Ponente
Descripción
Cardiac Positron Emission Tomography (PET) is a powerful molecular imaging technique, but its use is severely limited by poor spatial resolution. This limitation is particularly critical in preclinical studies with rodents, where small anatomical structures, high respiratory and heart rates exacerbate image blurring, partial volume effects, and quantitative errors [1]. These degradations arise from fundamental physical factors—positron range (PR), finite detector size, acollinearity, photon scatter—and are further amplified by physiological cardiac and respiratory motion [2]. To address this, we developed a novel deep learning-based super-resolution framework that integrates high-resolution ultrafast ultrasound (UUS) images as a priori anatomical and motion information to guide cardiac PET resolution recovery. Unlike previous approaches relying only on low-resolution PET data, this method leverages UUS images acquired simultaneously and co-registered with PET on the hybrid system PETRUS (PET/CT combined with UUS) [3], capturing fine structural boundaries and cardiac motion patterns.
Realistic PET data were generated using 50 digital mouse phantoms with the MOBY numerical phantom [4] to model diverse anatomies, cardiac and respiratory cycles. Physical and instrumental degradation factors -such as PR effects (simulated with PENEASY [5]), point spread function (PSF) blurring, and statistical noise - were applied to FDG activity maps, using experimentally measured parameters from the PETRUS scanner [6]. In parallel, corresponding UUS images were simulated for the same anatomical models using MUST simulations [7]. Two U-Net [8] convolutional neural networks were trained: one using only PET images, and another combining PET with co-registered UUS guidance.
The UUS-guided model clearly outperformed the PET-only model, achieving superior recovery of fine myocardial structures, higher SUV accuracy, and enhanced contrast in high-uptake regions (Fig.1). When applied to experimental datasets acquired with PETRUS, it produced sharper images with reduced noise and partial volume effects. These results show that incorporating anatomical and motion priors from UUS can overcome intrinsic resolution limits of cardiac PET, enabling more reliable quantification in cardiovascular diseases.
[1] J.J. Vaquero, P. Kinahan, Positron Emission Tomography: Current Challenges and Opportunities for Technological Advances in Clinical and Preclinical Imaging Systems, Annu. Rev. Biomed. Eng. 17, 385 (2015).
[2] M. Perez-Liva et al., Ultrafast Ultrasound Imaging for Super-Resolution Preclinical Cardiac PET, Mol. Imaging Biol. 22, 1342 (2020).
[3] J. Provost et al., Simultaneous positron emission tomography and ultrafast ultrasound for hybrid molecular, anatomical and functional imaging, Nat. Biomed. Eng. 2, 85–94 (2018).
[4] W.P. Segars et al., Development of a 4-D digital mouse phantom for molecular imaging research. Mol. Imaging Biol. 6, 149–159 (2004).
[5] J. Sempau et al., A PENELOPE‐based system for the automated Monte Carlo simulation of clinacs and voxelized geometries—application to far‐from‐axis fields. Med. Phys., 38(11), 5887-5895 (2011).
[6 ] M. Perez-Liva et al., Performance evaluation of the PET component of a hybrid PET/CT-ultrafast ultrasound imaging instrument, Phys. Med. Biol. 63, 19NT01 (2018).
[7] D. Garcia, Make the Most of MUST: An Open-Source Matlab Ultrasound Toolbox, IEEE IUS (2021).
[8] G. Du et al., Medical Image Segmentation Based on U-Net: A Review, J. Imaging Sci. Technol. 64(2) (2020).
