Physics-Informed autoencoder for DSC-MRI Perfusion post-processing: application to glioma grading

2025-10-13

AI Summary

Noise and motion artifacts in DSC-MRI perfusion imaging corrupt the deconvolution process, leading to biased cerebral blood flow (CBF) parameter estimation; moreover, existing deep learning methods rely on biased third-party deconvolution results as supervision. To address this, we propose a physics-guided autoencoder that embeds an analytical perfusion model—specifically, the singular value decomposition (SVD)-based residue function—into the decoder, enabling end-to-end, fully self-supervised training without external annotations. Our method directly reconstructs physiologically consistent perfusion parameters from raw time-signal curves. In glioma grading, it achieves performance comparable to state-of-the-art deconvolution algorithms (AUC ≥ 0.89), reduces computational cost by over 3×, and exhibits markedly improved robustness under high noise. The key innovation lies in the first deep integration of a differentiable biophysical model into an autoencoding architecture, thereby eliminating dependence on conventional deconvolution priors.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Computer Vision: Segmentation
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

DSC-MRI perfusion is a medical imaging technique for diagnosing and prognosing brain tumors and strokes. Its analysis relies on mathematical deconvolution, but noise or motion artifacts in a clinical environment can disrupt this process, leading to incorrect estimate of perfusion parameters. Although deep learning approaches have shown promising results, their calibration typically rely on third-party deconvolution algorithms to generate reference outputs and are bound to reproduce their limitations. To adress this problem, we propose a physics-informed autoencoder that leverages an analytical model to decode the perfusion parameters and guide the learning of the encoding network. This autoencoder is trained in a self-supervised fashion without any third-party software and its performance is evaluated on a database with glioma patients. Our method shows reliable results for glioma grading in accordance with other well-known deconvolution algorithms despite a lower computation time. It also achieved competitive performance even in the presence of high noise which is critical in a medical environment.

Problem

• Addressing noise and artifact disruption in DSC-MRI perfusion analysis
• Overcoming reliance on third-party deconvolution algorithms in deep learning
• Improving glioma grading accuracy under high-noise clinical conditions

Innovation

• Physics-informed autoencoder decodes perfusion parameters analytically
• Self-supervised training without third-party deconvolution algorithms
• Robust performance maintained under high noise conditions