1
|
Olin AB, Hansen AE, Rasmussen JH, Jakoby B, Berthelsen AK, Ladefoged CN, Kjær A, Fischer BM, Andersen FL. Deep learning for Dixon MRI-based attenuation correction in PET/MRI of head and neck cancer patients. EJNMMI Phys 2022; 9:20. [PMID: 35294629 PMCID: PMC8927520 DOI: 10.1186/s40658-022-00449-z] [Citation(s) in RCA: 4] [Impact Index Per Article: 2.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 10/27/2021] [Accepted: 03/02/2022] [Indexed: 11/10/2022] Open
Abstract
Background Quantitative whole-body PET/MRI relies on accurate patient-specific MRI-based attenuation correction (AC) of PET, which is a non-trivial challenge, especially for the anatomically complex head and neck region. We used a deep learning model developed for dose planning in radiation oncology to derive MRI-based attenuation maps of head and neck cancer patients and evaluated its performance on PET AC. Methods Eleven head and neck cancer patients, referred for radiotherapy, underwent CT followed by PET/MRI with acquisition of Dixon MRI. Both scans were performed in radiotherapy position. PET AC was performed with three different patient-specific attenuation maps derived from: (1) Dixon MRI using a deep learning network (PETDeep). (2) Dixon MRI using the vendor-provided atlas-based method (PETAtlas). (3) CT, serving as reference (PETCT). We analyzed the effect of the MRI-based AC methods on PET quantification by assessing the average voxelwise error within the entire body, and the error as a function of distance to bone/air. The error in mean uptake within anatomical regions of interest and the tumor was also assessed. Results The average (± standard deviation) PET voxel error was 0.0 ± 11.4% for PETDeep and −1.3 ± 21.8% for PETAtlas. The error in mean PET uptake in bone/air was much lower for PETDeep (−4%/12%) than for PETAtlas (−15%/84%) and PETDeep also demonstrated a more rapidly decreasing error with distance to bone/air affecting only the immediate surroundings (less than 1 cm). The regions with the largest error in mean uptake were those containing bone (mandible) and air (larynx) for both methods, and the error in tumor mean uptake was −0.6 ± 2.0% for PETDeep and −3.5 ± 4.6% for PETAtlas. Conclusion The deep learning network for deriving MRI-based attenuation maps of head and neck cancer patients demonstrated accurate AC and exceeded the performance of the vendor-provided atlas-based method both overall, on a lesion-level, and in vicinity of challenging regions such as bone and air.
Collapse
Affiliation(s)
- Anders B Olin
- Department of Clinical Physiology, Nuclear Medicine and PET & Cluster for Molecular Imaging, Rigshospitalet, University of Copenhagen, Blegdamsvej 9, 2100, Copenhagen, Denmark.
| | - Adam E Hansen
- Department of Clinical Physiology, Nuclear Medicine and PET & Cluster for Molecular Imaging, Rigshospitalet, University of Copenhagen, Blegdamsvej 9, 2100, Copenhagen, Denmark.,Faculty of Health and Medical Science, University of Copenhagen, Copenhagen, Denmark.,Department of Radiology, Rigshospitalet, University of Copenhagen, Copenhagen, Denmark
| | - Jacob H Rasmussen
- Department of Otorhinolaryngology, Head and Neck Surgery and Audiology, Rigshospitalet, University of Copenhagen, Copenhagen, Denmark
| | - Björn Jakoby
- Siemens Healthcare GmbH, Erlangen, Germany.,University of Surrey, Guildford, Surrey, UK
| | - Anne K Berthelsen
- Department of Oncology, Section of Radiotherapy, Rigshospitalet, University of Copenhagen, Copenhagen, Denmark
| | - Claes N Ladefoged
- Department of Clinical Physiology, Nuclear Medicine and PET & Cluster for Molecular Imaging, Rigshospitalet, University of Copenhagen, Blegdamsvej 9, 2100, Copenhagen, Denmark
| | - Andreas Kjær
- Department of Clinical Physiology, Nuclear Medicine and PET & Cluster for Molecular Imaging, Rigshospitalet, University of Copenhagen, Blegdamsvej 9, 2100, Copenhagen, Denmark
| | - Barbara M Fischer
- Department of Clinical Physiology, Nuclear Medicine and PET & Cluster for Molecular Imaging, Rigshospitalet, University of Copenhagen, Blegdamsvej 9, 2100, Copenhagen, Denmark.,King's College London and Guy's and St Thomas' PET Centre, School of Biomedical Engineering and Imaging Sciences, King's College London, King's Health Partners, London, UK
| | - Flemming L Andersen
- Department of Clinical Physiology, Nuclear Medicine and PET & Cluster for Molecular Imaging, Rigshospitalet, University of Copenhagen, Blegdamsvej 9, 2100, Copenhagen, Denmark
| |
Collapse
|
2
|
Wan S. Positron emission tomography/magnetic resonance imaging in musculoskeletal benign conditions. THE QUARTERLY JOURNAL OF NUCLEAR MEDICINE AND MOLECULAR IMAGING : OFFICIAL PUBLICATION OF THE ITALIAN ASSOCIATION OF NUCLEAR MEDICINE (AIMN) [AND] THE INTERNATIONAL ASSOCIATION OF RADIOPHARMACOLOGY (IAR), [AND] SECTION OF THE SOCIETY OF... 2022; 66:31-42. [PMID: 35119250 DOI: 10.23736/s1824-4785.22.03443-4] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [MESH Headings] [Track Full Text] [Subscribe] [Scholar Register] [Indexed: 06/14/2023]
Abstract
A wide spectrum of benign musculoskeletal (orthopedic and rheumatological) conditions affect the general population. Collectively, these are common, and they can inflict significant morbidity with resultant negative impact on the quality of life of patients. For many of these conditions, there is established evidence for research and clinical use of PETCT and MRI for assessment of disease. Introduction of integrated PET/MRI around a decade ago brought optimism that combining the strength of PET and MRI techniques on a single platform could have synergistic effect to benefit imaging assessment of patients, including in the context of benign musculoskeletal conditions. This review specifically focuses on the progress that has been made. This aims to showcase clinical studies derived primarily from the integrated PET/MRI platforms for the evaluation of common orthopedic and rheumatological conditions. Despite enthusiasm and progress by early adopters of the PET/MRI technology, significant barriers are present for its wider adoption, validation, and translation to routine clinical practice. Attenuation correction is a particular challenge which affects regions close to the skeleton and impacts PET/MRI assessment of musculoskeletal disorders. Continued effort on research and validation, as well as promotion of its multimodal multiparametric capability to clinical and pharmaceutical stakeholders, and increased availability through increased adoption of PET/MRI scanners internationally, may accelerate its translation into routine clinical practice in this domain.
Collapse
Affiliation(s)
- Simon Wan
- Institute of Nuclear Medicine, University College Hospital, London, UK -
| |
Collapse
|
3
|
Bogdanovic B, Solari EL, Villagran Asiares A, McIntosh L, van Marwick S, Schachoff S, Nekolla SG. PET/MR Technology: Advancement and Challenges. Semin Nucl Med 2021; 52:340-355. [PMID: 34969520 DOI: 10.1053/j.semnuclmed.2021.11.014] [Citation(s) in RCA: 1] [Impact Index Per Article: 0.3] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 11/19/2021] [Revised: 11/25/2021] [Accepted: 11/29/2021] [Indexed: 01/07/2023]
Abstract
When this article was written, it coincided with the 11th anniversary of the installation of our PET/MR device in Munich. In fact, this was the first fully integrated device to be in clinical use. During this time, we have observed many interesting behaviors, to put it kindly. However, it is more critical that in this process, our understanding of the system also improved - including the advantages and limitations from a technical, logistical, and medical perspective. The last decade of PET/MRI research has certainly been characterized by most sites looking for a "key application." There were many ideas in this context and before and after the devices became available, some of which were based on the earlier work with integrating data from single devices. These involved validating classical PET methods with MRI (eg, perfusion or oncology diagnostics). More important, however, were the scenarios where intermodal synergies could be expected. In this review, we look back on this decade-long journey, at the challenges overcome and those still to come.
Collapse
Affiliation(s)
- Borjana Bogdanovic
- Department of Nuclear Medicine, Klinikum rechts der Isar, Technische Universität München, Munich, Germany
| | - Esteban Lucas Solari
- Department of Nuclear Medicine, Klinikum rechts der Isar, Technische Universität München, Munich, Germany
| | - Alberto Villagran Asiares
- Department of Nuclear Medicine, Klinikum rechts der Isar, Technische Universität München, Munich, Germany
| | - Lachlan McIntosh
- Department of Physical Sciences, Peter MacCallum Cancer Centre, Melbourne, Australia
| | - Sandra van Marwick
- Department of Nuclear Medicine, Klinikum rechts der Isar, Technische Universität München, Munich, Germany
| | - Sylvia Schachoff
- Department of Nuclear Medicine, Klinikum rechts der Isar, Technische Universität München, Munich, Germany
| | - Stephan G Nekolla
- Department of Nuclear Medicine, Klinikum rechts der Isar, Technische Universität München, Munich, Germany; DZHK (German Centre for Cardiovascular Research), partner site Munich Heart Alliance, Munich, Germany.
| |
Collapse
|
4
|
Brancato V, Borrelli P, Alfano V, Picardi M, Mascalchi M, Nicolai E, Salvatore M, Aiello M. The impact of MR-based attenuation correction in spinal cord FDG-PET/MR imaging for neurological studies. Med Phys 2021; 48:5924-5934. [PMID: 34369590 PMCID: PMC9293017 DOI: 10.1002/mp.15149] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 10/13/2020] [Revised: 05/30/2021] [Accepted: 07/24/2021] [Indexed: 11/21/2022] Open
Abstract
Purpose Positron emission tomography (PET) attenuation correction (AC) in positron emission tomography‐magnetic resonance (PET/MR) scanners constitutes a critical and barely explored issue in spinal cord investigation, mainly due to the limitations in accounting for highly attenuating bone structures which surround the spinal canal. Our study aims at evaluating the clinical suitability of MR‐driven AC (MRAC) for 18‐fluorodeoxy‐glucose positron emission tomography (18F‐FDG‐PET) in spinal cord. Methods Thirty‐six patients, undergoing positron emission tomography‐computed tomography (PET/CT) and PET/MR in the same session for oncological examination, were retrospectively analyzed. For each patient, raw PET data from PET/MR scanner were reconstructed with 4‐ and 5‐class MRAC maps, generated by hybrid PET/MR system (PET_MRAC4 and PET_MRAC5, respectively, where PET_MRAC is PET images reconstructed using MR‐based attenuation correction map), and an AC map derived from CT data after a custom co‐registration pipeline (PET_rCTAC, where PET_rCTAC is PET images reconstructed using CT‐based attenuation correction map), which served as reference. Mean PET standardized uptake values (SUVm) were extracted from the three reconstructed PET images by regions of interest (ROIs) identified on T2‐weighted MRI, in the spinal cord, lumbar cerebrospinal fluid (CSF), and vertebral marrow at five levels (C2, C5, T6, T12, and L3). SUVm values from PET_MRAC4 and PET_MRAC5 were compared with each other and with the reference by means of paired t‐test, and correlated using Pearson's correlation (r) to assess their consistency. Cohen's d was calculated to assess the magnitude of differences between PET images. Results SUVmvalues from PET_MRAC4 were lower than those from PET_MRAC5 in almost all analyzed ROIs, with a mean difference ranging from 0.03 to 0.26 (statistically significant in the vertebral marrow at C2 and C5, spinal cord at T6 and T2, and CSF at L3). This was also confirmed by the effect size, with highest values at low spinal levels (d = 0.45 at T12 in spinal cord, d = 0.95 at L3 in CSF). SUVm values from PET_MRAC4 and PET_MRAC5 showed a very good correlation (0.81 < r < 0.97, p < 0.05) in all spinal ROIs. Underestimation of SUVm between PET_MRAC4 and PET_rCTAC was observed at each level, with a mean difference ranging from 0.02 to 0.32 (statistically significant in the vertebral marrow at C2 and T6, and CSF at L3). Although PET_MRAC5 underestimates PET_rCTAC (mean difference ranging from 0.02 to 0.3), an overall decrease in effect size could be observed for PET_MRAC5, mainly at lower spinal levels (T12, L3). SUVm from both PET_MRAC4 and PET_MRAC5 methods showed r value from good to very good with respect to PET_rCTAC (0.67 < r < 0.9 and 0.73 < r < 0.94, p < 0.05, respectively). Conclusions Our results showed that neglecting bones in AC can underestimate the FDG uptake measurement of the spinal cord. The inclusion of bones in MRAC is far from negligible and improves the AC in spinal cord, mainly at low spinal levels. Therefore, care must be taken in the spinal canal region, and the use of AC map reconstruction methods accounting for bone structures could be beneficial.
Collapse
Affiliation(s)
| | | | | | - Marco Picardi
- Department of Clinical Medicine and Surgery, Federico II University Medical School, Naples, Italy
| | - Mario Mascalchi
- «Mario Serio» Department of Experimental and Clinical Biomedical Sciences, University of Florence, Italy
| | | | | | | |
Collapse
|
5
|
Rao F, Yang B, Chen YW, Li J, Wang H, Ye H, Wang Y, Zhao K, Zhu W. A novel supervised learning method to generate CT images for attenuation correction in delayed pet scans. COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE 2020; 197:105764. [PMID: 33010702 DOI: 10.1016/j.cmpb.2020.105764] [Citation(s) in RCA: 3] [Impact Index Per Article: 0.8] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Subscribe] [Scholar Register] [Received: 05/19/2020] [Accepted: 09/14/2020] [Indexed: 06/11/2023]
Abstract
BACKGROUND AND OBJECTIVES Attenuation correction is important for PET image reconstruction. In clinical PET/CT scans, the attenuation information is usually obtained by CT. However, additional CT scans for delayed PET imaging may increase the risk of cancer. In this paper, we propose a novel CT generation method for attenuation correction in delayed PET imaging that requires no additional CT scans. METHODS As only PET raw data is available for the delayed PET scan, routine image registration methods are difficult to use directly. To solve this problem, a reconstruction network is developed to produce pseudo PET images from raw data first. Then a second network is used to generate the CT image through mapping PET/CT images from the first scan to the delayed scan. The inputs of the second network are the two pseudo PET images from the first and delayed scans, and the CT image from the first scan. The labels are taken from the ground truth CT image in the delayed scan. The loss function contains an image similarity term and a regularization term, which reflect the anatomy matching accuracy and the smoothness of the non-rigid deformation field, respectively. RESULTS We evaluated the proposed method with simulated and clinical PET/CT datasets. Standard Uptake Value was computed and compared with the gold standard (with coregistered CT for attenuation correction). The results show that the proposed supervised learning method can generate PET images with high quality and quantitative accuracy. For the test cases in our study, the average MAE and RMSE of the proposed supervised learning method were 4.61 and 22.75 respectively, and the average PSNR between the reconstructed PET image and the ground truth PET image was 62.13 dB. CONCLUSIONS The proposed method is able to generate accurate CT images for attenuation correction in delayed PET scans. Experiments indicate that the proposed method outperforms traditional methods with respect to quantitative PET image accuracy.
Collapse
Affiliation(s)
- Fan Rao
- Research Center for Healthcare Data Science, Zhejiang Lab, Hangzhou, China
| | - Bao Yang
- Research Center for Healthcare Data Science, Zhejiang Lab, Hangzhou, China
| | - Yen-Wei Chen
- Research Center for Healthcare Data Science, Zhejiang Lab, Hangzhou, China
| | - Jingsong Li
- Research Center for Healthcare Data Science, Zhejiang Lab, Hangzhou, China
| | - Hongkai Wang
- School of Biomedical Engineering, Dalian University of Technology, Dalian, China
| | - Hongwei Ye
- MinFound Medical Systems Co., Ltd. Hangzhou, China
| | - Yaofa Wang
- MinFound Medical Systems Co., Ltd. Hangzhou, China
| | - Kui Zhao
- PET Center, The First Affiliated Hospital, College of Medicine, Zhejiang University, Zhejiang, China
| | - Wentao Zhu
- Research Center for Healthcare Data Science, Zhejiang Lab, Hangzhou, China.
| |
Collapse
|
6
|
Ladefoged CN, Hansen AE, Henriksen OM, Bruun FJ, Eikenes L, Øen SK, Karlberg A, Højgaard L, Law I, Andersen FL. AI-driven attenuation correction for brain PET/MRI: Clinical evaluation of a dementia cohort and importance of the training group size. Neuroimage 2020; 222:117221. [PMID: 32750498 DOI: 10.1016/j.neuroimage.2020.117221] [Citation(s) in RCA: 37] [Impact Index Per Article: 9.3] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 11/08/2019] [Revised: 07/15/2020] [Accepted: 07/28/2020] [Indexed: 11/27/2022] Open
Abstract
INTRODUCTION Robust and reliable attenuation correction (AC) is a prerequisite for accurate quantification of activity concentration. In combined PET/MRI, AC is challenged by the lack of bone signal in the MRI from which the AC maps has to be derived. Deep learning-based image-to-image translation networks present itself as an optimal solution for MRI-derived AC (MR-AC). High robustness and generalizability of these networks are expected to be achieved through large training cohorts. In this study, we implemented an MR-AC method based on deep learning, and investigated how training cohort size, transfer learning, and MR input affected robustness, and subsequently evaluated the method in a clinical setup, with the overall aim to explore if this method could be implemented in clinical routine for PET/MRI examinations. METHODS A total cohort of 1037 adult subjects from the Siemens Biograph mMR with two different software versions (VB20P and VE11P) was used. The software upgrade included updates to all MRI sequences. The impact of training group size was investigated by training a convolutional neural network (CNN) on an increasing training group size from 10 to 403. The ability to adapt to changes in the input images between software versions were evaluated using transfer learning from a large cohort to a smaller cohort, by varying training group size from 5 to 91 subjects. The impact of MRI sequence was evaluated by training three networks based on the Dixon VIBE sequence (DeepDixon), T1-weighted MPRAGE (DeepT1), and ultra-short echo time (UTE) sequence (DeepUTE). Blinded clinical evaluation relative to the reference low-dose CT (CT-AC) was performed for DeepDixon in 104 independent 2-[18F]fluoro-2-deoxy-d-glucose ([18F]FDG) PET patient studies performed for suspected neurodegenerative disorder using statistical surface projections. RESULTS Robustness increased with group size in the training data set: 100 subjects were required to reduce the number of outliers compared to a state-of-the-art segmentation-based method, and a cohort >400 subjects further increased robustness in terms of reduced variation and number of outliers. When using transfer learning to adapt to changes in the MRI input, as few as five subjects were sufficient to minimize outliers. Full robustness was achieved at 20 subjects. Comparable robust and accurate results were obtained using all three types of MRI input with a bias below 1% relative to CT-AC in any brain region. The clinical PET evaluation using DeepDixon showed no clinically relevant differences compared to CT-AC. CONCLUSION Deep learning based AC requires a large training cohort to achieve accurate and robust performance. Using transfer learning, only five subjects were needed to fine-tune the method to large changes to the input images. No clinically relevant differences were found compared to CT-AC, indicating that clinical implementation of our deep learning-based MR-AC method will be feasible across MRI system types using transfer learning and a limited number of subjects.
Collapse
Affiliation(s)
- Claes Nøhr Ladefoged
- Department of Clinical Physiology, Nuclear Medicine & PET, Rigshospitalet, University of Copenhagen, Denmark.
| | - Adam Espe Hansen
- Department of Clinical Physiology, Nuclear Medicine & PET, Rigshospitalet, University of Copenhagen, Denmark
| | - Otto Mølby Henriksen
- Department of Clinical Physiology, Nuclear Medicine & PET, Rigshospitalet, University of Copenhagen, Denmark
| | - Frederik Jager Bruun
- Department of Clinical Physiology, Nuclear Medicine & PET, Rigshospitalet, University of Copenhagen, Denmark
| | - Live Eikenes
- Department of Circulation and Medical Imaging, Norwegian University of Science and Technology, Trondheim, Norway
| | - Silje Kjærnes Øen
- Department of Circulation and Medical Imaging, Norwegian University of Science and Technology, Trondheim, Norway
| | - Anna Karlberg
- Department of Circulation and Medical Imaging, Norwegian University of Science and Technology, Trondheim, Norway; (c)Department of Radiology and Nuclear Medicine, St. Olavs hospital, Trondheim University Hospital, Trondheim, Norway
| | - Liselotte Højgaard
- Department of Clinical Physiology, Nuclear Medicine & PET, Rigshospitalet, University of Copenhagen, Denmark
| | - Ian Law
- Department of Clinical Physiology, Nuclear Medicine & PET, Rigshospitalet, University of Copenhagen, Denmark
| | - Flemming Littrup Andersen
- Department of Clinical Physiology, Nuclear Medicine & PET, Rigshospitalet, University of Copenhagen, Denmark
| |
Collapse
|
7
|
Abstract
PURPOSE OF REVIEW Hybrid PET- MRI is a technique that has the ability to improve diagnostic accuracy in many applications, whereas PET and MRI performed separately often fail to provide accurate responses to clinical questions. Here, we review recent studies and current developments in PET-MRI, focusing on clinical applications. RECENT FINDINGS The combination of PET and MRI imaging methods aims at increasing the potential of each individual modality. Combined methods of image reconstruction and correction of PET-MRI attenuation are being developed, and a number of applications are being introduced into clinical practice. To date, the value of PET-MRI has been demonstrated for the evaluation of brain tumours in epilepsy and neurodegenerative diseases. Continued advances in data analysis regularly improve the efficiency and the potential application of multimodal biomarkers. SUMMARY PET-MRI provides simultaneous of anatomical, functional, biochemical and metabolic information for the personalized characterization and monitoring of neurological diseases. In this review, we show the advantage of the complementarity of different biomarkers obtained using PET-MRI data. We also present the recent advances made in this hybrid imaging modality and its advantages in clinical practice compared with MRI and PET separately.
Collapse
|