Introduction

Glioma is a heterogeneous group of primary brain tumors, with considerable chemical and histological heterogeneity, posing significant challenges to their treatment and management. Gliomas accounted for 26.3%, of all primary Central Nervous System (CNS) tumors and approximately 80% of malignant brain tumors [ 1 , 2 ].

Glioma treatment typically employs a multimodal approach, aiming to remove or reduce tumor size while effectively managing associated symptoms [ 3 - 6 ]. Tumor surgery is often the first step to preserving brain function while removing the tumor completely. Due to the aggressive behavior, postoperative external beam radiation, concurrent Temozolomide (TMZ) chemotherapy, and additional adjuvant temozolomide were administered after surgery to target remaining tumor cells and lower the chance of recurrence [ 7 , 8 ]. The use of these combination therapies has been associated with improved median survival rates for glioma patients [ 9 ]. Despite the postoperative care, the median Overall Survival (OS) for these patients is just fifteen months [ 10 ]. Gliomas present notable challenges in clinical management due to their diverse histological and molecular characteristics. One of the critical issues in glioma treatment is the accurate differentiation between post treatment or treatment-related changes, including Pseudoprogression (PSP), Radio-necrosis (RN), and true glioma progression, including progression/recurrence because each requires distinct therapeutic interventions with a significant effect on patient prognosis and outcomes. Moreover, monitoring tumor progression is essential in these patients after treatment, as it helps determine the next steps in their care. Confirming whether a tumor is recurring or progressing informs surgeons and patients about the best treatment strategy [ 11 , 12 ].

Previous studies have indicated the importance of imaging modalities including Positron Emission Tomography (PET) and Magnetic Resonance Imaging (MRI) in accurately characterizing glioma progression, pre and post treatment management [ 13 - 18 ]. Advanced imaging techniques, such as amino acid PET including O-(2-[18F] fluoroethyl)-L-tyrosine (18F-FET), 18F-fluorodeoxyglucose (18F-FDG), 11C-methionine (11C-MET), 3,4-dihydroxy-6-[18F]-fluoro-l-phenylalanine (18F-DOPA), and 11C-choline (11C-CHO) [ 10 , 19 - 21 ] and cerebral blood volume (CBV) evaluation with perfusion-weighted MRI, and Magnetic Resonance Spectroscopy (MRS), suggest deeper analysis of tumor characteristics and microenvironment than conventional MRI. Several studies represented superior performance for amino acid PET, such as MET PET (accuracy of 89.6%) [ 19 ], DOPA (accuracy of 82%) [ 22 ], and FET (accuracy ranging between 81% and 99%) [ 23 , 24 ] for distinguishing glioma post treatment recurrence/progression and provide complementary and comparable information to the MRI as a gold standard. They represented that amino acid PET can be an essential tool for distinguishing tumor progression/recurrence from PSP, and radiation necrosis [ 10 ].

In recent years, radiomics, Artificial Intelligence (AI), Deep Learning (DL), and Machine Learning (ML) have enhanced research on glioma patients [ 12 , 13 , 15 , 17 , 25 - 29 ]. Accordingly, PET imaging and advancements in radiomics analysis hold promise in providing quantitative and qualitative insights into glioma behavior and helping in distinct diagnosis, distinguishing recurrence from treatment-associated modifications, and treatment strategizing [ 15 ].

Regardless of growing interest in the use of radiomics and PET imaging for glioma characterization, a need for a systematic review to explore the application of PET-based radiomic models has remained for prognosis and post-treatment assessment in glioma. This study aimed to systematically investigate the capabilities of PET-based radiomics and AI models for the investigation of prognosis and differentiation post-treatment changes, including progression, recurrence, PSP, and radio-necrosis in glioma patients. The present study aimed to clarify the diagnostic and prognostic utility of these models and identify areas that need further research and clinical application.

Material and Methods

Eligibility Criteria

The present study included cross-sectional, case-control, cohort, and studies assessing the relevance of PET-based AI models (radiomics, ML, and DL models) in assessing prognosis and differentiating glioma progression or recurrence from treatment-related changes. Case reports, case series, review articles, editorials, and commentary and abstracts of articles were excluded from the study. The study also excluded papers evaluating MRI-based AI models, studies assessing glioma segmentation and Isocitrate Dehydrogenase (IDH)-mutant with AI.

Search Strategy

In June 2024, the following databases were electronically searched from January 1^st, 2005, to June 31^st, 2024 including Web of Science databases, MEDLINE, PubMed, and EMBASE using a detailed search strategy with no language restriction. The search terms used were “amino acid PET”, “PET”, “glioblastoma”, “glioma”, “positron emission tomography”, “machine learning”, “deep learning”, “radiomics”, “artificial intelligence”, “AI”, “prognosis”, “outcome”, “post treatment changes”, “treatment-related changes”, “progression”, “true progression” “pseudo-progression”, and “necrosis”. The study’s methodology was derived from an initial review of related literature and a comprehensive internet search approach of systematic reviews of radiation therapy outcomes and PET based radiomics, ML, and DL models. The search strategy was reviewed and commented on by two independent medical librarians. In addition, the reference list of related papers included in the study was examined. Afterward, the results were hand-searched based on the titles and abstracts to exclude studies according to exclusion criteria.

Selection process

The titles and abstracts of the recognized citations were reviewed by two independent reviewers using a standardized screening guide. We retrieved the full text of all citations recognized as qualified by at least one reviewer. A standardized, pilot-tested format was then used by two reviewers to independently screen the full text for eligibility. Third reviewers were contacted to resolve disagreements.

Data abstraction

The selected articles were abstracted by two independent reviewers based on a standard grid comprising the following items: purpose of the study, comparison method, training, dataset, models, validation, test, Area Under Curve (AUC), sensitivity and specificity, accuracy, and conclusion. In addition, to assess the methodological quality of the involved studies, two reviewers used the Radiomic Quality Score (RQS) [ 30 ], a tool based on sixteen items that indicates the quality of radiomics study. Results of analyses restricted to PET based AI models were recorded for prognosis, post treatment changes, and radiotherapy outcomes in glioma patients. PRISMA checklist was used to evaluate the overall quality of evidence for each outcome.

Results

Study selection

A total of 1056 articles were found in the initial literature review. Title and abstract screening were done, and after that 112 studies were selected for full-text screening, of which 101 studies were excluded. Finally, 11 articles were included in the systematic review ( Figure 1). Duplicate and conference data were the most common exclusions.

Figure 1. Database and register flow diagram for identifying studies

PET radiotracer

The majority of included studies (10 studies) (Table 1) focused on radiolabeled amino acid PET. Amino acid radiotracers, including 6 studies used O-(2-[18F] fluoroethyl)-L-tyrosine (FET) [ 31 - 37 ], 2 studies utilized L-3,4-dihydroxy-6-[18F]-fluoro-phenylalanine (18F-DOPA), and 2 studies used 11C-methionine (MET). In addition, one study used a nucleoside analog probe: 3’-deoxy-3’-18F-fluorothymidine (FLT) (Table 1).

PET Radiomics for differentiation post treatment changes from glioma progression

The results revealed that radiomics, ML, and DL models based on PET imaging demonstrate superior performance in differentiating post treatment changes in comparison to conventional PET parameters, such as maximum Tumor-to-Brain ratios (TBRmax), mean Tumor-to-Brain ratios (TBRmean), and Time to Peak (TTP) models in glioma patients [ 13 , 32 , 33 , 37 ]. In addition, the model integrating radiomics features and the conventional PET parameters achieved superior diagnostic performance compared to radiomics and conventional parameter models solely (AUC of 0.85, 0.85, 0.78, sensitivity of 0.81, 0.73, 0.66, and specificity of 0.70, 0.80, 0.80, respectively) in distinguishing treatmentrelated changes from glioma progression [ 33 ]. The dynamic PET imaging dataset outperformed static imaging-based models, with the combined dynamic and static feature models. The AUC values for dynamic, combined static/dynamic, and static models were 0.805, 0.79, and 0.715, respectively, in a cohort of 85 patients [ 13 ]. The random forest model utilizing MET PET achieved highest performance in distinguishing recurrent brain tumor from radiation necrosis, with an AUC of 0.98 in 41 patients [ 38 ]. The FET PET radiomics models achieved an AUC of 0.93 for Linear Discriminant Analysis (LDA)-model [ 31 ], 0.85 for Logistic regression model [ 33 ], 0.74 [ 37 ], and 0.85 [ 35 ] for random forest in post treatment differentiation. The findings from the 18F-DOPA PET radiomics models demonstrated AUC values of 0.715 for static features and 0.805 for dynamic features when using the ElasticNet logistic regression model [ 13 ], For the XGBoost (XGB) model, the AUC values were 0.715 and 0.755 for static and dynamic datasets, respectively, while the random forest model achieved AUC values of 0.832 and 0.749 for static and dynamic datasets, respectively [ 13 ]. Table 2 showed the results of PET based radiomics models for differentiation treatment related changes in glioma patients.

PET Radiomics for glioma survival stratification and prognosis

Generally, PET based DL and radiomics models represented higher performance than conventional models for predicting tumor proliferation, recurrence and survival of glioma patients (Table 2).

As compared to conventional clinical parameters alone, the combination of clinical parameters and radiomics using pretreatment dynamic FET PET data improved the prognostic accuracy for short-term survival assessment of patients with newly diagnosed IDH-wildtype glioblastoma [ 36 ]. The FET-PET demonstrated potential in predicting recurrence in glioblastoma (GBM) patients undergoing re-irradiation, with an AUC and sensitivity of 0.66 and 0.78, respectively [ 34 ]. In addition, the 3D-DenseNet model based on postoperative 11C-MET PET demonstrated strong performance in identifying residual tumors, achieving an AUC of 0.95. In terms of time to recurrence (TTR) and OS prognosis, the 3D-DenseNet model based on MET PET, which incorporated age and O6-methylguanine-DNA methyltransferase (MGMT) promoter methylation status achieved the best performance, with concordance indices of 0.68 for TTR and 0.65 for OS [ 40 ]. In addition, 18F-DOPA delta radiomics, using Recursive Feature Elimination (RFE) and Support Vector Machine (SVM) represented high performance in predicting progressionfree survival in rare highgrade glioma, with a concordance index of 0.751 (95% CI: 0.716–0.784, P-value=0.06) [ 25 ].

Discussion

The present study is a systematic review on relevant clinical topics and challenges of glioma prognosis and distinguish treatment related changes, including progression, PSP, and RN using PET based radiomics models. This review summarizes the key findings and discusses their implications for practice. Such a review is essential for advancing personalized medicine in the care of glioma patients and ultimately improving patients’ treatment outcomes and quality of life.

In the course of glioma treatment, progression or recurrence, PSP, and RN are all potential outcomes [ 41 ]. Approximately 60% of patients with low-grade gliomas experience tumor recurrence within five years [ 42 , 43 ], while around 40% of grade III gliomas and 90% of grade IV gliomas progress within two years [ 42 , 44 ]. Recurrence in glioma, particularly in high-grade forms, such as glioblastoma, remains a significant challenge in neuro-oncology. Gliomas, characterized by diffuse infiltration of adjacent brain tissue, frequently recur despite aggressive treatment approaches, including surgery, radiation, and chemotherapy. Recurrence is typically localized to the primary tumor site, few centimeters around the tumor bed and the resection site, although multifocal or distant recurrences can also occur [ 45 , 46 ]. Even with the addition of temozolomide to GBM radiation therapy, the most common site of initial recurrence remains local failure [ 4 , 45 , 47 - 49 ]. The mechanisms driving glioma recurrence are complex and multifactorial, involving intrinsic and adaptive tumor cell resistance, tumor heterogeneity, and immune evasion [ 50 , 51 ].

Abbasi et al. represented that PSP happened in 36% of (95% confidence range, 33–40%) high-grade glioma patients and can occur in up to 20% of individuals following routine temozolomide chemoradiotherapy [ 52 ]. Through PSP, inflammation, edema, endothelial damage, abnormal capillary permeability, Blood Brain Barrier (BBB) disruption and oligodendroglia injury can cause a new or increased contrast enhanced lesion after chemoradiation (mostly during six months after treatment) [ 2 , 18 , 41 ]. The chance of developing diseases is raised within three months to years after the therapy and occurs mostly in MGMT-methylated tumors treated with TMZ [ 2 , 41 , 53 ].

Furthermore, among patients with malignant gliomas, RN is a severe radiotherapy-induced local tissue response. In most cases, it occurs within three to twelve months of radiation therapy, but it can occur years later as well [ 53 ]. RN is characterized by endothelial damage, severe neurotoxicity, the release of tumor necrosis factor-alpha (TNF-), damage to the BBB, glial damage, and deteriorating of edema that causes the emergence of new regions displaying abnormal improvement simulating true progression and recurrence [ 2 , 53 ]. Several pathological criteria classified RN from other glioma-post treatment circumstances, including histological examination, gliosis, edema, hyalinization, endothelial thickening, thrombosis, vessel occlusion, and fibrinoid deposition [ 2 , 54 ]. The majority of radiation necrosis occurs in regions that receive the highest radiation dose, typically near the tumor site and within the resected tumor surgical cavity [ 54 ]. On traditional MRI, it can be difficult to distinguish RN from tumor progression, as contrast-enhancing masses on T1-weighted contrast-enhanced imaging are often similar to those seen in tumor recurrence [ 2 , 55 ].

Radiotracers play a crucial role in diagnosing and treating gliomas. Traditional MRI is extremely difficult to use in routine clinical care or clinical trials to determine whether a tumor has responded, especially in pre-contrast T1-weighted images [ 56 ]. In order to manage brain tumors, Response Assessment in Neuro-Oncology group (RANO) recommended radiolabeled amino acid PET [ 57 ]. Among the various tracers in PET imaging for glioma, 18F-FET, 11C-MET, 18F-DOPA, and 11C-CHO are particularly notable, due to their ability to highlight various aspects of tumor biology and metabolism [ 13 , 17 , 58 , 59 ].

The results demonstrated the higher performance (higher AUC, sensitivity and specificity) of radiomics, ML, and DL models over conventional PET parameters-based models in distinguishing glioma treatment related changes [ 13 , 31 , 33 ]. This enhanced performance may be attributed to the ability of radiomics analysis to extract a diverse set of quantitative features from medical images, including texture, shape, and intensity characteristics. These comprehensive features can capture subtle changes in tissue characteristics that might not be apparent through traditional visual inspection or even conventional PET metrics including Standard Uptake Value (SUV) or TBR. These additional features can provide a more precise understanding of the tumor microenvironment and response to therapy, leading to improved diagnostic accuracy [ 13 , 60 ].

In addition, radiomics models use advanced mathematical algorithms to analyze the complex relationships within the extracted features. These algorithms can recognize patterns and correlations that are difficult for human observers to perceive [ 33 , 60 ]. Furthermore, advanced techniques, such as principal component analysis, recursive feature elimination, and logistic regression are applied in radiomics models to achieve higher predictive values and better distinguish treatment-related changes from tumor progression or recurrence [ 60 ].

FET, as a tyrosine analog, which is absorbed into the glioma cells via the L-type amino acid transporter, is used to monitor the amino acid transport capacity of brain tumors, related to tumor proliferation. The high sensitivity and specificity of FET PET in distinguishing glioma recurrence have been confirmed by several studies [ 61 , 62 ]. Sensitivity typically ranges from 70% to 90%, indicating the percentage of actual glioma recurrences correctly identified by the test/model [ 62 ]. In evaluating model performance, sensitivity refers to the True Positive Rate (TPR), which estimates the proportion of real positive cases that the model correctly identifies (i.e., accurate detection of glioma recurrence). The specificity of a model refers to its ability to correctly identify actual negative cases (i.e., accurate detection of non-recurrence). High values for both sensitivity and specificity indicate that FET PET performs exceptionally well in accurately detecting recurrences and excluding non-recurrences in patients with gliomas [ 62 ]. A high sensitivity is essential for minimizing false negatives, which could lead to missed diagnoses of recurrences. Meanwhile, FET PET has a specificity of 60% to 90%, indicating a higher proportion of true negative results. This is a measure of the percentage of glioma cases that are correctly identified as negative by the test or model. High specificity helps reduce false positives and thus avoids unnecessary invasive procedures or treatments [ 34 - 36 ]. Results showed high sensitivity and specificity in detecting glioma recurrence and making it a valuable tool for monitoring therapeutic response and diagnosing disease progression [ 21 , 62 - 65 ]. FET is the most widely used clinical and available tracer, and these results demonstrate its usefulness in detecting glioma recurrence accurately and precisely. This could explain why most of the studies reviewed used FET-PET models [ 23 , 31 , 32 , 34 - 36 ]. Paprottka et al. concluded that the random forest model, combining FET PET data with advanced MRI imaging techniques, assesses disease progression with a sensitivity of 91% and a specificity of 70% [ 35 ]. Kebir et al. also developed a linear discriminant model by using FET PET image radiomics information, achieving excellent detection of pseudoprogrossion in IDH-Wildtype glioblastoma with an AUC of 0.93 in comparison to conventional PET parameters model TBRmax and TBRmean with AUC 0.68 and 0.74, respectively [ 31 ].

Another amino acid radiotracer is [11C] MET, which is actively transported into glioma cells. Tumor amino acid uptake by MET reflects tumor cell proliferation and angiogenesis. MET PET imaging has proven useful in detecting glioma recurrence and guiding treatment decisions, showing the highest sensitivity (90%) and specificity (87%) among the available radiopharmaceuticals, demonstrated superior diagnostic power for recurrence detecting. As it displays an impressive 90% sensitivity, it can detect even the smallest signs of recurrence, preventing missed diagnoses and ensuring nearly all cases are identified. Furthermore, its exceptional specificity of 87% allows for the confident exclusion of non-recurrence cases, thereby enhancing diagnostic reliability. These remarkable attributes solidify MET’s position as a leading radiopharmaceutical for the precise evaluation of glioma recurrence [ 66 , 67 ]. A group of researchers demonstrated the high performance of a radiomic approach using a random forest classifier to distinguish recurrent tumors from RN with MET. Their results showed an area under the curve (AUC) of 0.98, along with a sensitivity of 0.90 and a specificity of 0.939 [ 38 ]. Random forest is an ensemble learning algorithm that generates multiple decision trees during the training phase and determines the final classification based on the majority vote derived from the predictions of individual trees [ 68 ]. L. Breiman developed the random forest model in 2001, and it has since become highly successful in both classification and regression tasks due to its robustness and ability to handle complex datasets [ 69 ]. Multiple decision trees are combined to make predictions by averaging in settings where the number of variables exceeds the number of observations [ 69 , 70 ]. It can also be adapted to a variety of ad hoc learning tasks and returns variable importance measures, which help identify the most influential features in the model’s predictions [ 70 ]. Its advantage over other models lies in its capability to effectively handle complex datasets, deliver high predictive accuracy, and incorporate an inherent mechanism for evaluating feature importance. Stems from its ability to handle complex datasets, provide high accuracy, and provide built-in feature importance evaluation [ 69 , 71 ].

DOPA serves as a precursor to dopamine and norepinephrine, neurotransmitters implicated in pain modulation and stress response. Its accumulation in glioma cells, indicates the presence of functional dopamine transporters, which are often overexpressed in high grade glioma including glioblastoma. Studies have indicated that DOPA PET can effectively differentiates glioma recurrence from treatment-related changes, and providing valuable information for patient management [ 10 ]. Recent studies reported that DOPA PET exhibits higher sensitivity and comparable specificity to FET PET in detecting glioma recurrence [ 10 , 61 , 72 ]. Notably, only two studies reported higher sensitivity with FET PET compared to DOPA PET. It is important to mention that the current evidence supporting comparative evaluation of these two imaging methods for assessing glioma recurrence is of very low quality. Therefore, further research is necessary to obtain more definitive and clinically meaningful results [ 10 , 61 , 72 ]. The selection of a specific radiotracer for PET imaging in glioma is determined by the particular tumor biological characteristics to be evaluated, including functional aspects of proliferation such as amino acid transport and glucose metabolism to neurotransmitter synthesis. Each radiotracer provides distinct insights into the pathophysiology of glioma, thereby assisting in diagnosis, staging, and monitoring of tumor progression and treatment response [ 73 , 74 ].

Ahrari et al. employed dynamic radiomics models and static DOPA PET including random forest, ElasticNet logistic regression (LR), and XGBoost to differentiate high-grade glioma progression from treatment-related changes. Their results represented AUCs of 0.834 for the best radiomics model (i.e., random forest) which slightly outperformed the reference tumor-to-background (TBR) with an AUC of 0.792 (P<0.001). Additionally, their findings revealed a lower AUC of 0.79 for the LR model compared to the other models [ 13 ].

Interpreting these findings within the context of specific studies and clinical scenarios is essential, considering several factors including the imaging protocols, treatment effects, and patient population characteristics. It is important to recognize that imaging protocols can influence the reproducibility, robustness, and generalizability of radiomics features and models. When most data are derived from one or a few centers with specific protocols, the model may not generalize effectively to data acquired using different protocols. Therefore, standardization or harmonization of imaging protocols is essential to minimize these effects and improve the clinical utility of PET based radiomics [ 75 ]. In addition, glioma characteristics vary with age, as older patients often have more aggressive tumors and distinct genetic mutations (e.g., IDH mutations) [ 6 ]. Consequently, radiomics features derived from PET images may reflect these biological differences, affecting model performance across different age groups. Moreover, variations in follow-up duration and the timing of early or late changes in imaging biomarkers may lead to misinterpreted if follow-up timing is not standardized, resulting in biased or incomplete training data for the models.

Another key application of radiomics models is in predicting tumor prognosis and patient survival. By analyzing radiomic features extracted from PET scans, researchers may be able to predict tumor prognosis and patient survival rates [ 36 ]. For instance, specific texture patterns or intensity distributions may be associated with tumor aggressiveness, thereby influencing survival time. Carles et al. demonstrated that radiomics texture features derived from FET PET images were most effective in predicting time-to-progression (P=0.001) in glioblastoma. They suggested that FET-PET radiomics could play a valuable role in prognostic evaluation and in identifying glioblastoma patients who may benefit from re-irradiation [ 34 ].

Radiomic models can also support risk classification, helping recognize glioma patients at high risk of recurrence or poor response to standard treatments thereby informing personalized treatment strategies [ 25 ]. Shahzadi et al. demonstrated that a 3D-DenseNet model, based on MET-PET integrated with age and MGMT status, achieved the highest performance in predicting overall survival and time to recurrence in glioblastoma patients, with Concordance Indices of 0.65 for overall survival and 0.68 for time to recurrence [ 40 ].

The present study has several limitations. The small number of studies included and their heterogeneity particularly in terms of patient demographics, follow-up durations, glioma subtypes, and imaging protocols, may limit the generalizability of the findings. Consequently, further investigations involving additional models and PET radiotracers are necessary to validate these results with greater accuracy. Moreover, selection bias may have influenced the studies, as patients unable to undergo surgery or chemoradiotherapy often due to advanced-stage cancer or significant comorbidities are frequently excluded. Additionally, confounding factors such as tumor grade, gender, race, and other comorbid conditions may affect treatment outcomes and radiomics model performance. Therefore, these variables should ideally be controlled in future analyses to improve the reliability of the findings.

Challenges and Future Directions

Although the potential benefits of PET based radiomic models in glioma management are significant, several challenges need to be addressed. The most important issue is the need for standardized protocols for image acquisition, feature extraction, and model development to ensure reproducibility in studies. Additionally, precise validation against clinical and pathological criteria for establishing the reliability and accuracy of these models, and integrating radiomic features with genomic, proteomic, and other clinical data can increase predictive power. Furthermore, overcoming barriers to clinical implementation, such as regulatory approval and integration into routine and standard clinical workflows, will be essential to realize the full potential of these models to personalized medicine in neuro-oncology, and paving the way for more effective therapeutic strategies [ 76 ].

Clinical Application

Integrating radiomics models with clinical protocols for managing glioma has the potential to revolutionize personalized care. Radiomics can extract high-dimensional imaging features from PET and other modalities, providing non-invasive biomarkers that complement conventional clinical factors. When incorporated into existing protocols, these models can help physicians as a diagnostic aid tool for tumor classification, guide treatment planning, and predict outcomes of glioma patients with greater precision before the treatment. Furthermore, radiomics-based models could assist in preoperative assessments, optimize radiotherapy dosing, and differentiate post-treatment changes from recurrence for each patient. However, successful integration requires rigorous validation, standardization of imaging protocols, and harmonization across clinical centers to ensure the models are reproducible, robust, and applicable in diverse healthcare settings.

Conclusion

This systematic review highlights the potential of PET-based radiomics as a promising approach for assessing post-treatment changes and predicting prognosis in glioma patients. The extraction and analysis of quantitative features from PET images, facilitated by machine learning (ML) and artificial intelligence (AI) models, demonstrate superior capability in distinguishing tumor progression or recurrence from treatment-related changes compared to conventional methods. Such advancements have great potential to enable personalized treatment strategies for improving overall survival and quality of life in glioma patients. Continued research and technological innovations are anticipated to overcome existing limitations, further enhancing the precision and clinical applicability of these tools for glioma management.

Acknowledgment

Special thanks are expressed to all those who have contributed in various ways to this work.

Authors’ Contribution

All authors contributed to the study’s conception and design. The idea for the article was for M. Shakeri. The literature search was performed by M. Shakeri, H. Ghadiri and SM. Hosseini, studies selection was done by A. Amraee, and L. Darvish, data abstraction and data analysis were performed by M. Shakeri, A. Amraee, F. Farkhondeh and A. Mostaar. The first draft of the manuscript was written by M. Shakeri and H. Ghadiri, and all authors commented on previous versions of the manuscript. All authors read and approved of the final manuscript.

Ethical Approval

This systematic review is a part of our study. Approval was granted by the Ethics Committee of the Tehran University of Medical Sciences with approval number: IR.TUMS.MEDICINE.REC.1400.1080.

Funding

This study was financially supported by Tehran University of Medical Sciences under Grant number 55747.

Conflict of Interest

None

References

1. Ostrom QT, Price M, Neff C, Cioffi G, Waite KA, Kruchko C, Barnholtz-Sloan JS. CBTRUS Statistical Report: Primary Brain and Other Central Nervous System Tumors Diagnosed in the United States in 2016-2020. Neuro Oncol. 2023; 25(12 Suppl 2):iv1-99. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
2. Alizadeh M, Broomand Lomer N, Azami M, Khalafi M, Shobeiri P, Arab Bafrani M, Sotoudeh H. Radiomics: The New Promise for Differentiating Progression, Recurrence, Pseudoprogression, and Radionecrosis in Glioma and Glioblastoma Multiforme. Cancers (Basel). 2023; 15(18):4429. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
3. Birzu C, French P, Caccese M, Cerretti G, Idbaih A, Zagonel V, Lombardi G. Recurrent Glioblastoma: From Molecular Landscape to New Treatment Perspectives. Cancers (Basel). 2020; 13(1):47. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
4. Brandes AA, Tosoni A, Franceschi E, Sotti G, Frezza G, Amistà P, et al. Recurrence pattern after temozolomide concomitant with and adjuvant to radiotherapy in newly diagnosed patients with glioblastoma: correlation With MGMT promoter methylation status. J Clin Oncol. 2009; 27(8):1275-9. DOI | PubMed
5. Drake LR, Hillmer AT, Cai Z. Approaches to PET Imaging of Glioblastoma. Molecules. 2020; 25(3):568. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
6. Kim M, Ladomersky E, Mozny A, Kocherginsky M, O’Shea K, Reinstein ZZ, et al. Glioblastoma as an age-related neurological disorder in adults. Neurooncol Adv. 2021; 3(1):1-13. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
7. Jovčevska I, Kočevar N, Komel R. Glioma and glioblastoma - how much do we (not) know? Mol Clin Oncol. 2013; 1(6):935-41. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
8. Tan AC, Ashley DM, López GY, Malinzak M, Friedman HS, Khasraw M. Management of glioblastoma: State of the art and future directions. CA Cancer J Clin. 2020; 70(4):299-312. DOI | PubMed
9. Rončević A, Koruga N, Soldo Koruga A, Rončević R, Rotim T, Šimundić T, et al. Personalized Treatment of Glioblastoma: Current State and Future Perspective. Biomedicines. 2023; 11(6):1579. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
10. Santo G, Laudicella R, Linguanti F, Nappi AG, Abenavoli E, Vergura V, et al. The Utility of Conventional Amino Acid PET Radiotracers in the Evaluation of Glioma Recurrence also in Comparison with MRI. Diagnostics (Basel). 2022; 12(4):844. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
11. Cui M, Zorrilla-Veloz RI, Hu J, Guan B, Ma X. Diagnostic Accuracy of PET for Differentiating True Glioma Progression From Post Treatment-Related Changes: A Systematic Review and Meta-Analysis. Front Neurol. 2021; 12:671867. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
12. Dhermain FG, Hau P, Lanfermann H, Jacobs AH, Van Den Bent MJ. Advanced MRI and PET imaging for assessment of treatment response in patients with gliomas. Lancet Neurol. 2010; 9(9):906-20. DOI | PubMed
13. Ahrari S, Zaragori T, Rozenblum L, Oster J, Imbert L, Kas A, Verger A. Relevance of Dynamic 18F-DOPA PET Radiomics for Differentiation of High-Grade Glioma Progression from Treatment-Related Changes. Biomedicines. 2021; 9(12):1924. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
14. Albert NL, Galldiks N, Ellingson BM, Van Den Bent MJ, Chang SM, Cicone F, et al. PET-based response assessment criteria for diffuse gliomas (PET RANO 1.0): a report of the RANO group. Lancet Oncol. 2024; 25(1):e29-41. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
15. Alongi P, Arnone A, Vultaggio V, Fraternali A, Versari A, Casali C, et al. Artificial Intelligence Analysis Using MRI and PET Imaging in Gliomas: A Narrative Review. Cancers (Basel). 2024; 16(2):407. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
16. Bailo M, Pecco N, Callea M, Scifo P, Gagliardi F, Presotto L, et al. Decoding the Heterogeneity of Malignant Gliomas by PET and MRI for Spatial Habitat Analysis of Hypoxia, Perfusion, and Diffusion Imaging: A Preliminary Study. Front Neurosci. 2022; 16:885291. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
17. Bonte S, Donche S, Henrotte M, Van Holen R, Goethals I. OS6. 3 Radiomics and machine learning on [18F] FET PET and T1ce MRI discriminate between low-grade and high-grade glioma. Neuro-Oncol. 2018; 20(suppl_3):iii226. DOI
18. Bhattacharya K, Rastogi S, Mahajan A. Post-treatment imaging of gliomas: challenging the existing dogmas. Clin Radiol. 2024; 79(3):e376-92. DOI | PubMed
19. D’Souza MM, Sharma R, Jaimini A, Panwar P, Saw S, Kaur P, et al. 11C-MET PET/CT and advanced MRI in the evaluation of tumor recurrence in high-grade gliomas. Clin Nucl Med. 2014; 39(9):791-8. DOI | PubMed
20. Eisazadeh R, Shahbazi-Akbari M, Mirshahvalad SA, Pirich C, Beheshti M. Application of Artificial Intelligence in Oncologic Molecular PET-Imaging: A Narrative Review on Beyond [18F]F-FDG Tracers Part II. [18F]F-FLT, [18F]F-FET, [11C]C-MET and Other Less-Commonly Used Radiotracers. Semin Nucl Med. 2024; 54(2):293-301. DOI | PubMed
21. Lohmeier J, Bohner G, Siebert E, Brenner W, Hamm B, Makowski MR. Quantitative biparametric analysis of hybrid 18F-FET PET/MR-neuroimaging for differentiation between treatment response and recurrent glioma. Sci Rep. 2019; 9(1):14603. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
22. Herrmann K, Czernin J, Cloughesy T, Lai A, Pomykala KL, Benz MR, et al. Comparison of visual and semiquantitative analysis of 18F-FDOPA-PET/CT for recurrence detection in glioblastoma patients. Neuro Oncol. 2014; 16(4):603-9. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
23. Maurer GD, Brucker DP, Stoffels G, Filipski K, Filss CP, Mottaghy FM, et al. 18F-FET PET Imaging in Differentiating Glioma Progression from Treatment-Related Changes: A Single-Center Experience. J Nucl Med. 2020; 61(4):505-11. DOI | PubMed
24. Bashir A, Mathilde Jacobsen S, Mølby Henriksen O, Broholm H, Urup T, Grunnet K, et al. Recurrent glioblastoma versus late posttreatment changes: diagnostic accuracy of O-(2-[18F]fluoroethyl)-L-tyrosine positron emission tomography (18F-FET PET). Neuro Oncol. 2019; 21(12):1595-606. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
25. Ahrari S, Zaragori T, Zinsz A, Oster J, Imbert L, Verger A. Application of PET imaging delta radiomics for predicting progression-free survival in rare high-grade glioma. Sci Rep. 2024; 14(1):3256. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
26. Barry N, Rowshanfarzad P, Francis RJ, Nowak AK, Ebert MA. Repeatability of image features extracted from FET PET in application to post-surgical glioblastoma assessment. Phys Eng Sci Med. 2021; 44(4):1131-40. DOI | PubMed
27. Booth TC, Larkin TJ, Yuan Y, Kettunen MI, Dawson SN, Scoffings D, et al. Analysis of heterogeneity in T2-weighted MR images can differentiate pseudoprogression from progression in glioblastoma. PLoS One. 2017; 12(5):e0176528. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
28. Ehret F, Kaul D, Clusmann H, Delev D, Kernbach JM. Machine Learning-Based Radiomics in Neuro-Oncology. Acta Neurochir Suppl. 2022; 134:139-51. DOI | PubMed
29. Jin W, Fatehi M, Abhishek K, Mallya M, Toyota B, Hamarneh G. Artificial intelligence in glioma imaging: challenges and advances. J Neural Eng. 2020; 17(2):021002. DOI | PubMed
30. Lambin P, Leijenaar RTH, Deist TM, Peerlings J, De Jong EEC, Van Timmeren J, et al. Radiomics: the bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol. 2017; 14(12):749-62. DOI | PubMed
31. Kebir S, Schmidt T, Weber M, Lazaridis L, Galldiks N, Langen KJ, et al. A Preliminary Study on Machine Learning-Based Evaluation of Static and Dynamic FET-PET for the Detection of Pseudoprogression in Patients with IDH-Wildtype Glioblastoma. Cancers (Basel). 2020; 12(11):3080. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
32. Lohmann P, Elahmadawy MA, Gutsche R, Werner JM, Bauer EK, Ceccon G, et al. FET PET Radiomics for Differentiating Pseudoprogression from Early Tumor Progression in Glioma Patients Post-Chemoradiation. Cancers (Basel). 2020; 12(12):3835. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
33. Müller M, Winz O, Gutsche R, Leijenaar RTH, Kocher M, Lerche C, et al. Static FET PET radiomics for the differentiation of treatment-related changes from glioma progression. J Neurooncol. 2022; 159(3):519-29. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
34. Carles M, Popp I, Starke MM, Mix M, Urbach H, Schimek-Jasch T, et al. FET-PET radiomics in recurrent glioblastoma: prognostic value for outcome after re-irradiation? Radiat Oncol. 2021; 16(1):46. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
35. Paprottka KJ, Kleiner S, Preibisch C, Kofler F, Schmidt-Graf F, Delbridge C, et al. Fully automated analysis combining [18F]-FET-PET and multiparametric MRI including DSC perfusion and APTw imaging: a promising tool for objective evaluation of glioma progression. Eur J Nucl Med Mol Imaging. 2021; 48(13):4445-55. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
36. Li Z, Holzgreve A, Unterrainer LM, Ruf VC, Quach S, Bartos LM, et al. Combination of pre-treatment dynamic [18F]FET PET radiomics and conventional clinical parameters for the survival stratification in patients with IDH-wildtype glioblastoma. Eur J Nucl Med Mol Imaging. 2023; 50(2):535-45. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
37. Lohmann P, Kocher M, Ruge MI, Visser-Vandewalle V, Shah NJ, Fink GR, et al. PET/MRI Radiomics in Patients With Brain Metastases. Front Neurol. 2020; 11:1. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
38. Hotta M, Minamimoto R, Miwa K. 11C-methionine-PET for differentiating recurrent brain tumor from radiation necrosis: radiomics approach with random forest classifier. Sci Rep. 2019; 9(1):15666. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
39. Mitamura K, Yamamoto Y, Kudomi N, Maeda Y, Norikane T, Miyake K, Nishiyama Y. Intratumoral heterogeneity of 18F-FLT uptake predicts proliferation and survival in patients with newly diagnosed gliomas. Ann Nucl Med. 2017; 31(1):46-52. DOI | PubMed
40. Shahzadi I, Seidlitz A, Beuthien-Baumann B, Zwanenburg A, Platzek I, Kotzerke J, et al. Radiomics for residual tumour detection and prognosis in newly diagnosed glioblastoma based on postoperative [11C] methionine PET and T1c-w MRI. Sci Rep. 2024; 14(1):4576. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
41. Ari AP, Akkurt BH, Musigmann M, Mammadov O, Blömer DA, Kasap DNG, et al. Pseudoprogression prediction in high grade primary CNS tumors by use of radiomics. Sci Rep. 2022; 12(1):5915. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
42. Jin S, Chen W, Guo X, Xing H, Yang H, Liu Q, et al. A prognostic model for overall survival in recurrent glioma patients treated with bevacizumab-containing therapy. Discov Oncol. 2024; 15(1):85. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
43. Teng C, Zhu Y, Li Y, Dai L, Pan Z, Wanggou S, Li X. Recurrence- and Malignant Progression-Associated Biomarkers in Low-Grade Gliomas and Their Roles in Immunotherapy. Front Immunol. 2022; 13:899710. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
44. García-Cabezas S, Rivin Del Campo E, Solivera-Vela J, Palacios-Eito A. Re-irradiation for high-grade gliomas: Has anything changed? World J Clin Oncol. 2021; 12(9):767-86. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
45. Kirkpatrick JP, Sampson JH. Recurrent malignant gliomas. Semin Radiat Oncol. 2014; 24(4):289-98. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
46. Hess CF, Schaaf JC, Kortmann RD, Schabet M, Bamberg M. Malignant glioma: patterns of failure following individually tailored limited volume irradiation. Radiother Oncol. 1994; 30(2):146-9. DOI | PubMed
47. McDonald MW, Shu HK, Curran WJ Jr, Crocker IR. Pattern of failure after limited margin radiotherapy and temozolomide for glioblastoma. Int J Radiat Oncol Biol Phys. 2011; 79(1):130-6. DOI | PubMed
48. Milano MT, Okunieff P, Donatello RS, Mohile NA, Sul J, Walter KA, Korones DN. Patterns and timing of recurrence after temozolomide-based chemoradiation for glioblastoma. Int J Radiat Oncol Biol Phys. 2010; 78(4):1147-55. DOI | PubMed
49. Pan H, Alksne J, Mundt AJ, Murphy KT, Cornell M, Kesari S, Lawson JD. Patterns of imaging failures in glioblastoma patients treated with chemoradiation: a retrospective study. Med Oncol. 2012; 29(3):2040-5. DOI | PubMed
50. Sharma P, Aaroe A, Liang J, Puduvalli VK. Tumor microenvironment in glioblastoma: Current and emerging concepts. Neurooncol Adv. 2023; 5(1):vdad009. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
51. Taylor KR, Barron T, Hui A, Spitzer A, Yalçin B, Ivec AE, et al. Glioma synapses recruit mechanisms of adaptive plasticity. Nature. 2023; 623(7986):366-74. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
52. Abbasi AW, Westerlaan HE, Holtman GA, Aden KM, Van Laar PJ, Van Der Hoorn A. Incidence of Tumour Progression and Pseudoprogression in High-Grade Gliomas: a Systematic Review and Meta-Analysis. Clin Neuroradiol. 2018; 28(3):401-11. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
53. Zikou A, Sioka C, Alexiou GA, Fotopoulos A, Voulgaris S, Argyropoulou MI. Radiation Necrosis, Pseudoprogression, Pseudoresponse, and Tumor Recurrence: Imaging Challenges for the Evaluation of Treated Gliomas. Contrast Media Mol Imaging. 2018; 2018:6828396. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
54. Brandsma D, Stalpers L, Taal W, Sminia P, Van Den Bent MJ. Clinical features, mechanisms, and management of pseudoprogression in malignant gliomas. Lancet Oncol. 2008; 9(5):453-61. DOI | PubMed
55. Reddy K, Westerly D, Chen C. MRI patterns of T1 enhancing radiation necrosis versus tumour recurrence in high-grade gliomas. J Med Imaging Radiat Oncol. 2013; 57(3):349-55. DOI | PubMed
56. Ellingson BM, Wen PY, Cloughesy TF. Modified Criteria for Radiographic Response Assessment in Glioblastoma Clinical Trials. Neurotherapeutics. 2017; 14(2):307-20. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
57. Saidijam M, Afshar S, Ahmad I, Patching S. Nucleoside transporters in PET imaging of proliferating cancer cells using 3ꞌ-deoxy-3ꞌ-[18F] fluoro-L-thymidine. Journal of Diagnostic Imaging in Therapy. 2018; 5(1):1-13. DOI
58. Castelo-Branco M, Moreira A. Combined PET/MRI in brain glioma imaging. In New Insights Into Glioblastoma. Academic Press; 2023. p. 155-65.
59. Evangelista L, Cuppari L, Bellu L, Bertin D, Caccese M, Reccia P, et al. Comparison Between 18F-Dopa and 18F-Fet PET/CT in Patients with Suspicious Recurrent High Grade Glioma: A Literature Review and Our Experience. Curr Radiopharm. 2019; 12(3):220-8. DOI | PubMed
60. Eertink JJ, Zwezerijnen GJC, Cysouw MCF, Wiegers SE, Pfaehler EAG, Lugtenburg PJ, et al. Comparing lesion and feature selections to predict progression in newly diagnosed DLBCL patients with FDG PET/CT radiomics features. Eur J Nucl Med Mol Imaging. 2022; 49(13):4642-51. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
61. Tang C, Ruan R, Xiong Z. Comparison between [18F] FET PET/MRI and [18F] FET PET/CT in the diagnosis of glioma recurrence: a systematic review and meta-analysis. Clin Transl Imaging. 2023; 11(5):479-91. DOI
62. Yu P, Wang Y, Su F, Chen Y. Comparing [18F]FET PET and [18F]FDOPA PET for glioma recurrence diagnosis: a systematic review and meta-analysis. Front Oncol. 2024; 13:1346951. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
63. Lohmann P, Lerche C, Stoffels G, Filss CP, Stegmayr C, Neumaier B, et al. P09. 26 FET PET radiomics-diagnosis of pseudoprogression in glioblastoma patients based on textural features. Neuro-Oncology. 2017; 19(suppl_3):iii75. DOI
64. Manzarbeitia-Arroba B, Hodolic M, Pichler R, Osipova O, Soriano-Castrejón ÁM, García-Vicente AM. 18F-Fluoroethyl-L Tyrosine Positron Emission Tomography Radiomics in the Differentiation of Treatment-Related Changes from Disease Progression in Patients with Glioblastoma. Cancers (Basel). 2023; 16(1):195. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
65. Wirsching HG, Roelcke U, Weller J, Hundsberger T, Hottinger AF, Von Moos R, et al. MRI and 18FET-PET Predict Survival Benefit from Bevacizumab Plus Radiotherapy in Patients with Isocitrate Dehydrogenase Wild-type Glioblastoma: Results from the Randomized ARTE Trial. Clin Cancer Res. 2021; 27(1):179-88. DOI | PubMed
66. Seifert R, Kersting D, Rischpler C, Opitz M, Kirchner J, Pabst KM, et al. Clinical Use of PET/MR in Oncology: An Update. Semin Nucl Med. 2022; 52(3):356-64. DOI | PubMed
67. Jadvar H, Connolly LP, Fahey FH, Shulkin BL. PET and PET/CT in pediatric oncology. Semin Nucl Med. 2007; 37(5):316-31. DOI | PubMed
68. Tabassum M, Suman AA, Suero Molina E, Pan E, Di Ieva A, Liu S. Radiomics and Machine Learning in Brain Tumors and Their Habitat: A Systematic Review. Cancers (Basel). 2023; 15(15):3845. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
69. Breiman L. Random forests. Machine Learning. 2001; 45:5-32. DOI
70. Rigatti SJ. Random Forest. J Insur Med. 2017; 47(1):31-9. DOI | PubMed
71. Sarica A, Cerasa A, Quattrone A. Random Forest Algorithm for the Classification of Neuroimaging Data in Alzheimer’s Disease: A Systematic Review. Front Aging Neurosci. 2017; 9:329. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
72. Galldiks N, Lohmann P, Cicone F, Langen KJ. FET and FDOPA PET imaging in glioma. In: Glioma Imaging: Physiologic, Metabolic, and Molecular Approaches. Cham: Springer International Publishing; 2019.
73. Muthukumar S, Darden J, Crowley J, Witcher M, Kiser J. A Comparison of PET Tracers in Recurrent High-Grade Gliomas: A Systematic Review. Int J Mol Sci. 2022; 24(1):408. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
74. Castello A, Castellani M, Florimonte L, Ciccariello G, Mansi L, Lopci E. PET radiotracers in glioma: a review of clinical indications and evidence. Clin Transl Imaging. 2022; 10(5):535-51. DOI
75. Gutsche R, Scheins J, Kocher M, Bousabarah K, Fink GR, Shah NJ, et al. Evaluation of FET PET Radiomics Feature Repeatability in Glioma Patients. Cancers (Basel). 2021; 13(4):647. Publisher Full Text | DOI | PubMed [ PMC Free Article ]
76. Jiang YQ, Gao Q, Chen H, Shi XX, Wu JB, Chen Y, et al. Positron Emission Tomography-Based Short-Term Efficacy Evaluation and Prediction in Patients With Non-Small Cell Lung Cancer Treated With Hypo-Fractionated Radiotherapy. Front Oncol. 2021; 11:590836. Publisher Full Text | DOI | PubMed [ PMC Free Article ]