Disentangled Learning Improves Implicit Neural Representations for Medical Reconstruction

TL;DR

DisINR is a novel framework for medical image reconstruction using implicit neural representations that disentangles shared and subject-specific features, enabling efficient training and improved image quality.

Contribution

DisINR introduces a disentangled INR architecture with shared and subject-specific modules, pre-trained from raw data, enhancing reconstruction accuracy and efficiency in medical imaging.

Findings

DisINR outperforms existing INRs in reconstruction accuracy.

DisINR reduces training time by optimizing only subject-specific encoders during testing.

DisINR effectively preserves learned priors during test-time adaptation.

Abstract

Implicit neural representations (INRs) have emerged as a powerful paradigm for medical imaging via physics-informed unsupervised learning. Classical INRs optimize an entire network from scratch for each subject, leading to inefficient training and suboptimal imaging quality. Recent initialization-based approaches attempt to inject population priors into pre-trained networks, yet they rely on high-quality images and often suffer from catastrophic forgetting during fine-tuning. We present DisINR, a novel INR framework that explicitly disentangles shared and subject-specific representations. DisINR introduces a shared encoder-decoder pair and subject-specific encoders, whose features are jointly decoded for image reconstruction. By integrating differentiable forward models, it pre-trains the shared modules directly from limited raw measurements, removing the need for pre-acquired high-quality images. During test-time adaptation, only the subject-specific encoder is optimized, while the shared pair remains frozen, effectively preserving learned priors. Extensive evaluations on three representative medical imaging tasks show that DisINR significantly outperforms state-of-the-art INRs in both reconstruction accuracy and efficiency.