SafetyInsights.org

This systematic review of 28 studies explored the application of LLMs for lung cancer care and management.

Probably few surprises here. And it’s focused mostly on LLMs, rather than specialised AI models.

Extracts:

·        The review identified 7 primary application domains of LLMs in LC: auxiliary diagnosis, information extraction, question answering, scientific research, medical education, nursing support, and treatment decision-making

·        LLMs have demonstrated the ability to “automatically extract medical records, popularize general knowledge about LC, and assist clinical diagnosis and treatment”

·        LLMs can “extract clinical features by applying natural language processing methods”, enabling the diagnosis of “clinical staging, histological type, lung-RADS (Reporting and Data System) score, and metastasis sites of LC”

·        LLMs can help to “automatically generate medical records during lung nodule screening sessions” and “to generate Lung-RADS scores based on low-dose CT reports for LC screening, achieving up to 83.6% accuracy”

·        LLMs have been used for “disseminating general knowledge about LC” and assisting communication between patients and health care providers

·        Within controlled settings, LLMs can “perform preliminary analyses of medical images and textual data and… offer diagnostic and therapeutic suggestions for LC”

·        And this includes “generating more detailed chemotherapy or radiotherapy plans, and predicting outcome indicators”

·        LLMs “improve researchers’ efficiency” by classifying patients for clinical trial eligibility screening or extracting “safety and efficacy information from clinical trial abstracts”

·        Limitations exist, of course, where most current research is retrospective datasets and single-centre studies with “limited prospective, multicenter clinical validation”

·        General LLMs “may have gaps in domain-specific LC knowledge, and their outputs are prone to hallucinations and insufficient citations”

·        “LLMs cannot fully replace medical professionals,” necessitating a “human-on-the-loop” human-machine collaboration model

Shout me a coffee (one-off or monthly recurring)

Ref: Zhong R, Chen S, Li Z, Gao T, Su Y, Zhang W, Liu D, Gao L, Hu K Large Language Models in Lung Cancer: Systematic Review J Med Internet Res 2025;27:e74177

Study link: https://doi.org/10.2196/74177