U-TTT: Towards Generalizable PET Image Denoising via Test-Time Training

2026-06-09

AI Summary

This study addresses the limited generalization of existing deep learning–based PET image denoising methods under variations in dose levels or scanner types, which hinders their clinical deployment. To overcome this challenge, the authors propose U-TTT, a 3D U-Net architecture integrated with test-time training (TTT) that leverages a self-supervised mechanism to dynamically adapt model parameters during inference for sample-specific denoising. The method innovatively incorporates dual TTT pathways operating in the spatial domain (S-TTT) and frequency domain (F-TTT) to jointly correct structural degradation and recover high-frequency details. Experimental results demonstrate that U-TTT achieves state-of-the-art performance on unseen dose and scanner conditions, significantly enhancing cross-domain robustness and generalization capability.

Abstract

Existing deep learning models for Positron Emission Tomography (PET) image denoising often suffer from severe performance degradation under distribution shifts, fundamentally restricting their robust clinical deployment. This lack of generalization stems from the conventional paradigm of fixed-parameter models that cannot adapt to variations in test data (e.g., dose levels or scanner types) after training. To overcome this limitation and achieve robust generalization, we introduce U-TTT, a novel U-shaped model that integrates Test-Time Training (TTT) layers to dynamically adjust model parameters during inference through self-supervision, thereby adapting to the specific characteristics of each test instance. Furthermore, to comprehensively capture the complex degradations of 3D PET data, U-TTT features a dual-domain adaptation mechanism comprising a Spatial Test-Time Training (S-TTT) layer and a Frequency Test-Time Training (F-TTT) layer. The S-TTT layer captures and corrects spatial structural degradations, while the F-TTT layer suppresses global noise spectra and restores delicate high-frequency details. Extensive experiments demonstrate that U-TTT achieves state-of-the-art PET denoising performance and exhibits superior generalization under challenging distribution shifts, including both unseen dose levels and unseen scanners. Our code will be available at https://github.com/Yaziwel/U-TTT.

Problem

• PET image denoising
• distribution shift
• generalization
• test-time adaptation
• clinical deployment

Innovation

• Test-Time Training
• PET image denoising
• domain adaptation
• dual-domain learning
• generalizable deep learning