Development and Validation of a Deep Learning-Enhanced Prediction Model for the Likelihood of Pulmonary Embolism

Abstract

Introduction Pulmonary embolism (PE) is a common and potentially fatal disease, and timely and accurate assessment of the risk of PE occurrence in patients with Deep Vein Thrombosis (DVT) is crucial. This study aims to develop a precise and efficient deep learning-based PE risk prediction model, PE-Mind. Materials and Methods We first preprocessed and reduced the high-dimensional clinical features collected from patients. The 37 most important clinical features were grouped, sorted, and connected to capture potential associations between them. The proposed model utilizes a convolutional approach, including three custom-designed residual modules. To validate the model's superiority, we also compared it with five mainstream models. Results The results show that PE-Mind demonstrated the highest accuracy and reliability, achieving an accuracy of 0.7826 and an area under the receiver operating characteristic curve of 0.8641 on the prospective test set, outperforming other models. Based on this, we have also developed a Web server, PulmoRiskAI, for real-time physician operation. Conclusions The proposed method has the potential to become a practical clinical tool, providing doctors with more accurate PE risk assessments and timely identification of high-risk patients.