oncology

TRACE 提出了一种可部署的树-关系结构增强框架，用于肿瘤学大语言模型：将昂贵的离线结构学习与轻量在线推理分离，通过 LM 损失推导的证据来精炼医学概念树结构，并在推理时作为紧凑提示证据检索，在十项肿瘤学分类任务和 MedQuAD CancerGov QA 上超越了标准 RAG 与 GraphRAG。

This paper proposes Clinical Guardrail Probes (CG-Probes) to measure risks in patient-facing AI assistants for oncology by recovering guardrail directions from query embeddings using linear probing methods.

This research paper investigates how creatine uptake enhances antitumor immune responses, based on scientific findings in immunology and oncology.

MIST introduces a novel multimodal approach for survival prediction in oncology by using genomic-guided histology attention, showing improved performance across multiple external datasets.

The paper evaluates a configurable multi-agent system (nMAS) for extracting structured oncology data from fragmented clinical documents, achieving high performance compared to a baseline model.

This paper introduces the Oncology Decision Boundary Benchmark and evaluates nine frontier large language models, finding systematic blind spots in guideline-conformant oncology decision-making, indicating that single models cannot serve as sole bases for clinical decisions.

MultiSigBERT is a novel multimodal framework for survival analysis in oncology that uses path signature transforms to integrate electronic health record data for improved risk prediction.

Virologist Beata Halassy self-treated her recurrent breast cancer with lab-prepared viruses, achieving remission and redirecting her research to oncolytic virotherapy.

DoGMA is a central-dogma-guided foundation model for pan-cancer multi-omics analysis, using a Transformer-MoE architecture with directed attention to align DNA-RNA-protein flows and pretraining via masked hierarchical omics reconstruction. It shows strong performance across cancer representation learning, survival prediction, and metastasis prediction tasks.

Introduces OncoTriad-QA, a patient-level benchmark integrating radiology, pathology, genomics, and clinical data for pan-cancer reasoning, along with OncoVLM, a reference multimodal model that outperforms existing medical LLMs after fine-tuning.

Microsoft Research and the Broad Institute, with support from Dana-Farber Cancer Institute, are collaborating on Project Ex Vivo to use AI for better understanding cancer cell states and advancing precision oncology.

Evaluates agentic LLM systems for generating breast cancer treatment recommendations using 72 clinical cases, finding that the best system (Claude Opus 4.8 with D&C+SA pipeline) achieved a global score of 0.594 but remains insufficient for unsupervised clinical use due to persistent errors.

SAGEAgent is an LLM-based clinical agent that sequentially decides which diagnostic modalities to acquire for cancer patients to balance predictive accuracy with clinical invasiveness, reducing acquisition burden by 55% while maintaining competitive survival prediction performance.

This paper introduces the Large Cancer Assistant (LCA), a model-agnostic orchestration framework for scalable clinical decision support in oncology that decouples multimodal data ingestion from AI inference using a 7-tuple architecture and Algorithmic Impermeability.

Big progress against two of the most deadly cancers reported from randomized trials, as highlighted by Eric Topol during the ASCO conference.

Daraxonrasib received a standing ovation at ASCO as Revolution Medicines' breakthrough against pancreatic cancer, celebrated by over 40,000 oncologists, entrepreneurs, investors, and patient advocates.

This paper presents a deployment-oriented stress-testing framework to evaluate how well large language models identify side effects of breast cancer radiation treatments. The study highlights limitations in LLM reliability, such as sensitivity to minor documentation changes and under-recall of rare side effects, suggesting that grounding outputs in clinician-curated lists improves robustness.

A new study proposes viewing tumors as organized ecosystems rather than random mutations, using AI to analyze spatial organization, immune localization, and signaling environments in oncology.

FD-NL2SQL is a feedback-driven natural language to SQL system for clinical oncology databases that improves with use through clinician edits and logic-based SQL augmentation. The system decomposes natural language questions into predicates, retrieves expert-verified exemplars, and synthesizes executable SQL with continuous learning capabilities.