Affordable and real-time antimicrobial resistance prediction from multimodal electronic health records

Antimicrobial resistance (AMR) is a growing concern, causing significant health risks and loss of life. Machine learning techniques can help predict AMR by utilizing data-driven models. This study uses deep learning techniques and multimodal data from electronic health records (EHR) to predict AMR. The MIMIC-IV database is used to generate structured input sources for AMR tasks, and a multimodality fusion approach is used to determine resistance based on antibiotics or pathogens. This approach lays the groundwork for deploying multimodal DL techniques in clinical practice.