The article discusses the importance of joint analysis of structured and unstructured data in Electronic Health Records (EHR). It introduces a novel multimodal feature embedding generative model and contrastive loss for improved EHR feature representation. The theoretical analysis demonstrates the effectiveness of multimodal learning compared to single-modality approaches, with a focus on privacy-preserving algorithms. Simulation studies validate the proposed algorithm's performance under various configurations and its clinical utility in real-world EHR data.
A otro idioma
del contenido fuente
arxiv.org
Ideas clave extraídas de
by Tianxi Cai,F... a las arxiv.org 03-25-2024
https://arxiv.org/pdf/2403.14926.pdfConsultas más profundas