AI-based Virtual Synthesis of Methionine PET from Contrast-enhanced MRI: Development and External Validation Study
Carbon 11 (11C)–methionine is a useful PET radiotracer for the management of patients with glioma, but radiation exposure and lack of molecular imaging facilities limit its use.
To generate synthetic methionine PET images from contrast-enhanced (CE) MRI through an artificial intelligence (AI)–based image-to-image translation model and to compare its performance for grading and prognosis of gliomas with that of real PET.
Materials and Methods
An AI-based model to generate synthetic methionine PET images from CE MRI was developed and validated from patients who underwent both methionine PET and CE MRI at a university hospital from January 2007 to December 2018 (institutional data set). Pearson correlation coefficients for the maximum and mean tumor to background ratio (TBRmax and TBRmean, respectively) of methionine uptake and the lesion volume between synthetic and real PET were calculated. Two additional open-source glioma databases of preoperative CE MRI without methionine PET were used as the external test set. Using the TBRs, the area under the receiver operating characteristic curve (AUC) for classifying high-grade and low-grade gliomas and overall survival were evaluated.
The institutional data set included 362 patients (mean age, 49 years ± 19 [SD]; 195 female, 167 male; training, n = 294; validation, n = 34; test, n = 34). In the internal test set, Pearson correlation coefficients were 0.68 (95% CI: 0.47, 0.81), 0.76 (95% CI: 0.59, 0.86), and 0.92 (95% CI: 0.85, 0.95) for TBRmax, TBRmean, and lesion volume, respectively. The external test set included 344 patients with gliomas (mean age, 53 years ± 15; 192 male, 152 female; high grade, n = 269). The AUC for TBRmax was 0.81 (95% CI: 0.75, 0.86) and the overall survival analysis showed a significant difference between the high (2-year survival rate, 27%) and low (2-year survival rate, 71%; P < .001) TBRmax groups.
The AI-based model-generated synthetic methionine PET images strongly correlated with real PET images and showed good performance for glioma grading and prognostication.
Published under a CC BY 4.0 license.
Synthetic methionine PET images generated from MRI using an artificial intelligence–based model strongly correlated with real methionine PET and showed good performance, similar to that of real methionine PET, for grading and prognosis of brain gliomas.
■ Using 32 075 paired images from 362 patients (training, n = 294; validation, n = 34; test, n = 34) who underwent methionine PET and contrast-enhanced (CE) MRI, artificial intelligence–based synthetic methionine PET from CE MRI had Pearson correlation coefficients of 0.68, 0.76, and 0.92 for maximum tumor to background ratio (TBRmax), mean TBR, and lesion volume, respectively, compared with real PET.
■ Using an external test set of 41 755 CE MRI studies from 344 patients with gliomas (high-grade, n = 269; low-grade, n = 75), the area under the receiver operating characteristic curve for classifying high- and low-grade gliomas from synthetic PET was 0.81 (TBRmax).
■ Overall survival showed significant differences between the high TBRmax group (2-year survival rate, 27%) and low TBRmax group (2-year survival rate, 71%; P < .001).
Carbon 11 (11C)–methionine is a PET radiotracer often used to evaluate methionine uptake in areas of active amino acid metabolism in patients with suspected brain tumor (1). 11C-methionine is useful for glioma grading, prognosis prediction, and delineation of tumor extent (1–4). Guidelines and a prior study support the use of methionine PET for preoperative evaluation of patients with glioma (5–7). However, methionine PET examinations are resource-intensive and involve ionizing radiation (8). MRI is often used as the sole imaging modality to evaluate glioma, although methionine PET outperforms MRI for grading and delineation of tumor extent (5).
A correlation between methionine uptake and contrast-enhanced (CE) areas on MRI scans has been reported in gliomas (9–13). Image-to-image translation models involve the application of generative adversarial networks that can extract features from images (14–16). A previous study reported the successful generation of synthetic fluorine 18 fluorodeoxyglucose PET images from MRI (17). At present, few, if any, published studies have used artificial intelligence (AI) to generate synthetic methionine PET images from MRI.
The aim of this study was to generate synthetic methionine PET images from CE MRI through an AI-based image-to-image translation model and to compare the model performance for grading and prognosis of gliomas with that of real PET from previous studies.
Materials and Methods
This retrospective study was approved by the institutional ethics committee (approval ID: 2022–059), and the requirement for informed consent was waived for the institutional data set because the images were collected for a previous clinical or research purpose. This study complies with the Declaration of Helsinki and was structured according to Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis, or TRIPOD, guidelines (18).
To extract imaging features, an image-to-image translation technique was used (14,15). Paired images from PET and CE MRI examinations performed at a single university hospital were collected. These images were used to train, validate, and test the image-to-image translation AI model. The developed AI model was externally tested to show its performance for grading and prognosis of synthetic methionine PET images generated from CE MRI scans in patients with gliomas. Part of the data (56 of 362 patients) used in this study overlap with a previous study (19) that investigated the correlation between the prognosis of patients with glioma and imaging findings of real methionine PET, but did not explore the relationship between synthetic methionine PET and MRI.
First, an institutional data set for training, validating, and internally testing the AI model was prepared. Patients who underwent both methionine PET and MRI examinations at a single university hospital from January 2007 to December 2018 were retrospectively included. Exclusion criteria were as follows: (a) methionine PET and MRI examinations performed more than 1 month apart, (b) missing CE MRI studies, (c) methionine PET images showing no lesion, (d) methionine PET and CE MRI scans that failed registration, and (e) methionine PET or CE MRI scans showing artifacts. Acquisition parameters are detailed in Appendix S1. The diagnosis was established based on histopathologic assessment if the patient underwent surgery; if the patient did not undergo surgery, the diagnosis was established according to international guidelines for each disease (20–22).
Subsequently, an external test set that included MRI scans, histopathologic results, tumor grading, and prognostic data in patients diagnosed with primary gliomas who underwent preoperative CE MRI but not methionine PET were collected. These examination data, which are from several institutions in the United States and Italy, were collected from two open-source data sets of The Cancer Imaging Archive (23–25). The exclusion criteria for the external test set were as follows: (a) missing CE MRI studies, (b) CE MRI scans showing artifacts, (c) CE MRI studies showing postoperative change, (d) missing patient pathologic data, and (e) missing patient follow-up data.
Partition of Data Sets
All institutional patient data were randomly divided into training, validation, and internal test sets at a ratio of 8:1:1. Patient-based partition ensured that there was no overlap of images or patients among the respective sets. All patient data in the external test set were from countries different from that of the study institution and were used to demonstrate the generalization of the AI model.
Ground Truth Labeling
In methionine PET images, a tumor boundary was delineated with a threshold of 1.3 times the mean standardized uptake value (SUV) of reference regions of interest according to a previous study (26). Details of the radiologists’ labeling are shown in Appendix S1. Regarding the tumor to background ratios (TBRs) of methionine uptake, maximum TBR (TBRmax) and mean TBR (TBRmean) were defined as the maximum and mean SUV of the tumor, respectively, divided by the mean SUV of reference regions of interest (26).
AI Model Overview
The AI model was developed based on the pix2pix model, which is a generative adversarial network that performs image-to-image translation using extracted features between image pairs (27). The model has two important differences compared with conventional pix2pix development, including (a) the three-dimensional data handling method and (b) the data sampling method. These allow efficient feature learning regarding the methionine uptake in lesions. A schematic overview of the developed model is shown in Figure S1. The model was designed to create one synthetic methionine PET image from three consecutive sections of CE MRI scans, which reflect the three-dimensional information of the lesion. Both full and disease-cropped images were used during the learning process to allow extraction of the characteristics of both normal and abnormal methionine uptake. This data sampling method enables the AI model to efficiently learn both local lesion information and information from the entire brain. Details regarding this image processing are shown in Appendix S1 and Figures S2–S5.
After these preprocessing steps, the pix2pix model (27) was trained, tuned, and evaluated on the training, validation, and internal test sets, respectively. As reported in the original pix2pix publication, 200 training epochs were used. These models were built with PyTorch (version 1.10.0; https://pytorch.org/). The source code for the developed AI model is available online (https://github.com/synthetic-METPET). Further details regarding the model and machine environment are outlined in Appendix S1.
In the internal test set, Pearson correlation coefficients and Bland-Altman plots were calculated for TBRmax, TBRmean, and lesion volume between the synthetic and real methionine PET images. The 95% CIs were calculated by a bootstrap method (bootstrap sample size = 1000). Using the external test set, area under the receiver operating characteristic curve (AUC) analysis was performed using the TBRs of the synthetic methionine PET images to differentiate between high-grade and low-grade gliomas. Sample size calculations for the AUC analysis are available in Appendix S1. Additionally, overall survival analysis was performed using the Kaplan-Meier method. The optimal cutoff value for overall survival analysis using the TBRs of the synthetic methionine PET images was determined by the Youden index in the internal test set. This cutoff was used to stratify patients in the external test set into high-risk and low-risk groups. The survival curves for the high- and low-risk groups were statistically compared using the log-rank test. Overall survival was defined as the time between the date of diagnosis and death. When the death of a patient was not able to be confirmed, that patient was defined as censored. The follow-up interval for censored patients was the time between the date of diagnosis and last contact. P < .05 was considered indicative of a statistically significant difference. All statistics were performed at the examination level using R (version 4.0.0; http://www.r-project.org). Additional analyses are described in Appendix S1.
Characteristics of the Data Sets
Of the 1295 methionine PET examinations performed at a single institution from January 2007 to December 2018, 698 were excluded due to missing corresponding MRI studies within 1 month of the methionine PET examination. Among the 597 remaining methionine PET and MRI pairs, 122 examination pairs were excluded due to missing CE MRI scans, registration failure, artifacts, or no lesions present on the methionine PET images. Overall, 32 075 image pairs from 475 methionine PET and CE MRI examination pairs in 362 patients (mean age, 49 years ± 19 [SD]; 195 female, 167 male) were included in the institutional data set. The training set included 25 848 image pairs from 384 examination pairs in 294 patients (mean age, 48 years ± 18; 162 female, 132 male); the validation set included 3099 image pairs from 46 examinations in 34 patients (mean age, 54 years ± 17; 18 male, 16 female ); and the internal test set included 3128 image pairs from 45 examination pairs in 34 patients (mean age, 50 years ± 17; 17 female, 17 male). In the external test set, 457 MRI studies were collected from two open-source data sets of patients with gliomas provided by The Cancer Imaging Archive. Among them, 113 MRI studies were excluded due to postoperative changes, artifacts, no CE MRI scans, no patient pathologic data, or no patient follow-up data. Finally, 41 755 images were included from 344 CE MRI examinations in 344 patients (mean age, 53 years ± 15; 192 male, 152 female; 269 high-grade and 75 low-grade gliomas). Flowcharts of the data sets are shown in Figure 1. Tables 1 and 2 show patient demographics and characteristics of the data sets. The vendors of the PET and MRI systems are shown in Table S1.
Model Development Results in the Internal Test Set
Pearson correlation coefficients between the synthetic and real methionine PET images were 0.68 (95% CI: 0.47, 0.81) for TBRmax, 0.76 (95% CI: 0.59, 0.86) for TBRmean, and 0.92 (95% CI: 0.85, 0.95) for lesion volume (all P < .001) (Table 3). The mean differences in TBRmax, TBRmean, and lesion volume between the synthetic and real methionine PET were 0.05 (95% CI: −1.9, 2), 0.02 (−0.45, 0.49), and 5990 mm3 (95% CI: −27 000, 39 000), respectively. Figure 2 shows the scatterplots and Bland-Altman plots of TBRmax, TBRmean, and lesion volume for the internal test set. Results of additional analyses in the internal test set are described in Appendix S1. Figure S6 shows the AUC analysis for differentiation between high- and low-grade gliomas in the internal test set. Figure 3 shows representative images of methionine PET and CE MRI scans from the internal test set. In addition, representative methionine PET and CE MRI scans of radiation necrosis and tumor progression cases are shown in Figure S7.
Model Evaluation Results in the External Test Set
The AUCs for differentiating between high- and low-grade gliomas were 0.81 (95% CI: 0.75, 0.86) for TBRmax and 0.78 (95% CI: 0.73, 0.84) for TBRmean. Worse prognosis was found in the group with a TBRmax higher than 3.8 (2-year survival rate, 27%) compared with the group with a lower TBRmax (2-year survival rate, 71%; P < .001). Worse prognosis was observed in the group with a TBRmean higher than 1.6 (2-year survival rate, 30%) compared with the group with a lower TBRmean (2-year survival rate, 81%; P < .001). Table 3 shows the results for the external test set. Figures 4 and 5 show the receiver operating characteristic and Kaplan-Meier curves, respectively, for the external test set. Representative synthetic methionine PET images and their associated CE MRI scans from the external test set are shown in Figure 6. Results of additional analyses in the external test set are described in Appendix S1. Figure S8 shows the AUC analysis for differentiation between isocitrate dehydrogenase–mutant and isocitrate dehydrogenase wild-type gliomas. Figure S9 demonstrates the reader performance test. Table S2 shows a summary of AUC analyses for differentiation between high- and low-grade gliomas in the internal and external test sets.
Carbon 11 (11C)–methionine is a useful PET radiotracer for the management of glioma, but its use is limited due to a lack of molecular imaging facilities. In this study, we developed an artificial intelligence model using an image-to-image translation technique to generate synthetic methionine PET images from contrast-enhanced (CE) MRI. In the internal test set, Pearson correlation coefficients between synthetic and real PET were 0.68 (95% CI: 0.47, 0.81), 0.76 (95% CI: 0.59, 0.86), and 0.92 (95% CI: 0.85, 0.95) for the maximum tumor to background ratio (TBRmax), mean tumor to background ratio (TBRmean), and lesion volume, respectively (all P < .001). In the external test set, the areas under the receiver operating characteristic curve for differentiating between high-grade and low-grade gliomas with synthetic methionine PET were 0.81 (95% CI: 0.75, 0.86) for TBRmax and 0.78 (95% CI: 0.73, 0.84) for TBRmean. The group with TBRmax and TBRmean values higher than the cutoff had significantly shorter overall survival than the group with lower ratios (both P < .001). These findings support the clinical utility of synthetic methionine PET images generated from CE MRI.
Synthetic methionine PET can have additive value for classifying and prognosticating patients with glioma, with similar performance to real methionine PET. Results from the internal test set of this study showed strong correlations between the TBRs of the real and synthetic methionine PET images. In the external test set, synthetic methionine PET showed good performance in differentiating between high- and low-grade gliomas. Prior studies have demonstrated TBRs from real methionine PET to be useful for glioma grading and prognosis prediction (12,28–30). Furthermore, previous studies of real methionine PET for differentiation between high- and low-grade gliomas reported that the AUC of TBRs ranged from 0.72 to 0.92 (31–33). The AUCs of TBRs for the synthetic methionine PET in our external test set were similar. In addition, overall survival of patients with gliomas was significantly shorter in the higher TBR group than in the lower TBR group when using real methionine PET (29). The results of overall survival analysis using the TBRs for the synthetic methionine PET in our external test set were consistent with these reports. Although our data were different from previous reports and cannot simply be compared, this suggests the potential clinical usefulness of synthetic methionine PET.
There was a strong correlation between the lesion volume of real and synthetic methionine PET images. In patients with suspected glioma, preoperative MRI is usually performed to evaluate tumor extent (2). However, high-grade gliomas may extend beyond the CE areas on CE MRI scans (34,35), and hyperintense areas on T2-weighted and fluid-attenuated inversion recovery images can represent both tumor cells and peritumoral edema (36). Methionine PET examinations can evaluate the extent of metabolically active gliomas (37), and the use of methionine PET examinations has been recommended for preoperatively evaluating the tumor extent in patients with gliomas (5–7). Synthetic methionine PET images may be useful for the evaluation of the extent of metabolically active gliomas.
Although further validation is needed, there are several hypotheses as to why our AI model could generate synthetic methionine PET images similar to real methionine PET images from CE MRI scans. Both methionine uptake and enhancement of CE MRI scans are reported to be related to glioma cell proliferation and angiogenesis (9–12). Methionine uptake and T1 relaxation time have been shown to be associated with tumor cell density in gliomas without enhancement on CE MRI scans (13). Another study showed that the extent of methionine uptake can predict the extent of the contrast-enhancing lesion on CE MRI scans in glioblastoma (38). Our model may automatically learn these reported relationships as it generates synthetic methionine PET images from CE MRI.
Our study had limitations. First, this was a retrospective study with a relatively small number of images and patients. Second, there were no methionine PET images in the external test set.
In conclusion, we developed an artificial intelligence model based on the pix2pix model, a generative adversarial network, to generate synthetic methionine PET images from contrast-enhanced (CE) MRI scans. These synthetic methionine PET images strongly correlated with real methionine PET images and showed good performance for grading and prognostication of gliomas, similar to that of real methionine PET from previous studies. Future prospective and large-scale studies are warranted to evaluate the clinical value of synthetic methionine PET based on CE MRI.
Author contributions: Guarantors of integrity of entire study, H. Takita, K.N., Y. Mitsuyama, D.U.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; agrees to ensure any questions related to the work are appropriately resolved, all authors; literature research, H. Takita, Y. Mitsuyama, D.U.; clinical studies, K.N.; experimental studies, H. Takita, T.M., Y.K., T.U.; statistical analysis, H. Takita, H. Tatekawa, Y. Mitsuyama, D.U.; and manuscript editing, H. Takita, H. Tatekawa, Y. Mitsuyama, S.L.W., Y. Miki, D.U.
H. Takita supported by Bayer Yakuhin (Bayer Academic Support: BASJ20220408012) and Japan Society for the Promotion of Science (JSPS) KAKENHI (23K14899). H. Tatekawa supported by JSPS KAKENHI (22K15866).
- 1. . Value of 11C-methionine PET in imaging brain tumours and metastases. Eur J Nucl Med Mol Imaging 2013;40(4):615–635. Crossref, Medline, Google Scholar
- 2. . Delineation of brain tumor extent with [11C]L-methionine positron emission tomography: local comparison with stereotactic histopathology. Clin Cancer Res 2004;10(21):7163–7170. Crossref, Medline, Google Scholar
- 3. . 11C-methionine PET for differential diagnosis of low-grade gliomas. Neurology 1998;50(5):1316–1322. Crossref, Medline, Google Scholar
- 4. . Cerebral glioma: evaluation with methionine PET. Radiology 1993;186(1):45–53. Link, Google Scholar
- 5. . Response Assessment in Neuro-Oncology working group and European Association for Neuro-Oncology recommendations for the clinical use of PET imaging in gliomas. Neuro Oncol 2016;18(9):1199–1208. Crossref, Medline, Google Scholar
- 6. . Joint EANM/EANO/RANO practice guidelines/SNMMI procedure standards for imaging of gliomas using PET with radiolabelled amino acids and [18F]FDG: version 1.0. Eur J Nucl Med Mol Imaging 2019;46(3):540–557. Crossref, Medline, Google Scholar
- 7. . Preoperative [11C]methionine PET to personalize treatment decisions in patients with lower-grade gliomas. Neuro Oncol 2022;24(9):1546–1556. Crossref, Medline, Google Scholar
- 8. . Brain tumour imaging with PET: a comparison between [18F]fluorodopa and [11C]methionine. Eur J Nucl Med Mol Imaging 2003;30(11):1561–1567. Crossref, Medline, Google Scholar
- 9. . Transport mechanisms of 3-[123I]iodo-alpha-methyl-L-tyrosine in a human glioma cell line: comparison with [3H]methyl]-L-methionine. J Nucl Med 2000;41(7):1250–1255. Medline, Google Scholar
- 10. . Methyl-[11C]- l-methionine uptake as measured by positron emission tomography correlates to microvessel density in patients with glioma. Eur J Nucl Med Mol Imaging 2003;30(6):868–873. Crossref, Medline, Google Scholar
- 11. . MRI enhancement and microvascular density in gliomas. Correlation with tumor cell proliferation. Invest Radiol 1999;34(6):427–434. Crossref, Medline, Google Scholar
- 12. . Stereotactic comparison among cerebral blood volume, methionine uptake, and histopathology in brain glioma. AJNR Am J Neuroradiol 2007;28(3):455–461. Medline, Google Scholar
- 13. . Magnetic Resonance Relaxometry for Tumor Cell Density Imaging for Glioma: An Exploratory Study via 11C-Methionine PET and Its Validation via Stereotactic Tissue Sampling. Cancers (Basel) 2021;13(16):4067. Crossref, Medline, Google Scholar
- 14. . Deep learning. Nature 2015;521(7553):436–444. Crossref, Medline, Google Scholar
- 15. . Deep Learning-A Technology With the Potential to Transform Health Care. JAMA 2018;320(11):1101–1102. Crossref, Medline, Google Scholar
- 16. . Conditional Generative Adversarial Nets. arXiv preprint arXiv:1411.1784. https://arxiv.org/abs/1411.1784. Posted November 6, 2014. Accessed September 10, 2022. Google Scholar
- 17. . BPGAN: Brain PET synthesis from MRI using generative adversarial network for multi-modal Alzheimer’s disease diagnosis. Comput Methods Programs Biomed 2022;217:106676. Crossref, Medline, Google Scholar
- 18. . Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): explanation and elaboration. Ann Intern Med 2015;162(1):W1–W73. Crossref, Medline, Google Scholar
- 19. . Maximum 11C-methionine PET uptake as a prognostic imaging biomarker for newly diagnosed and untreated astrocytic glioma. Sci Rep 2022;12(1):546. Crossref, Medline, Google Scholar
- 20. . Immunotherapy response assessment in neuro-oncology: a report of the RANO working group. Lancet Oncol 2015;16(15):e534–e542. Crossref, Medline, Google Scholar
- 21. . Response assessment in neuro-oncology (a report of the RANO group): assessment of outcome in trials of diffuse low-grade gliomas. Lancet Oncol 2011;12(6):583–593. Crossref, Medline, Google Scholar
- 22. . Updated response assessment criteria for high-grade gliomas: response assessment in neuro-oncology working group. J Clin Oncol 2010;28(11):1963–1972. Crossref, Medline, Google Scholar
- 23. . The Cancer Imaging Archive (TCIA): maintaining and operating a public information repository. J Digit Imaging 2013;26(6):1045–1057. Crossref, Medline, Google Scholar
- 24. . The Cancer Genome Atlas Low Grade Glioma Collection (TCGA-LGG), The Cancer Imaging Archive. 2016. Google Scholar
- 25. . The Cancer Genome Atlas Glioblastoma Multiforme Collection (TCGA-GBM), The Cancer Imaging Archive. 2016. Google Scholar
- 26. . Prognostic value of volume-based measurements on (11)C-methionine PET in glioma patients. Eur J Nucl Med Mol Imaging 2015;42(7):1071–1080. Crossref, Medline, Google Scholar
- 27. . Image-to-Image Translation with Conditional Adversarial Networks. In:
2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017; 5967–5976. Google Scholar
- 28. . A comparative study of thallium-201 SPET, carbon-11 methionine PET and fluorine-18 fluorodeoxyglucose PET for the differentiation of astrocytic tumours. Eur J Nucl Med 1998;25(9):1261–1269. Crossref, Medline, Google Scholar
- 29. . 11C-methionine PET as a prognostic marker in patients with glioma: comparison with 18F-FDG PET. Eur J Nucl Med Mol Imaging 2005;32(1):52–59. Crossref, Medline, Google Scholar
- 30. . Preoperative evaluation of 54 gliomas by PET with fluorine-18-fluorodeoxyglucose and/or carbon-11-methionine. J Nucl Med 1998;39(5):778–785. Medline, Google Scholar
- 31. . 11C-methionine PET for grading and prognostication in gliomas: a comparison study with 18F-FDG PET and contrast enhancement on MRI. J Nucl Med 2012;53(11):1709–1715. Crossref, Medline, Google Scholar
- 32. . Discrimination between primary low-grade and high-grade glioma with 11C-methionine PET: a bivariate diagnostic test accuracy meta-analysis. Br J Radiol 2018;91(1082):20170426. Crossref, Medline, Google Scholar
- 33. . Oligodendroglial component complicates the prediction of tumour grading with metabolic imaging. Eur J Nucl Med Mol Imaging 2015;42(6):896–904. Crossref, Medline, Google Scholar
- 34. . Imaging-based stereotaxic serial biopsies in untreated intracranial glial neoplasms. J Neurosurg 1987;66(6):865–874. Crossref, Medline, Google Scholar
- 35. . Magnetic resonance imaging and histopathology of cerebral gliomas. Neuroradiology 1992;34(6):463–469. Crossref, Medline, Google Scholar
- 36. . L-(methyl-11C) methionine positron emission tomography for target delineation in resected high-grade gliomas before radiotherapy. Int J Radiat Oncol Biol Phys 2005;63(1):64–74. Crossref, Medline, Google Scholar
- 37. . Quantitative volumetric analysis of gliomas with sequential MRI and 11C-methionine PET assessment: patterns of integration in therapy planning. Eur J Nucl Med Mol Imaging 2012;39(5):771–781. Crossref, Medline, Google Scholar
- 38. . Prediction of nuclide accumulation spread based on the volume of enhancing MRI lesion in glioblastoma patients. J Neurosurg Sci 2021. https://doi.org/10.23736/S0390-5616.21.05353-4. Published online October 14, 2021. Crossref, Medline, Google Scholar
Article HistoryReceived: Nov 24 2022
Revision requested: Feb 1 2023
Revision received: June 30 2023
Accepted: June 12 2023
Published online: Aug 01 2023