Learning Brain Tumor Representation in 3D High-Resolution MR Images via Interpretable State Space Models

TL;DR

This paper introduces an interpretable state-space-model-based autoencoder for 3D high-resolution MR images, improving representation learning, interpretability, and accuracy in neuro-oncology tasks.

Contribution

It presents a novel SSM-based masked autoencoder that scales ViT-like models to 3D MR images and enables direct visualization of latent features in the input space.

Findings

Achieved state-of-the-art accuracy in neuro-oncology classification tasks.

Enhanced interpretability of learned representations through latent-to-spatial mapping.

Demonstrated effective handling of high-resolution 3D MR data with improved efficiency.

Abstract

Learning meaningful and interpretable representations from high-dimensional volumetric magnetic resonance (MR) images is essential for advancing personalized medicine. While Vision Transformers (ViTs) have shown promise in handling image data, their application to 3D multi-contrast MR images faces challenges due to computational complexity and interpretability. To address this, we propose a novel state-space-model (SSM)-based masked autoencoder which scales ViT-like models to handle high-resolution data effectively while also enhancing the interpretability of learned representations. We propose a latent-to-spatial mapping technique that enables direct visualization of how latent features correspond to specific regions in the input volumes in the context of SSM. We validate our method on two key neuro-oncology tasks: identification of isocitrate dehydrogenase mutation status and 1p/19q co-deletion classification, achieving state-of-the-art accuracy. Our results highlight the potential of SSM-based self-supervised learning to transform radiomics analysis by combining efficiency and interpretability.