Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Video

Updated: Jun 30, 2026

Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System

Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System

Published on: July 11, 2025

Challenges in AI Based Tumor Board Case Summarization and Recommendations.

Wen-Wai Yim, Hendrik Damm, Tabea M G Pakull

Research Square

|June 29, 2026

Summary

This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

Extracting Medical Information From Unstructured Clinical Text Using Large Language Models to Enhance Health Care Interoperability: Proof-of-Concept Study.

Journal of medical Internet research·2026

Same author

Recent Advances in Automated Mitosis Detection in Digital Pathology: A PRISMA-Guided Systematic Review with Evaluation-Regime Stratification (2018-2025).

Biomedicines·2026

Same author

Cardiac dysfunction during immune checkpoint inhibitor therapy: association with extracardiac immune-related adverse events.

Cardio-oncology (London, England)·2026

Same author

Can one model fit all? Evaluating foundation models for time series forecasting across clinical medicine.

Artificial intelligence in medicine·2026

Same author

Refining the treatment of immune checkpoint inhibitor-associated myocarditis: Challenges, innovations and rechallenge considerations.

British journal of pharmacology·2026

Same author

Multimodal phenotypic classification of generalized anxiety and panic using structural MRI data and psychosocial factors: machine learning results from the German National Cohort (NAKO) study.

Translational psychiatry·2026

Same journal

Somatosensory cortex shapes perceptual decision bias via the superior colliculus.

Research square·2026

Same journal

Combinatorial Targeting of Avapritinib-Driven MAP Kinase Activation in High-Grade Glioma.

Research square·2026

Same journal

Supporting Implementation of the National Standards for Cancer Survivorship Care: Development of the Cancer Survivorship Maturity Model (CSMM).

Research square·2026

Same journal

Operationalizing a walking exercise prescription based on 6-minute walk test results.

Research square·2026

Same journal

Age but not sex modifies lymphoid immune responses in murine sepsis.

Research square·2026

Same journal

Indirect effect, through aspects of neighborhood affluence and racial/ethnic composition, of receiving a Section 8 voucher on the prevalence of psychiatric disorders among boys and girls in the Moving to Opportunity study.

Research square·2026

See all related articles

Large language models (LLMs) show promise in summarizing tumor board cases but struggle with generating treatment recommendations. Expert evaluations highlight limitations in current AI for oncology decision support.

Area of Science:

Artificial Intelligence in Medicine
Oncology Decision Support
Clinical Informatics

Background:

Tumor boards are multidisciplinary meetings crucial for cancer patient care planning.
Evaluating the efficacy of AI, specifically large language models (LLMs), in supporting tumor board tasks is essential.

Purpose of the Study:

To formally define tumor board case summarization, options generation, and outcome prediction as AI tasks.
To assess the performance of state-of-the-art LLMs across multiple medical institutions for these tasks.
To introduce and validate novel metrics for evaluating AI performance in clinical settings.

Main Methods:

Utilized datasets from over four medical institutions.
Evaluated LLM performance (DeepSeek, GPT, Gemini, Qwen families) using human expert judgments (~13k assessments).

Related Experiment Videos

Last Updated: Jun 30, 2026

Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System

Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System

Published on: July 11, 2025

Developed and compared TBFact metrics against LLM-as-Judge metrics for free-text evaluation.

Main Results:

Clinician ratings for LLM-generated case summaries ranged from 3.57-4.59/5.0.
LLMs struggled with generating treatment recommendations, averaging 2.0-3.6/5.0.
Modest correlations (0.2) for summarization and higher (0.6) for recommendations were observed with LLM-as-Judge metrics.

Conclusions:

LLMs demonstrate capability in case summarization but require significant improvement for recommendation generation.
Novel metrics like TBFact show promise for explainable AI evaluation in clinical contexts.
Expert evaluations revealed complex relationships between granular scores and overall assessments, challenging traditional evaluation assumptions.