Development of Synthetic Medical Data Generation Technology to Predict Postoperative Complications

Overview

\<Development of synthetic medical data generation technology to predict postoperative complications\> In order to develop a model for predicting the occurrence of complications after surgery, it is necessary to establish a cohort along with statistical indicators related to the occurrence of complications. This study aims to combine synthetic medical data based on actual clinical data and develop a predictive model based on synthetic medical data. This will allow researchers to conduct research only with synthetic data without dealing with actual medical data, allowing them to use and process data without legal constraints, and to create as much data as they want based on various preprocessed, standardized, and labeled raw data. Patients from three hospitals in Korea (Seoul National University Hospital, Seoul National University Bundang Hospital, Seoul Metropolitan City-Boramae Medical Center) were enrolled for the study. Medical data (both clinical and laboratory) from 410,000 patients who were conducted surgery between 2005 and 2020 were collected to evaluate the performance of the prediction model using AKI-based prediction model development and external verification. Based on the collected patient data, synthetic medical data were combined using the machine learning algorithm, and the anonymity and re-identification of the synthesized medical data were evaluated. Also, the development of AI-based prediction model using synthetic medical data and the actual medical data model were compared.

Conditions