Reimagining Cancer Care With Generative Artificial Intelligence: The Promise of Large Language Models.

Area of Science:

• Artificial Intelligence in Medicine
• Oncology Informatics

Background:

• Large language models (LLMs) demonstrate advanced natural language processing capabilities, opening new avenues in healthcare.
• Cancer patient journeys generate rich, longitudinal data, ideal for LLM analysis.

Purpose of the Study:

• To review fundamental concepts of LLMs, including architecture, training, and optimization.
• To explore the application of LLMs in oncology for patients, clinicians, researchers, and learners.
• To address the limitations and risks associated with LLM implementation in cancer care.

Main Methods:

• Review of LLM fundamentals: definitions, architecture, training, and optimization techniques (prompt engineering, retrieval-augmented generation).
• Exploration of multiagentic frameworks for LLMs.
• Synthesis of current research on LLM benefits and risks in oncology.
• Identification of future research directions.

Main Results:

• LLMs can generalize across diverse NLP tasks, with significant potential in oncology due to extensive patient data.
• Potential benefits span patients, oncologists, researchers, and learners, enhancing various aspects of cancer care.
• Identified risks include model hallucinations, data bias, patient privacy concerns, and potential clinician deskilling.

Conclusions:

• LLMs show substantial promise for transforming oncology, but careful consideration of risks is crucial.
• Future research must prioritize patient perspectives, clinical benchmarks, real-world validation, and deeper understanding of LLM reasoning.