OBJECTIVES: In the field of scientific research, retrieving and interpreting data from complex graphs and plots remains a difficult and error-prone task. The goal of this study is to evaluate the ability of AI, with or without computer vision (CV), to revolutionize the interpretation of graphical representations of Kaplan-Meier curves that are broadly encountered in research publications.

METHODS: Two novel methodologies were used in this study to automate graph analysis. The first uses a Python 3.10-based web application that combines GPT-4o and Flask to provide researchers with an easy-to-use interface. This platform allows users to capture graph images from PDFs or upload them directly, leading GPT-4o (AI) to perform extensive analysis, employ a variety of statistical methods, and derive logical conclusions. The second method combines computer vision (CV) techniques for extracting accurate data points from graphs with an AI model for advanced statistical analysis.

RESULTS: The digitization of complex survival curves yielded insightful results, with key metrics analysed including, Median Survival Time, and Hazard Ratio along with confidence interval. A subject matter expert (SME) manually compared the outcomes of both methodologies with the original findings. The first approach (AI) achieved accuracy ranging from 87 to 100 percent. The second strategy (AI with CV) yielded mixed results: certain parameters were quite close to the original data, while others had discrepancies.

CONCLUSIONS: AI-powered Kaplan-Meier survival curve analysis produced encouraging results. The future research to focus on improving the output for AI with CV to increase the reliability and consistency. Generative AI has the potential to greatly improve automated graph analysis in scientific research, allowing for more precise and insightful data interpretations along with saving time and effort.