Federated Linear Mixed Effects Modeling for Voxel-Based Morphometry

• Published in: 2023 IEEE 20th International Symposium on Biomedical Imaging (ISBI) pp. 1 - 4
• Main Authors: Basodi, Sunitha, Raja, Rajikha, Gazula, Harshvardhan, Romero, Javier Tomas, Panta, Sandeep, Calhoun, Vince D.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 18.04.2023
• Subjects: Analytical models; Biological system modeling; COINSTAC; Collaboration; Data privacy; Federated learning; Linear mixed-effects (LME); Magnetic resonance imaging; Memory management; Neuroimaging
• DOI: 10.1109/ISBI53787.2023.10230684

We propose a federated linear mixed-effects (LME) model to perform a large-scale analysis of data gathered from different collaborations without the need to pool or share the actual data. This method is efficient as it overcomes the hurdles of data privacy for sharing and has lower bandwidth and memory requirements for analysis than the centralized modeling approach. Our decentralized LME approach is optimized for neuroimaging data. We evaluate our model using features extracted from structural magnetic resonance imaging (sMRI) data. Results demonstrate that federated (decentralized) LME models achieve similar performance compared to the models trained with all the data in one location.