ALL-PET: A Low-resource and Low-shot PET Foundation Model in Projection Domain

Author: Bin Huang, Kang Chen, Bingxuan Li, Huafeng Liu, Qiegen Liu

Building large-scale foundation model for PET imaging is hindered by limited access to labeled data and insufficient computational resources. To overcome data scarcity and efficiency limitations, we propose ALL-PET, a low-resource, low-shot PET foundation model operating directly in the projection domain. ALL-PET leverages a latent diffusion model (LDM) with three key innovations. First, we design a Radon mask augmentation strategy (RMAS) that generates over 200,000 structurally diverse training samples by projecting randomized image-domain masks into sinogram space, significantly improving generalization with minimal data. This is extended by a dynamic multi-mask (DMM) mechanism that varies mask quantity and distribution, enhancing data diversity without added model complexity. Second, we implement positive/negative mask constraints to embed strict geometric consistency, reducing parameter burden while preserving generation quality. Third, we introduce transparent medical attention (TMA), a parameter-free, geometry-driven mechanism that enhances lesion-related regions in raw projection data. Lesion-focused attention maps are derived from coarse segmentation, covering both hypermetabolic and hypometabolic areas, and projected into sinogram space for physically consistent guidance. The system supports clinician-defined ROI adjustments, ensuring flexible, interpretable, and task-adaptive emphasis aligned with PET acquisition physics. Experimental results show ALL-PET achieves high-quality sinogram generation using only 500 samples, with performance comparable to models trained on larger datasets. ALL-PET generalizes across tasks including low-dose reconstruction, attenuation correction, delayed-frame prediction, and tracer separation, operating efficiently with memory use under 24GB.

Overview of the ALL-PET foundation model and traininginference pipeline.

RMAS and DMM are used for PET sinogram synthesis while TMA is used for downstream tasks. By incorporating DDPM/DDIM/DDBM into the latent diffusion, ALL-PET can be utilized as supervised and unsupervised model. Prompt is used for generating different tracer or anatomical regions and controllable parameters can output specified size of the sinograms.

Synthetic projection-domain data generated by ALL-PET

Synthetic projection-domain data generated by ALL-PET and their corresponding reconstructed PET images, shown for multiple radiotracers e.g., 18F-FDG, 18F-DOPA, 68Ga-PSMA and different anatomical regions e.g., Brain, Trunk, Chest, Abdomen.

Task-specified paired synthetic data generated by ALL-PET

Task-specified paired synthetic data generated by ALL-PET, showing projection-domain data and their corresponding reconstructed PET images for five representative downstream tasks e.g., low-dose and normal-dose data, before and after attenuation correction, first scan and delayed scan synthesis, dual-tracer separation and tracer conversion.

Visual juxtaposition of generated outputs from each method

visual juxtaposition of generated outputs from each method, with red boxes denoting regions enlarged for detailed inspection

Generation of PET Projection Domain

# Generation of PET_brain images # Generation of PET_body images

Other Related Projects

• Raysolution_PET_Data [Data]

• Diffusion Transformer Model with Compact Prior for Low-dose PET Reconstruction [Paper] [Code]

• RED: Residual Estimation Diffusion for Low-Dose PET Sinogram Reconstruction [Paper] [Code]

• Diffusion Transformer Meets Random Masks: An Advanced PET Reconstruction Framework [Paper] [Code]

• Double-Constraint Diffusion Model with Nuclear Regularization for Ultra-low-dose PET Reconstruction [Paper] [Code]

• Temporal Image Sequence Separation in Dual-tracer Dynamic PET with an Invertible Network [Paper] [Code]

• Synthetic CT Generation via Variant Invertible Network for Brain PET Attenuation Correction [Paper] [Code]

• Spatial-Temporal Guided Diffusion Transformer Probabilistic Model for Delayed Scan PET Image Prediction [Paper] [Code]

• PET Tracer Separation using Conditional Diffusion Transformer with Multi-latent Space Learning [Paper]

• A Prior-Guided Joint Diffusion Model in Projection Domain for PET Tracer Conversion [Paper] [Code]

• Positron Emission Tomography Tracer Conversion via Variable Augmented Invertible Network [Paper]