Project Grant 2436216

This $200,000 Project Grant award from the National Science Foundation's Engineering program (CFDA 47.041) aims to develop the mathematical and statistical foundations for a Digital Twin (DT) system to enhance neurophysiological modeling and uncertainty quantification for individuals with Autism Spectrum Disorder (ASD). The key products and services to be delivered include:

1. Computational models based on Conditional Variational Auto-Encoders (CVAE) and longitudinal CVAE to analyze brain activities, integrate diverse imaging data, and model neurodevelopmental processes.

2. A novel bilevel formulation for multi-distribution fine-tuning techniques on pre-trained foundational models and a fast algorithm to learn from heterogeneous data sources to predict ASD outcomes.

3. A model-free conformal prediction procedure to ensemble predictions from multiple models obtained with different modalities and progression simulations, integrating various types of uncertainties.

4. A DT-based reinforcement learning framework to recommend personalized treatment/intervention plans that improve online learning efficiency and clinical outcomes.

The project aims to create a unified DT system that can enable individualized models, anticipate progression, and adjust treatment proactively, ultimately enhancing care and promoting community well-being. The funding period is from January 1, 2025 to December 31, 2027.