Exploiting network optimization stability for enhanced PET image denoising using deep image prior

2025-02-16

AI Summary

PET low-dose imaging is fundamentally limited by statistical noise, and existing deep learning denoising methods often over-smooth structures, compromising quantitative accuracy. To address this, we propose a stability-aware conditional Deep Image Prior (DIP) denoising framework: for the first time, it explicitly models the stability of multi-step intermediate DIP outputs during optimization, generating a spatially adaptive stability map that guides weighted linear fusion between the noisy input and the DIP reconstruction. The method requires no additional training or supervised labels and explicitly preserves fine anatomical structures and quantitative fidelity. Evaluated on FDG brain PET data across multiple dose levels, it significantly outperforms state-of-the-art methods—achieving higher peak signal-to-noise ratio (PSNR) and improved background noise suppression. Region-of-interest (ROI) analysis confirms absence of quantitative bias. Moreover, it effectively reduces background variability while preserving peak-to-valley ratios even in full-dose scans.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Learning & Optimization for CV
• Reasoning under Uncertainty: Stochastic Optimization

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Data and user privacy-enhancing technologies for the Web
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

PET is affected by statistical noise due to constraints on tracer dose and scan duration, impacting both diagnostic performance and quantitative accuracy. While deep learning (DL)-based PET denoising methods have been used to improve image quality, they may introduce over-smoothing, compromising quantitative accuracy. We propose a method for making a DL solution more reliable and apply it to the conditional deep image prior (DIP). We introduce the idea of stability information in the optimization process of conditional DIP, enabling the identification of unstable regions within the network's optimization trajectory. Our method incorporates a stability map, which is derived from multiple intermediate outputs of moderate network at different optimization steps. The final denoised image is then obtained by computing linear combination of the DIP output and the original reconstructed image, weighted by the stability map. Our method effectively reduces noise while preserving small structure details in brain FDG images. Results demonstrated that our approach outperformed existing methods in peak-to-valley ratio and noise suppression across various low-dose levels. Region-of-interest analysis confirmed that the proposed method maintains quantitative accuracy without introducing under- or over-estimation. We applied our method to full-dose PET data to assess its impact on image quality. The results revealed that the proposed method significantly reduced background noise while preserving the peak-to-valley ratio at a level comparable to that of unfiltered full-dose PET images. The proposed method introduces a robust approach to DL-based PET denoising, enhancing its reliability and preserving quantitative accuracy. This strategy has the potential to advance performance in high-sensitivity PET scanners, demonstrating that DL can extend PET imaging capabilities beyond low-dose applications.

Problem

• Enhance PET image denoising reliability
• Preserve quantitative accuracy in PET images
• Reduce noise while maintaining structural details

Innovation

• Stability map enhances PET denoising
• Linear combination preserves image details
• Conditional DIP improves quantitative accuracy