Large Language Models Demonstrate Potential in Predicting Liver Cancer Treatment Outcomes

Recent advancements in artificial intelligence have highlighted the promising role of large language models (LLMs) in oncology, particularly in predicting how patients with liver cancer respond to various treatments. A study spearheaded by Prof. Li Hai at the Hefei Institutes of Physical Science, Chinese Academy of Sciences, has systematically examined the efficacy of LLMs like GPT-4, GPT-4o, Google Gemini, and DeepSeek in forecasting treatment outcomes for hepatocellular carcinoma (HCC), one of the most prevalent and lethal cancers worldwide.

HCC patients often face limited effective treatment options, especially in advanced stages, where combination therapies involving immune checkpoint inhibitors and targeted treatments are used. However, only about 30% of patients respond successfully, underscoring the crucial need for better predictive tools to tailor therapies effectively. The study utilized clinical and imaging data from 186 patients with inoperable HCC and evaluated the models' prediction capabilities through zero-shot learning, meaning the models were not customized with prior liver cancer data.

To improve accuracy, researchers implemented various decision strategies, including voting rules and logical combinations, culminating in a hybrid model called Gemini-GPT. Notably, this model matched the predictive performance of experienced senior doctors and outperformed less experienced clinicians in accuracy and speed. Gemini-GPT consistently produced reliable predictions across different treatment modalities and disease stages, emphasizing its potential as a decision-support tool.

Further refinement of the model with logical strategies enhanced its utility in real-world clinical settings, making it a promising asset for personalized medicine in oncology. According to Prof. Li Hai, this research signifies a significant step toward integrating trustworthy AI systems in cancer care, providing clinicians with a powerful tool to improve treatment planning and patient outcomes.

The findings, published in the Journal of Medical Systems, point toward a future where AI and human expertise work hand-in-hand to deliver tailored therapies, ultimately improving prognosis and quality of life for liver cancer patients. This study demonstrates that with continued development, LLMs can not only understand language but also reason, predict, and support critical medical decisions in complex clinical scenarios.

Source: Medical Xpress