Self is the Best Learner: CT-free Ultra-Low-Dose PET Organ Segmentation via Collaborating Denoising and Segmentation Learning

2025-03-05

AI Summary

Organ segmentation in ultra-low-dose PET (5% of standard dose) without CT guidance remains challenging due to severe noise and lack of anatomical priors. Method: We propose the first end-to-end co-learning framework that eliminates CT registration dependency: modeling low-dose PET (LDPET) as a natural masked observation of full-dose PET (FDPET), employing a shared CNN/Transformer encoder with dual decoders (for denoising and segmentation), and—novelly—embedding CT-derived anatomical priors into the denoising process to enhance boundary awareness. Inspired by masked autoencoding, the framework jointly optimizes denoising and segmentation. Results: Evaluated on ¹⁸F-FDG and ⁶⁸Ga-FAPI datasets, our method achieves mean Dice scores of 73.11% and 73.97% across 18 organs, significantly outperforming CT-registration-based baselines. This establishes a new paradigm for CT-free, annotation-efficient PET organ segmentation.

Technology Category

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

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Security and Privacy: Data transparency and provenance
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Organ segmentation in Positron Emission Tomography (PET) plays a vital role in cancer quantification. Low-dose PET (LDPET) provides a safer alternative by reducing radiation exposure. However, the inherent noise and blurred boundaries make organ segmentation more challenging. Additionally, existing PET organ segmentation methods rely on co-registered Computed Tomography (CT) annotations, overlooking the problem of modality mismatch. In this study, we propose LDOS, a novel CT-free ultra-LDPET organ segmentation pipeline. Inspired by Masked Autoencoders (MAE), we reinterpret LDPET as a naturally masked version of Full-Dose PET (FDPET). LDOS adopts a simple yet effective architecture: a shared encoder extracts generalized features, while task-specific decoders independently refine outputs for denoising and segmentation. By integrating CT-derived organ annotations into the denoising process, LDOS improves anatomical boundary recognition and alleviates the PET/CT misalignments. Experiments demonstrate that LDOS achieves state-of-the-art performance with mean Dice scores of 73.11% (18F-FDG) and 73.97% (68Ga-FAPI) across 18 organs in 5% dose PET. Our code is publicly available.

Problem

• Develops CT-free ultra-low-dose PET organ segmentation.
• Addresses noise and blurred boundaries in low-dose PET.
• Reduces PET/CT misalignments and improves anatomical recognition.

Innovation

• CT-free ultra-LDPET organ segmentation pipeline
• Shared encoder with task-specific decoders
• Integrates CT annotations for improved boundary recognition