Capacity of Understanding the Future Approaches in Cancer Treatment by Multiple Models of Artificial Intelligence.

Abstract

Artificial intelligence (AI) has emerged as a popular tool in education for disease treatment, not only for patients but also for physicians and scientists. We aimed to explore the educational values of different AI models in future disease treatment by providing them with real-world obstacles in cancer treatment for the most serious types of breast cancer and chondrosarcoma. We first asked seven large AI models to predict the future treatment approaches that would lead to a better outcome for triple-negative breast cancer (TNBC) and dedifferentiated chondrosarcoma (DDCS). We then requested each model to select the best one and provide supporting evidence. Next, the models were requested to provide a plan or clinical trial to test the treatment approach. Our test obtained ten treatment approaches for TNBC and DDCS from each of the seven models. Together, a total of 18 different unique approaches were suggested for TNBC and 34 for DDCS. Modified and/or extended usage of antibody-drug conjugates are predominantly selected by models as the best approach for TNBC. Combined immune checkpoint inhibition usage and isocitrate dehydrogenase (IDH) inhibitors were favored by models for DDCS. Specialized CAR-T cell therapy and clustered regularly interspaced short palindromic repeats (CRISPR)-based gene editing were selected by majority of AI models as high risk and high reward approaches. Our study indicated that most AI models are capable of keeping up with updated cancer research. However, for patients and physicians, consultation of multiple AI models may gain a better understanding of the pros and cons of a variety of approaches for cancer treatment.