count":0,"fair_f":null,"fair_a":null,"fair_i":null,"fair_r":null,"fair_zscore":null,"fair_rationale":null,"fair_model":null,"fair_agent_version":null,"fair_fulltext_source":null,"fair_has_llm":null,"fair_computed_at":null,"clinical_trials":[],"software_tools":[],"db_accessions":[],"linked_datasets":[],"topics":[]}