Skip to main content
[Preprint]. 2024 Mar 15:2024.03.14.24304230. [Version 1] doi: 10.1101/2024.03.14.24304230

Figure 2.

Figure 2.

Schematic Representation of the Proposed Multimodal Model: The diagram illustrates our proposed framework, wherein we concatenate embeddings derived from 6 hours’ time-series measurements using LSTM and clinical notes processed by BioMedBERT. This combined representation is utilized for the prediction of AKI/CRRT, aiming to forecast clinical manifestation at least 12 hours in advance.