Researchers have developed a new artificial intelligence framework called FedDOSE that improves the diagnosis of autism and ADHD by analyzing brain scans from multiple hospitals without sharing sensitive patient data.
The research
A team led by Deepank Girish and Yi Hao Chan at Nanyang Technological University created FedDOSE, a federated learning approach that addresses two major challenges in brain imaging: privacy and data variability. Federated learning allows multiple institutions to collaboratively train a model without moving raw data, but standard methods struggle with differences in scanning equipment and patient populations across sites.
FedDOSE specifically tackles these "site effects" while also capturing dynamic brain connectivity—how different brain regions interact over time—rather than just static snapshots. The framework uses a Modularity-Guided Tucker Decomposition block to efficiently process high-dimensional brain connectivity data and identify modular spatial-temporal patterns. Each site generates class-specific prototypes, which are then aligned across sites using optimal transport and Procrustes analysis.
Testing on three public datasets—ABIDE-I, ABIDE-II, and ADHD-200—FedDOSE outperformed existing methods in detecting both autism spectrum disorder (ASD) and attention-deficit hyperactivity disorder (ADHD). These datasets include thousands of brain scans collected from many research centers, making them ideal for evaluating multi-site performance.
Why it matters
For individuals and families affected by ASD or ADHD, this research could lead to more accurate, earlier diagnoses. By enabling AI models to learn from larger, more diverse datasets—without sacrificing privacy—FedDOSE helps overcome a major bottleneck in medical imaging. The dynamic connectivity approach also captures more nuanced brain activity, potentially revealing biomarkers that static methods miss. While the current study focuses on diagnosis, the same framework could be adapted to other brain imaging tasks, such as predicting treatment response or tracking disease progression.
What you can do
If you’re curious about your own cognitive health, stay informed about advances in brain imaging and AI. For a quick snapshot of your cognitive abilities, consider taking a validated test. Regular mental stimulation, physical exercise, and good sleep remain the pillars of brain health, regardless of technological developments.
Source: arXiv q-bio.NC
Curious about your own brain? Take our free adaptive IQ test or try 306 brain training levels.
