Related Experiment Video
Updated: Jun 24, 2025
Whole-brain Segmentation and Change-point Analysis of Anatomical Brain MRI—Application in Premanifest Huntington's Disease
Published on: June 9, 2018
Multimodal Predictive Modeling: Scalable Imaging Informed Approaches to Predict Future Brain Health
Meenu Ajith1, Jeffrey S Spence2, Sandra B Chapman2
1Tri-Institutional Center for Translational Research in Neuroimaging and Data Science (TReNDS), Georgia State University, Georgia Institute of Technology, and Emory University, 55 Park Pl NE, Atlanta, 30303, GA, USA.
Predicting future brain health is enhanced by integrating neuroimaging data. An image-assisted approach using a partially conditional variational autoencoder (PCVAE) shows superior prediction of future brain health compared to assessment-only or neuroimaging-only methods.
Area of Science:
- Neuroscience
- Computational Psychiatry
- Medical Imaging Analysis
Background:
- Predicting future brain health is challenging, requiring integration of diverse data.
- Neural patterns from neuroimaging offer early indicators of behavioral and psychological states.
Purpose of the Study:
- To introduce and evaluate a multimodal predictive modeling approach for future brain health.
- To compare an imaging-informed methodology against traditional assessment-only and neuroimaging-only methods.
Main Methods:
- Developed an image-assisted approach using a partially conditional variational autoencoder (PCVAE).
- Integrated static functional network connectivity (sFNC) from resting-state functional magnetic resonance imaging (rs-fMRI) with behavioral assessments.
- Evaluated against support vector regression (SVR) and random forest (RF) models.
Main Results:
- The image-assisted method demonstrated superior performance in predicting future brain health constructs and their longitudinal changes.
- The PCVAE model effectively utilized neuroimaging data during training to enhance predictions from assessment data alone.
- Outperformed assessment-only and neuroimaging-only approaches in predictive accuracy.
Conclusions:
- Neuroimaging-informed predictive modeling holds significant potential for understanding brain health.
- The proposed multimodal approach advances the prediction of cognitive performance and neural connectivity relationships.
- This study highlights the value of integrating neuroimaging with behavioral data for robust brain health predictions.
