A deep learning-based ECMO injury risk prediction method and system

Abstract

A deep learning-based ECMO injury risk prediction method and system, comprising multidimensional data acquisition, data processing enhancement, multidimensional injury risk prediction, injury treatment pathway prediction, and ECMO injury risk prediction. The invention pertains to the technical field of ECMO injury risk prediction, specifically a deep learning-based ECMO injury risk prediction method and system. The invention adopts a bidirectional prediction method integrating multidimensional injury risk prediction and treatment pathway prediction. By first predicting specific injury types, risks, and trends, combined with recommended treatment modalities, pathways, and therapeutic outcome predictions, the invention enhances the direct usability and result validity of ECMO injury risk prediction. A deep bidirectional long-term neural network incorporating multimodal fusion and multi-task learning is employed for multidimensional injury risk prediction. A feature-optimized treatment pathway prediction network is utilized for injury treatment pathway prediction, constructing a comprehensive treatment pathway prediction model.