Contrastive Self-supervised Learning for Neurodegenerative Disorder Classification

Vadym Gryshchuk ¹ Devesh Singh ^1* Stefan Teipel ^1,2 Martin Dyrba ^1*

• ¹ German Center for Neurodegenerative Diseases (DZNE), Rostock/Greifswald, Germany
• ² Department of Psychosomatic Medicine, Rostock University Medical Center, Rostock, Mecklenburg-Vorpommern, Germany

Neurodegenerative diseases such as Alzheimer's disease (AD) or frontotemporal lobar degeneration (FTLD) involve specific loss of brain volume, detectable in vivo using T1-weighted MRI scans. Supervised machine learning approaches classifying neurodegenerative diseases require diagnostic-labels for each sample. However, it can be difficult to obtain expert labels for a large amount of data. Self-supervised learning (SSL) offers an alternative for training machine learning models without data-labels. Methods:We investigated if the SSL models can be applied to distinguish between different neurodegenerative disorders in an interpretable manner. Our method comprises a feature extractor and a downstream classification head. A deep convolutional neural network, trained with a contrastive loss, serves as the feature extractor that learns latent representations. The classification head is a single-layer perceptron that is trained to perform diagnostic group separation. We used N=2694 T1-weighted MRI scans from four data cohorts: two ADNI datasets, AIBL and FTLDNI, including cognitively normal controls (CN), cases with prodromal and clinical AD, as well as FTLD cases differentiated into its phenotypes. Results:Our results showed that the feature extractor trained in a self-supervised way provides generalizable and robust representations for the downstream classification. For AD vs. CN, our model achieves 82% balanced accuracy on the test subset and 80% on an independent holdout dataset. Similarly, the Behavioral variant of frontotemporal dementia (BV) vs. CN model attains an 88% balanced accuracy on the test subset. The average feature attribution heatmaps obtained by the Integrated Gradient method highlighted hallmark regions, i.e., temporal gray matter atrophy for AD, and insular atrophy for BV. Conclusion: Our models perform comparably to state-of-the-art supervised deep learning approaches. This suggests that the SSL methodology can successfully make use of unannotated neuroimaging datasets as training data while remaining robust and interpretable.