CognitiveTwin Framework Predicts Alzheimer's Cognitive Decline
Researchers have developed a new digital twin framework called CognitiveTwin, aimed at predicting cognitive pathways in Alzheimer's disease. This cutting-edge model combines various types of long-term data, including cognitive evaluations, MRI scans, PET scans, cerebrospinal fluid markers, and genetic data. It uses a Transformer-based design to integrate these different data types and applies a Deep Markov Model to track changes over time. The framework was trained and validated using information from 1,666 patients in the TADPOLE dataset, which is part of the Alzheimer's Disease Neuroimaging Initiative. The study focused on assessing prediction accuracy, ensuring fairness across demographics, and maintaining resilience against incomplete data, all in an effort to understand the complexities of disease progression.
Key facts
- CognitiveTwin is a digital twin framework for predicting cognitive decline in Alzheimer's disease.
- It integrates multi-modal data: cognitive scores, MRI, PET, CSF biomarkers, and genetics.
- Uses Transformer-based architecture and Deep Markov Model.
- Trained and evaluated on 1,666 patients from the TADPOLE dataset.
- Assessed for prediction error, demographic fairness, and robustness to missing data.
- Addresses heterogeneity of Alzheimer's disease progression.
- Aims to provide reliable clinical tools with high accuracy and fairness.
- Published on arXiv with ID 2604.22428.
Entities
Institutions
- Alzheimer's Disease Neuroimaging Initiative
- TADPOLE