Survival Probability Extraction and Performance Comparison of Kaplan-Meier Curves

Mu Lifeng; Mao Longying; Mao Yun; Chen Xin; Chen Long; Yang Ming

Abstract：Objective:To plot Kaplan-Meier curves using simulated survival data and compare the characteristics of different survival probability extraction methods for Kaplan-Meier curves and their performance across various scenarios.Methods:Survival datasets were simulated using R-4.4.2 with parameters including sample sizes,censoring marker,and curve numbers.GetData Graph Digitizer,IPDfromKM getpoints,SurvdigitizeR survival_digitize were evaluated.A JavaScript script was developed to extract Kaplan-Meier curve.Root Mean Square Error(RMSE)was calculated to quantify deviations between digitized and true survival probabilities.Results:The JavaScript script method demonstrated the smallest RMSE across all simulated scenarios(RMSE=1.015×10-4),significantly outperforming the other three methods(P＜0.05),with statistically significant differences observed among methods(P＜0.05).Conclusion:For vector format illustrations,JavaScript scripts enable accurate and robust reverse engineering of Kaplan-Meier curves;for bitmaps,the GetData Graph Digitizer and SurvdigitizeR survival_digitize methods yield more accurate results,and the SurvdigitizeR survival_digitize method is the most efficient.Future research should focus on integrating intelligent algorithms for enhanced robustness and precision in survival data reconstruction.