Interpretation of Brain Morphology in Association to Alzheimer's Disease Dementia Classification Using Graph Convolutional Networks on Triangulated Meshes

Whole-brain radiotherapy associated with structural changes resembling aging as determined by anatomic surface-based deep learning

BACKGROUND

Brain metastases are the most common intracranial tumors in adults and are associated with significant morbidity and mortality. Whole-brain radiotherapy (WBRT) is used frequently in patients for palliation, but can result in neurocognitive deficits. While dose-dependent injury to individual areas such as the hippocampus has been demonstrated, global structural shape changes after WBRT remain to be studied.

METHODS

We studied healthy controls and patients with brain metastases and examined MRI brain anatomic surface data before and after WBRT. We implemented a validated graph convolutional neural network model to estimate patient's "brain age". We further developed a mixed-effects linear model to compare the estimated age of the whole brain and substructures before and after WBRT.

RESULTS

4220 subjects were analyzed (4148 healthy controls and 72 patients). The median radiation dose was 30 Gy (range 25-37.5 Gy). The whole brain and substructures underwent structural change resembling rapid aging in radiated patients compared to healthy controls; the whole brain "aged" 9.32 times faster, the cortex 8.05 times faster, the subcortical structures 12.57 times faster, and the hippocampus 10.14 times faster. In a subset analysis, the hippocampus "aged" 8.88 times faster in patients after conventional WBRT versus after hippocampal avoidance (HA)-WBRT.

CONCLUSIONS

Our findings suggest that WBRT causes the brain and its substructures to undergo structural changes at a pace up to 13x of the normal aging pace, where hippocampal avoidance offers focal structural protection. Correlating these structural imaging changes with neurocognitive outcomes following WBRT or HA-WBRT would benefit from future analysis.

TetCNN: Convolutional Neural Networks on Tetrahedral Meshes

Convolutional neural networks (CNN) have been broadly studied on images, videos, graphs, and triangular meshes. However, it has seldom been studied on tetrahedral meshes. Given the merits of using volumetric meshes in applications like brain image analysis, we introduce a novel interpretable graph CNN framework for the tetrahedral mesh structure. Inspired by ChebyNet, our model exploits the volumetric Laplace-Beltrami Operator (LBO) to define filters over commonly used graph Laplacian which lacks the Riemannian metric information of 3D manifolds. For pooling adaptation, we introduce new objective functions for localized minimum cuts in the Graclus algorithm based on the LBO. We employ a piece-wise constant approximation scheme that uses the clustering assignment matrix to estimate the LBO on sampled meshes after each pooling. Finally, adapting the Gradient-weighted Class Activation Mapping algorithm for tetrahedral meshes, we use the obtained heatmaps to visualize discovered regions-of-interest as biomarkers. We demonstrate the effectiveness of our model on cortical tetrahedral meshes from patients with Alzheimer's disease, as there is scientific evidence showing the correlation of cortical thickness to neurodegenerative disease progression. Our results show the superiority of our LBO-based convolution layer and adapted pooling over the conventionally used unitary cortical thickness, graph Laplacian, and point cloud representation.

Hippocampal representations for deep learning on Alzheimer's disease

Deep learning offers a powerful approach for analyzing hippocampal changes in Alzheimer's disease (AD) without relying on handcrafted features. Nevertheless, an input format needs to be selected to pass the image information to the neural network, which has wide ramifications for the analysis, but has not been evaluated yet. We compare five hippocampal representations (and their respective tailored network architectures) that span from raw images to geometric representations like meshes and point clouds. We performed a thorough evaluation for the prediction of AD diagnosis and time-to-dementia prediction with experiments on an independent test dataset. In addition, we evaluated the ease of interpretability for each representation-network pair. Our results show that choosing an appropriate representation of the hippocampus for predicting Alzheimer's disease with deep learning is crucial, since it impacts performance and ease of interpretation.