Abstract
PurposeWe compare the performance of three commonly used MRI‐guided attenuation correction approaches in torso PET/MRI, namely segmentation‐, atlas‐, and deep learning‐based algorithms.MethodsTwenty‐five co‐registered torso 18F‐FDG PET/CT and PET/MR images were enrolled. PET attenuation maps were generated from in‐phase Dixon MRI using a three‐tissue class segmentation‐based approach (soft‐tissue, lung, and background air), voxel‐wise weighting atlas‐based approach, and a residual convolutional neural network. The bias in standardized uptake value (SUV) was calculated for each approach considering CT‐based attenuation corrected PET images as reference. In addition to the overall performance assessment of these approaches, the primary focus of this work was on recognizing the origins of potential outliers, notably body truncation, metal‐artifacts, abnormal anatomy, and small malignant lesions in the lungs.ResultsThe deep learning approach outperformed both atlas‐ and segmentation‐based methods resulting in less than 4% SUV bias across 25 patients compared to the segmentation‐based method with up to 20% SUV bias in bony structures and the atlas‐based method with 9% bias in the lung. The deep learning‐based method exhibited superior performance. Yet, in case of sever truncation and metallic‐artifacts in the input MRI, this approach was outperformed by the atlas‐based method, exhibiting suboptimal performance in the affected regions. Conversely, for abnormal anatomies, such as a patient presenting with one lung or small malignant lesion in the lung, the deep learning algorithm exhibited promising performance compared to other methods.ConclusionThe deep learning‐based method provides promising outcome for synthetic CT generation from MRI. However, metal‐artifact and body truncation should be specifically addressed.
Highlights
Earlier approaches in this regard relied on bulk tissue segmentation,3,4 maximum likelihood reconstruction of attenuation and activity (MLAA),5,6 and atlas-based synthetic CT generation.7-10 More recently, with the revolution induced by the introduction of deep learning approaches,11 MRI-b ased sCT generation using one of the state-o f-the-art architectures of convolutional neural networks became the dominant trend or mainstream in this field.12-17 the exceptional capability of deep learning algorithms in providing satisfactory solutions to inverse problems has spurred novel frameworks for PET attenuation correction (AC) that were not feasible with other approaches
The standardized uptake value (SUV) bias estimated over the 35 volumes of interest (VOIs) drawn on the malignant lesions confirmed the overall better performance of the deep learning method, wherein a mean absolute error of less than 4% was achieved by the PET-DL approach compared to 5.6% and 10.1% by PET- Atlas and PET-Seg techniques reported in Table 3, respectively
Quantitative accuracy of the estimated CT values for major tissue classes by the different synthetic CT generation methods are presented in Supporting Information Table S3
Summary
Hybrid imaging in the form of PET/CT or PET/MRI is one of the modern and promising tools witnessed to empower a deeper understanding of the hallmarks of cancer. Since the emergence of hybrid PET/MR imaging, a countless number of approaches have been introduced to tackle the challenges of MRI-g uided attenuation correction (AC) to achieve the full quantitative potential of PET imaging. Earlier approaches in this regard relied on bulk tissue segmentation, maximum likelihood reconstruction of attenuation and activity (MLAA), and atlas-based synthetic CT (sCT) generation.7-10 More recently, with the revolution induced by the introduction of deep learning approaches, MRI-b ased sCT generation using one of the state-o f-the-art architectures of convolutional neural networks became the dominant trend or mainstream in this field.12-17 the exceptional capability of deep learning algorithms in providing satisfactory solutions to inverse problems has spurred novel frameworks for PET AC that were not feasible (or at least not providing a comparable performance) with other approaches. Since the emergence of hybrid PET/MR imaging, a countless number of approaches have been introduced to tackle the challenges of MRI-g uided attenuation correction (AC) to achieve the full quantitative potential of PET imaging.2 Earlier approaches in this regard relied on bulk tissue segmentation, maximum likelihood reconstruction of attenuation and activity (MLAA), and atlas-based synthetic CT (sCT) generation.7-10 More recently, with the revolution induced by the introduction of deep learning approaches, MRI-b ased sCT generation using one of the state-o f-the-art architectures of convolutional neural networks became the dominant trend or mainstream in this field.12-17 the exceptional capability of deep learning algorithms in providing satisfactory solutions to inverse problems has spurred novel frameworks for PET AC that were not feasible (or at least not providing a comparable performance) with other approaches. Investigation of outliers or potential sources of algorithms failure has not been sufficiently addressed
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
