Mapping and visualization of white matter deformation in glioma patients using deep learning

Motivation: Gliomas can induce alterations in white matter pathways, potentially resulting in structural anomalies that may contribute to functional impairments. Goal(s): We aimed to develop a robust model for mapping structural anomalies caused by gliomas. Approach: We developed a structural anomaly mapping pipeline using a convolutional autoencoder trained on randomly generated streamlines. By applying a clustering approach in the latent space, the model enables the mapping and visualization of structural anomalies induced by pathology. Results: The model effectively highlighted white matter deformations caused by gliomas, revealing variability in anomaly values based on tumor size and grade. Impact: Detecting and mapping white matter deformation is important for understanding the impact of gliomas on brain structure and its associated deficits.