Neural Pharmacodynamic State Space Modeling
This work addresses the challenge of accurate and interpretable modeling of patient biomarkers over time for clinical applications, representing an incremental advancement in neural network methods for healthcare.
The paper tackled the problem of overfitting in neural network models for predicting patient disease progression from high-dimensional longitudinal data, and proposed a deep generative model with a novel attention-based architecture inspired by treatment physics, resulting in significant improvements in generalization and interpretable insights into cancer progression dynamics on real-world clinical data.
Modeling the time-series of high-dimensional, longitudinal data is important for predicting patient disease progression. However, existing neural network based approaches that learn representations of patient state, while very flexible, are susceptible to overfitting. We propose a deep generative model that makes use of a novel attention-based neural architecture inspired by the physics of how treatments affect disease state. The result is a scalable and accurate model of high-dimensional patient biomarkers as they vary over time. Our proposed model yields significant improvements in generalization and, on real-world clinical data, provides interpretable insights into the dynamics of cancer progression.