Convolutional Neural Network and Bidirectional Long Short-Term Memory for Personalized Treatment Analysis Using Electronic Health Records

Correct precision techniques have far not been introduced for modeling the modality risk in Intensive Care Unit (ICU) patients. Traditional mortality risk prediction techniques effectively extract the data in longitudinal Electronic Health Records (EHRs), that ignore the difficult relationship and interactions among variables and time dependency in longitudinal records. The proposed work, developed the Convolutional Neural Network – Bidirectional-Long Short-Term Memory (CNN-Bi-LSTM) method for personalized treatment analysis using EHR data. The CNN extracts the significant features from relevant features, focused on spatial-based relationships. Then, the Bi-LSTM layer captured the sequential dependencies and temporal relationships in patient histories that are essential to understand the treatment results. The Circle Levy flight – Ladybug Beetle Optimization (CL-LBO) integrates the circle chaotic map and Levy flight process in traditional LBO to select relevant features for classification. The proposed method reached 99.85% accuracy, 99.60% precision, 99.50% recall, 99.55% f1-score, and 99.95% Area Under Curve (AUC) when compared to LSTM.

@article{Yavanamandha2025,
  title     = {Convolutional Neural Network and Bidirectional Long Short-Term Memory for Personalized Treatment Analysis Using Electronic Health Records},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {16},
  number    = {1},
  year      = {2025},
  publisher = {The Science and Information Organization},
  author    = {Prasanthi Yavanamandha and D. S. Rao},
  doi       = {10.14569/IJACSA.2025.0160136},
  url       = {https://doi.org/10.14569/IJACSA.2025.0160136}
}