Feasibility of New Method for the Prediction of Clinical Trial Results Using Compressive Sensing of Artificial Intelligence

Abstract

We propose a novel strategy to predict unknown data from a smaller sample size by using compressive sensing which is one of the information technology used in artificial intelligence. In this study, we applied the compressive sensing to an empirical sample of clinical trials comprising of two groups. Each group required the time data of 25 patients but the data of 12 more patients had to be acquired to satisfy alpha and beta errors. The unknown data were thus predicted using compressive sensing method. The calculation was repeated 1,000 times for each, and 1 million (=1,000 2) log-rank tests between the two groups were carried out following, which a histogram of the P-values was obtained. The information on the distribution of predicted data obtained by repeated calculations reflects the probability of statistical significance. The method could thus increase efficiency by saving on human resources, financial resources, and time. As such, further studies examining the feasibility of compressive sensing in clinical trials are warranted.