IVCVMar 28, 2023

SVD-DIP: Overcoming the Overfitting Problem in DIP-based CT Reconstruction

arXiv:2303.15748v35 citationsh-index: 44
Originality Incremental advance
AI Analysis

This addresses overfitting in unsupervised deep learning for CT reconstruction, offering a more stable method for medical imaging applications, though it is incremental as it builds on regularized fine-tuning of DIP.

The paper tackled the overfitting problem in deep image prior (DIP) for CT reconstruction by proposing SVD-DIP, which restricts learning to singular value adaptation, resulting in significantly improved stability and overcoming noise overfitting on real-measured μCT and medical datasets.

The deep image prior (DIP) is a well-established unsupervised deep learning method for image reconstruction; yet it is far from being flawless. The DIP overfits to noise if not early stopped, or optimized via a regularized objective. We build on the regularized fine-tuning of a pretrained DIP, by adopting a novel strategy that restricts the learning to the adaptation of singular values. The proposed SVD-DIP uses ad hoc convolutional layers whose pretrained parameters are decomposed via the singular value decomposition. Optimizing the DIP then solely consists in the fine-tuning of the singular values, while keeping the left and right singular vectors fixed. We thoroughly validate the proposed method on real-measured $μ$CT data of a lotus root as well as two medical datasets (LoDoPaB and Mayo). We report significantly improved stability of the DIP optimization, by overcoming the overfitting to noise.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes