A Generalizable Deep Voxel-Guided Morphometry Algorithm for Change Detection in Multiple Sclerosis

Anish Raj, Achim Gass, Philipp Eisele, Andreas Dabringhaus, Matthias Kraemer, Frank G. Zoellner

Keywords: Voxel-guided morphometry, multiple sclerosis, change maps, convolutional neural networks

Abstract: We present a deep learning-based approach to generate Voxel-Guided Morphometry (VGM) maps from MRI scans, aiming to improve the detection and monitoring of Multiple Sclerosis (MS) progression. Leveraging a 3D U-Net architecture with attention mechanisms and optimized by histogram matching, our model excels in processing three diverse datasets. It demonstrates enhanced accuracy in identifying MS-related changes, outperforming the reference method in mean absolute error by an average of 0.4%. Additionally, visual analysis confirmed our method yields more precise and stable VGM maps across all datasets, compared to the reference. This work underscores the potential of deep learning in MS progression and treatment assessment.

Submission Number: 11