You're using an outdated browser. Please upgrade to a modern browser for the best experience.
Glioblastoma Radiomics: Comparison
View Latest Version
Please note this is a comparison between Version 2 by Catherine Yang and Version 1 by Philip David Martin.

Glioblastoma (GBM) is a cancer with poor prognosis, its 5-year survival expectation is approximately 5%. Radiomics is a field of medical imaging analysis that focuses on the extraction of many quantitative imaging features related to shape, intensity and texture. These features are incorporated into models designed to predict important clinical or biological endpoints for patients. Glioblastoma multiforme (GBM) patients stand to benefit from this emerging research field as radiomics has the potential to assess the biological heterogeneity of the tumour, which contributes significantly to the inefficacy of current standard of care therapy.

  • radiomics
  • Glioblastoma
  • machine learning
  • MRI

1. Introduction

Glioblastoma multiforme (GBM) is the most common primary brain malignancy with an incidence of 3.1 cases per 100,000 adults in the United States [1]. Despite a multimodal treatment regime of maximal safe resection [2] and adjuvant chemoradiation therapy, the prognosis is poor for GBM patients. Median overall survival is approximately 15 months, and the 5-year survival rate is 5.1% [3].
The current standard of care for GBM consists of maximal safe resection followed by concurrent Temozolomide (TMZ, 75 mg/m2/day) and radiation therapy (60 Gy over 30 fractions delivered over six weeks). In elderly patients, it has been found that there is no significant difference in outcomes between patients treated with hypofractionated radiotherapy (40 Gy in 15 fractions over three weeks) and conventionally fractionated radiotherapy [4,5][4][5]. Recent developments in treatment include the use of low intensity, intermediate frequency electric fields. These Tumour Treating Fields (TTFs) have been found to arrest cell division and cause death in tumour cells [4]. Prognosis is limited by tumour progression, which occurs in over 80% of GBM patients, demonstrating the inefficacy of current treatment methods. Multiple factors have been cited as contributing to treatment failure [6]. These factors include intra-tumoural heterogeneity, referring to the physiological variations between different tumour regions, and inter-tumoural heterogeneity, referring to the physiological differences between tumours from different patients. This heterogeneity contributes to common treatment failure as homogeneous treatment prescriptions are ineffective in GBM treatment [7]. Treatment failures in GBM can be attributed to the tumour’s ability to dynamically adapt and develop mechanisms of treatment resistance. Resistance to chemotherapy arises from the presences of the blood–brain barrier (BBB) preventing chemotherapy drugs from reaching tumour cells and from inherent molecular resistance to these drugs [6]. Resistance to radiotherapy arises from tumour hypoxia reducing the treatment efficacy [8]. Additionally, peritumoural infiltration of tumour cells in the brain parenchyma is another important hallmark of GBM contributing to treatment failure and local relapse [9].
Magnetic Resonance Imaging (MRI) is the primary imaging technique used for the diagnosis, treatment planning and monitoring of GBM [10]. This is due to the excellent soft tissue contrast offered by MRI and the ability to change acquisition parameters to enhance contrast between normal and diseased tissue. Anatomical MRI sequences routinely acquired for the clinical management of GBM include T1-weighted (T1-w), T1-weighted contrast enhanced (T1CE), T2-weighted (T2-w) and T2 fluid attenuated inversion recovery (T2-FLAIR) imaging [11]. Gadolinium-based contrast agents, which are administered during T1CE imaging acquisition, enable improved delineation of the tumour boundaries. However, as selective accumulation in the tumour tissue relies on the extravasation of the contrast agent into the extracellular extravascular space, this technique is only useful for the delineated tumour regions with a disrupted BBB [9], leaving infiltrating tumour regions with an intact BBB invisible. The contrast-enhancing tumour volume delineated from T1CE imaging is normally used to define the gross tumour volume for planning radiotherapy. As such, this volume is often an underestimation of the ‘entire tumour burden’ [12]. FLAIR imaging is also utilised in delineating GBM tumour volumes on MRI. Hyperintense T2-FLAIR volumes are used in the delineation of tumour radiotherapy volumes, especially in defining areas of peritumoural cancer infiltration in oedema [13]; however, this has limitations as the prognostic value of FLAIR signal has not been clearly shown in GBM tumour response assessment [13]. Radiomics could help provide a method to quantify the extent of tumour infiltration beyond the contrast enhancing boundaries [12[12][14],14], potentially enabling the adaptation of treatment plans tailored to the characteristics of each patient’s individual tumour [15]. Functional imaging approaches such as quantitative MRI (qMRI) and Positron Emission Tomography (PET) have also been increasingly utilised in the management of GBM as they can provide information about the tumour physiological processes, including vascularisation, cellularity, hypoxia and metabolism [10,16][10][16].
Radiomics is a method of data analysis in which many quantitative features are extracted, and machine learning (ML) methods are used to establish correlations with patient clinical outcomes [17,18][17][18]. The underlying hypothesis of radiomics is that radiomic features reflect the biological characteristics of the tissue, which in turn are related to patient clinical outcomes. While previous studies have demonstrated that radiomics have the potential to predict clinical outcomes in GBM [19,20,21[19][20][21][22][23],22,23], its application is still far from clinical implementation [24,25][24][25]. Radiomics models for GBM are generally based on MRI due to its central role in the management of brain cancer patient; however, radiomics approaches using CT and PET imaging have also been investigated [26,27][26][27].
Radiomic models can be assessed based on three key factors: performance is defined by the accuracy of the model’s predictions; clinical utility determines whether the predictions can be used clinically to modify treatment and improve patient outcomes and transferability, which represents the ability of the model to maintain performance when applied to datasets with varied acquisition parameters [24,28][24][28].

2.Research in GBM Radiomics

Radiomics models have performed well predicting clinical and biological characteristics of tumours. The performance of these models has continued to improve as more data have become publicly available and innovative artificial intelligence methods have been utilised in radiomic analysis.

2.1. Potential Applications of Radiomics in GBM Patient Management

To date, clinical assessments of patient status have been primarily based on qualitative metrics [37][29]. This presents two main issues. First, basing clinical decisions on qualitative metrics introduces observer bias to treatment. Second, the qualitative metrics used are not sufficient to describe the functional characteristics of the tumour. Radiomics could help to address these issues by providing a quantitative, unbiased method of assessing tumour physiology [37][29]. Tasks in the management of GBM patients that radiomics are suited for including diagnosis, prognostication, stratification, response assessment and treatment planning. In the context of diagnosis, radiomics could reduce the need for invasive biopsies through the automatic detection and grading of brain tumours [47,55][30][31]. The ‘gold standard’ for the grading of gliomas is biopsy [56][32]. Interestingly, a recent study by Kobayashi et al. [39][33] was able to achieve an accuracy of 0.90 ± 0.03 using a fully automated radiomics model for distinguishing between high and low grade glioma. This was based on structural MRI sequences from the Brain Tumour Segmentation (BraTS) challenge [11,45,57][11][34][35]. The use of radiomics for these purposes could reduce the need for invasive surgical procedures as well as reducing observer bias. Prognostication predicts patient clinical outcomes and can be completed to identify if a patient would benefit from a more aggressive treatment regimen. Studies attempting to predict patient overall survival have generally performed well when incorporating advanced methods such as ML analysis or when incorporating multi-parametric imaging [19]. A prognostic task that radiomics could be useful for in GBM patient management is the identification of O-6-methylguanine-DNA methyltransferase (MGMT) promoter methylation status. MGMT is a repair protein which inversely correlates to patient survival [58][36]. MGMT methylation is both a prognostic and predictive marker, conferring an improved survival and better response to TMZ treatment [59][37]. Studies aiming to determine the MGMT methylation status [60,61][38][39] have seen some success using radiomic features extracted from structural MRI sequences. Improvements on these predictions and treatment implementation could help clinicians identify patients that may benefit from TMZ chemotherapy. Stratification classifies patients based on important clinical factors and can help determine if a patient may benefit from certain treatments. This is often completed in conjunction with prognostication. For example, a study by Kickingereder et al. [62][40] demonstrated the utility of radiomics in identifying patients who may benefit from anti-angiogenic therapies. This study was able to achieve an area under the receiver operator curve (AUC) of 0.792 for prediction of overall survival for patients treated with bevacizumab. Treatment response assessments in GBM clinical trials often rely on the updated Response Assessment in Neuro-Oncology (RANO) criteria [63][41]. These are based on changes in tumour volume and contrast enhancement observed on T1CE images. Radiomics could also be used to measure functional tumour changes occurring over time in response to treatment [14]. The ability of radiomics to quantify underlying tissue physiology makes it suitable for assessing tumour treatment response [18]. Diffusion-weighted imaging has been investigated as a potential predictor of early treatment response for GBM [64][42]. Recent studies implementing radiomics have shown to be able to differentiate between true tumour progression and pseudo-progression with a sensitivity of approximately 80% [20,27][20][27]. The combination of qMRI sequences and radiomics analyses has the potential to generate predictive models of tumour response, which could help clinicians make timely decisions for treatment adaptation [20]. Treatment planning is another promising application of radiomics in GBM management. For instance, radiotherapy stands to benefit from the integration of radiomics into clinical practice [15,65,66][15][43][44]. Radiomics may assist in the segmentation of radiotherapy target volumes and in measuring and predicting treatment response [67][45]. Identification of peritumoural infiltration is also a task which can improve the quality of surgery and radiotherapy by providing a means to better define tumour margins, and is well suited to be performed by radiomics [12,66][12][44]. Tumour hypoxia is a factor which reduces the efficacy of radiotherapy as DNA damage is reduced in hypoxic regions. Radiomics has shown potential in the identification of hypoxic tumour regions [8], which could be leveraged for dose escalation treatment strategies.

2.2. Existing Models in GBM Radiomics

To date, the best performing models have incorporated functional imaging and/or ML methods to improve predictions. A selection of radiomics models in GBM has been outlined below. These models have been selected as they address one of the key factors for the progress of radiomics towards clinical implementation (being clinical utility, performance and transferability). Table 21 reports recent studies with a focus on radiomics in GBM.
Table 21.
 A selection of modern radiomics models for predicting clinical and biological factors in GBM.
Author Model Description Conclusions Clinical Application Performance Patient Numbers
Kickingereder et al., 2016 [62][40] Stratification of patients into groups who were likely or not likely to benefit from anti-angiogenic therapies A radiomics model based on supervised principal component analysis is effective at stratifying patients into groups that can benefit from the addition of anti-angiogenic therapy Identification of which patients may benefit from certain therapies provides clinicians a convenient to tailor treatment regimens to individuals AUC = 0.792 172
Lao et al., 2017 [38][46] Deep features were extracted using transfer learning and implemented into a survival prediction model. This model utilised learned features, handcrafted radiomics features and clinical factors to produce a prediction of overall patient survival. Implementing learned features into a predictive radiomics model can improve the performance of a predictive model. A survival prediction model can be used to determine if a patient would benefit from a more aggressive treatment regimen. Improving performance by implementing learned features and clinical factors can build confidence in the model. AUC = 0.739 112
Shboul et al., 2019 [23] A fully automated segmentation pipeline using Deep Neural Networks was developed using the BraTS challenge dataset. Survival prediction was then performed using radiomic features extracted from this dataset. A fully automated framework for the delineation of GBM and patient survival prediction can be useful to reduce clinical workload and bias in the tasks of segmentation and survival prediction. A framework such as this can be used to provide a perform a tumour segmentation for the purpose of radiotherapy treatment planning. Survival predictions can be used to recommend a more or less aggressive treatment regimen as required Leave one out cross validation accuracy = 0.73 396 total
Park et al., 2020 [19] Survival Prediction based on T1 Post Contrast, T2 FLAIR and DSC MRI as well as clinical factors. By incorporating mpMRI as well as clinical factors, it is possible to achieve a high performing survival prediction model An accurate prediction of survival period can provide a quantitative measure of the severity of the disease. AUC = 0.74 216
Yan et al., 2020 [12] Identification of peritumoural invasive regions in GBM based on Structural, Perfusion-weighted and Diffusion-weighted MRI. Convolutional Neural Network was used along with radiomics to identify regions of peritumoural infiltration Lower intensity on Diffusion-weighted MRI and higher intensity on T1, FLAIR and Perfusion-weighted MRI was observed in peritumoural invasion areas. Identification of regions of peritumoural invasion will allow treatment plans to accurately target whole tumour volumes and improve local control. Accuracy = 78.5% 57
Suter et al., 2020 [36][47] Feature robustness was tested and models developed on single centre data were applied to multicentre data. In addition to this, a model developed using robust features on single centre data was tested on multicentre data. A large performance drop was found when models trained on single centre data were applied to multicentre data. This performance drop could be reduced when the model was restricted to robust features. Model transferability is an important factor in radiomic research. To develop transferable radiomics models, it will be necessary to develop models on multi-centre data and identify reproducible radiomic features. AUC reduced by 0.56 for single centre model tested on multicentre data 63 single centre patients, 76 multicentre data
Shim et al., 2021 [22] Prediction of recurrence pattern based on DSC MRI radiomics and neural networks, model produced to predict local and distant recurrence Quantitative measures of tumour perfusion can accurately predict recurrence patterns of tumour recurrence Identifying the likely course of tumour progression could enable early intervention or treatment plan adaptation. AUC = 0.969; AUC = 0.864 (local and distant) 192
DSC = Dynamic Susceptibility Contrast, mpMRI = multi-parametric MRI.
These models highlight how radiomics can build confidence to progress towards clinical implementation. Each model has its own clinical utility. Survival prediction models [19,23,38][19][23][46] can be used to determine if a patient may benefit from a more aggressive treatment regimen, the prediction of peritumoural infiltration [12] and the prediction of local vs. distant recurrence [22] could allow for improved treatment planning and stratification studies [62][40] can help determine if a patient can benefit from a certain treatment regimen. Innovative methods such as deep learning and mpMRI were implemented in these models to improve their performance. Suter et al. [36][47] investigated the repeatability of radiomic features and the performance of a model when applied to images acquired with different acquisition protocols to the training dataset. This study was an excellent example of a method that could be used to measure and improve transferability of radiomics models. One problem hindering the clinical translation of radiomic models is the lack of multi-centre data being utilised for training and technical validation. By training and validating a model on single-centre data, the issue of inter-centre variability is largely neglected. Inter-centre variability refers to variability in image intensity due to differences in scanner or acquisition protocols, which has been found to have a significant impact on MRI and MRI-derived radiomic features [51,68][48][49]. Technical validation studies aim to investigate the performance of models when applied to data not drawn from the same dataset used for training [69][50]. Some of these studies have been completed to date [25,36][25][47]; however, further studies will be required so that reproducibility of model results can be assured.

2.3. Challenges of Developing a Radiomics Model for Brain Cancer

For any model to reach clinical implementation, it needs to fulfill several criteria. It should provide valuable information that can improve the clinical workflow; it should produce accurate predictions of underlying physiology/clinical outcomes (clinical/biological validation); it should yield reproducible predictions on images acquired from different centres (technical validation); and it should be statistically sound, meaning that it has been trained on a large amount of diverse data [37][29]. A greater number of patients in the training cohort produces a more accurate model; thus, there is no upper limit on the number of patients required for the development of a radiomics model. As a guideline for a statistically robust dataset, there should be at least 10 patients per radiomic feature implemented in the model [70][51]. Maintaining both transferability and performance is a challenge when constructing any radiomics model. Transferability can be achieved in some capacity by restricting a model to features that are stable when acquired with different parameters [36][47]. It can be difficult to maintain performance of a model while restricted to these features. Additionally, difficulties can arise in radiomics due to the lack of confidence around the reproducibility of radiomic features. A lack of reproducibility in features will mean that radiomic models will be inaccurate in creating predictions [32][52]. Challenges arise in constructing radiomics models for brain cancer in establishing a biological ‘ground truth’ [25]. As radiomics models are often designed to predict biological characteristics of tissue, it is necessary to obtain the biological data for training these models. A biopsy is considered the gold standard for assessing biological function of tissue. However, using a biopsy in conjunction with radiomics is non-trivial for brain cancer. Challenges include that whole brain resection cannot be performed to investigate the biology of the whole brain; accurate image-guided biopsies are difficult to perform due to brain shifts on craniotomy [71][53]; challenging tumour location in critical areas of the brain prevent the performance of biopsies; and fragility of brain tissue compromises the retention of the tissue structural integrity. These factors make it difficult to assess the accuracy of radiomics models via image-guided biopsy. This creates challenges in assessing intratumoural heterogeneity, due to the difficulty in identifying biopsy location and retaining spatial information of the sampled tumour tissue. A study aiming to achieve spatial accuracy in GBM biopsy had a mean error of 1.5 ± 1.1 mm; however, this required MRI to be completed during the biopsy procedure [71][53]. Challenges in retaining spatial registration between the biopsy location and the region of interest marked on the image make it difficult to build radiomic models predicting biological heterogeneity with high accuracy. The alternative to biopsy is for an expert radiologist to label a tumour based on available imaging; however, this method may suffer from poor accuracy in comparison to biopsy. GBM also poses the challenge of having tumour infiltration that extends beyond the visible margins of contrast enhancement. Tumour infiltration is not visible on conventional MRI due to the presence of an intact BBB preventing the contrast agent from entering the tumour extracellular extravascular space. As such, accurate delineation of the tumour boundaries beyond the extent of contrast enhancement on T1CE images is extremely challenging, and clinicians are forced to rely on margins of uncertainty in planning treatments, which ultimately contributes to suboptimal treatment efficacy [72][54]. This presents both a challenge and an opportunity for radiomic research. Radiomic research can suffer from not having the full extent of the tumour delineated for feature extraction; however, it has also shown some promise in identifying regions of peritumoural infiltration by integrating machine learning methods into the analysis pipeline [12].

References

  1. Maher, E.A.; Bachoo, R.M. Glioblastoma. In Rosenberg’s Molecular and Genetic Basis of Neurological and Psychiatric Disease, 5th ed.; Academic Press: Cambridge, MA, USA, 2015; pp. 909–917. ISBN 9780124105294.
  2. Stupp, R.; Mason, W.P.; van den Bent, M.J.; Weller, M.; Fisher, B.; Taphoorn, M.J.; Belanger, K.; Brandes, A.A.; Marosi, C.; Bogdahn, U.; et al. Radiotherapy plus Concomitant and Adjuvant Temozolomide for Glioblastoma. N. Engl. J. Med. 2005, 352, 987–996.
  3. Australian Institute of Health and Welfare. Brain and Other Central Nervous System Cancers; Australian Institute of Health and Welfare: Canberra, Australia, 2017; p. 108.
  4. Fernandes, C.; Costa, A.; Osório, L.; Lago, R.C.; Linhares, P.; Carvalho, B.; Caeiro, C. Current Standards of Care in Glioblastoma Therapy. Glioblastoma 2017, 197–241.
  5. Roa, W.; Brasher, P.M.; Bauman, G.; Anthes, M.; Bruera, E.; Chan, A.; Fisher, B.; Fulton, D.; Gulavita, S.; Hao, C.; et al. Abbreviated course of radiation therapy in older patients with glioblastoma multiforme: A prospective randomized clinical trial. J. Clin. Oncol. Off. J. Am. Soc. Clin. Oncol. 2004, 22, 1583–1588.
  6. Ou, A.; Yung, W.A.; Majd, N. Molecular mechanisms of drug resistance in glioblastoma. Int. J. Mol. Sci. 2021, 22, 6385.
  7. Bergmann, N.; Delbridge, C.; Gempt, J.; Feuchtinger, A.; Walch, A.; Schirmer, L.; Bunk, W.; Aschenbrenner, T.; Liesche-Starnecker, F.; Schlegel, J. The Intratumoral Heterogeneity Reflects the Intertumoral Subtypes of Glioblastoma Multiforme: A Regional Immunohistochemistry Analysis. Front. Oncol. 2020, 10, 494.
  8. Beig, N.; Patel, J.; Prasanna, P.; Hill, V.; Gupta, A.; Correa, R.; Bera, K.; Singh, S.; Partovi, S.; Varadan, V.; et al. Radiogenomic analysis of hypoxia pathway is predictive of overall survival in Glioblastoma. Sci. Rep. 2018, 8, 7.
  9. Yan, J.L.; Li, C.; Boonzaier, N.R.; Fountain, D.M.; Larkin, T.J.; Matys, T.; van der Hoorn, A.; Price, S.J. Multimodal MRI characteristics of the glioblastoma infiltration beyond contrast enhancement. Ther. Adv. Neurol. Disord. 2019, 12, 1756286419844664.
  10. Shukla, G.; Alexander, G.S.; Bakas, S.; Nikam, R.; Talekar, K.; Palmer, J.D.; Shi, W. Advanced magnetic resonance imaging in glioblastoma: A review. Chin. Clin. Oncol. 2017, 6, 40.
  11. Menze, B.H.; Jakab, A.; Bauer, S.; Kalpathy-cramer, J.; Farahani, K.; Kirby, J.; Burren, Y.; Porz, N.; Slotboom, J.; Wiest, R.; et al. The multimodal brain tumor image segmentation benchmark (BRATS). IEEE Trans. Med. Imaging 2015, 34, 1993–2024.
  12. Yan, J.L.; Li, C.; Hoorn, A.v.d.; Boonzaier, N.R.; Matys, T.; Price, S.J. A Neural Network Approach to Identify the Peritumoral Invasive Areas in Glioblastoma Patients by Using MR Radiomics. Sci. Rep. 2020, 10, 9748.
  13. Garrett, M.D.; Yanagihara, T.K.; Yeh, R.; McKhann, G.M.; Sisti, M.B.; Bruce, J.N.; Sheth, S.A.; Sonabend, A.M.; Wang, T.J. Monitoring Radiation Treatment Effects in Glioblastoma: FLAIR Volume as Significant Predictor of Survival. Tomography 2017, 3, 131.
  14. Guo, L.; Wang, G.; Feng, Y.; Yu, T.; Guo, Y.; Bai, X.; Ye, Z. Diffusion and perfusion weighted magnetic resonance imaging for tumor volume definition in radiotherapy of brain tumors. Radiat. Oncol. 2016, 11, 123.
  15. Gardin, I.; Grégoire, V.; Gibon, D.; Kirisli, H.; Pasquier, D.; Thariat, J.; Vera, P. Radiomics: Principles and radiotherapy applications. Crit. Rev. Oncol./Hematol. 2019, 138, 44–50.
  16. Shaikh, F.; Dupont-Roettger, D.; Dehmeshki, J.; Awan, O.; Kubassova, O.; Bisdas, S. The Role of Imaging Biomarkers Derived From Advanced Imaging and Radiomics in the Management of Brain Tumors. Front. Oncol. 2020, 10, 1–6.
  17. Caruso, D.; Polici, M.; Zerunian, M.; Pucciarelli, F.; Guido, G.; Polidori, T.; Landolfi, F.; Nicolai, M.; Lucertini, E.; Tarallo, M.; et al. Radiomics in Oncology, Part 1: Technical Principles and Gastrointestinal Application in CT and MRI. Cancers 2021, 13, 2522.
  18. Caruso, D.; Polici, M.; Zerunian, M.; Pucciarelli, F.; Guido, G.; Polidori, T.; Landolfi, F.; Nicolai, M.; Lucertini, E.; Tarallo, M.; et al. Radiomics in Oncology, Part 2: Thoracic, Genito-Urinary, Breast, Neurological, Hematologic and Musculoskeletal Applications. Cancers 2021, 13, 2681.
  19. Park, J.E.; Kim, H.S.; Jo, Y.; Yoo, R.E.; Choi, S.H.; Nam, S.J.; Kim, J.H. Radiomics prognostication model in glioblastoma using diffusion- and perfusion-weighted MRI. Sci. Rep. 2020, 10, 4250.
  20. Patel, M.; Zhan, J.; Natarajan, K.; Flintham, R.; Davies, N.; Sanghera, P.; Grist, J.; Duddalwar, V.; Peet, A.; Sawlani, V. Machine learning-based radiomic evaluation of treatment response prediction in glioblastoma. Clin. Radiol. 2021, 76, 17–628.
  21. Yan, J.L.; Toh, C.H.; Ko, L.; Wei, K.C.; Chen, P.Y. A Neural Network Approach to Identify Glioblastoma Progression Phenotype from Multimodal MRI. Cancers 2021, 13, 2006.
  22. Shim, K.Y.; Chung, S.W.; Jeong, J.H.; Hwang, I.; Park, C.K.; Kim, T.M.; Park, S.H.; Won, J.K.; Lee, J.H.; Lee, S.T.; et al. Radiomics-based neural network predicts recurrence patterns in glioblastoma using dynamic susceptibility contrast-enhanced MRI. Sci. Rep. 2021, 11, 9974.
  23. Shboul, Z.A.; Alam, M.; Vidyaratne, L.; Pei, L.; Elbakary, M.I.; Iftekharuddin, K.M. Feature-Guided Deep Radiomics for Glioblastoma Patient Survival Prediction. Front. Neurosci. 2019, 13, 966.
  24. O’Connor, J.P.; Aboagye, E.O.; Adams, J.E.; Aerts, H.J.; Barrington, S.F.; Beer, A.J.; Boellaard, R.; Bohndiek, S.E.; Brady, M.; Brown, G.; et al. Imaging biomarker roadmap for cancer studies. Nat. Rev. Clin. Oncol. 2017, 14, 169.
  25. Hu, L.S.; Swanson, K.R. Roadmap for the clinical integration of radiomics in neuro-oncology. Neuro-Oncology 2020, 22, 743–745.
  26. Carles, M.; Popp, I.; Starke, M.M.; Mix, M.; Urbach, H.; Schimek-Jasch, T.; Eckert, F.; Niyazi, M.; Baltas, D.; Grosu, A.L. FET-PET radiomics in recurrent glioblastoma: Prognostic value for outcome after re-irradiation? Radiat. Oncol. 2021, 16, 46.
  27. Lohmann, P.; Elahmadawy, M.A.; Gutsche, R.; Werner, J.M.; Bauer, E.K.; Ceccon, G.; Kocher, M.; Lerche, C.W.; Rapp, M.; Fink, G.R.; et al. FET PET Radiomics for Differentiating Pseudoprogression from Early Tumor Progression in Glioma Patients Post-Chemoradiation. Cancers 2020, 12, 3835.
  28. Fournier, L.; Costaridou, L.; Bidaut, L.; Michoux, N.; Lecouvet, F.E.; de Geus-Oei, L.F.; Boellaard, R.; Oprea-Lager, D.E.; Obuchowski, N.A.; Caroli, A.; et al. Incorporating radiomics into clinical trials: Expert consensus endorsed by the European Society of Radiology on considerations for data-driven compared to biologically driven quantitative biomarkers. Eur. Radiol. 2021, 31, 6001–6012.
  29. Chaddad, A.; Kucharczyk, M.J.; Daniel, P.; Sabri, S.; Jean-Claude, B.J.; Niazi, T.; Abdulkarim, B. Radiomics in Glioblastoma: Current Status and Challenges Facing Clinical Implementation. Front. Oncol. 2019, 9, 374.
  30. Cho, H.H.; Lee, S.H.; Kim, J.; Park, H. Classification of the glioma grading using radiomics analysis. PeerJ 2018, 2018, e5982.
  31. Choi, S.W.; Cho, H.H.; Koo, H.; Cho, K.R.; Nenning, K.H.; Langs, G.; Furtner, J.; Baumann, B.; Woehrer, A.; Cho, H.J.; et al. Multi-Habitat Radiomics Unravels Distinct Phenotypic Subtypes of Glioblastoma with Clinical and Genomic Significance. Cancers 2020, 12, 1707.
  32. Komori, T. Grading of adult diffuse gliomas according to the 2021 WHO Classification of Tumors of the Central Nervous System. Lab. Investig. 2021, 102, 126–133.
  33. Kobayashi, K.; Miyake, M.; Takahashi, M.; Hamamoto, R. Observing deep radiomics for the classification of glioma grades. Sci. Rep. 2021, 11, 10942.
  34. Bakas, S.; Reyes, M.; Jakab, A.; Bauer, S.; Rempfler, M.; Crimi, A.; Shinohara, R.T.; Berger, C.; Ha, S.M.; Rozycki, M.; et al. Identifying the best machine learning algorithms for brain tumor segmentation, progression assessment, and overall survival prediction in the BRATS challenge. arXiv 2018, arXiv:1811.02629.
  35. Bakas, S.; Akbari, H.; Sotiras, A.; Bilello, M.; Rozycki, M.; Kirby, J.S.; Freymann, J.B.; Farahani, K.; Davatzikos, C. Advancing The Cancer Genome Atlas glioma MRI collections with expert segmentation labels and radiomic features. Sci. Data 2017, 4, 170117.
  36. Rivera, A.L.; Pelloski, C.E.; Gilbert, M.R.; Colman, H.; De La Cruz, C.; Sulman, E.P.; Bekele, B.N.; Aldape, K.D. MGMT promoter methylation is predictive of response to radiotherapy and prognostic in the absence of adjuvant alkylating chemotherapy for glioblastoma. Neuro-Oncology 2010, 12, 116.
  37. Hegi, M.E.; Diserens, A.C.; Gorlia, T.; Hamou, M.F.; de Tribolet, N.; Weller, M.; Kros, J.M.; Hainfellner, J.A.; Mason, W.; Mariani, L.; et al. MGMT Gene Silencing and Benefit from Temozolomide in Glioblastoma. N. Engl. J. Med. 2005, 352, 997–1003.
  38. Sasaki, T.; Kinoshita, M.; Fujita, K.; Fukai, J.; Hayashi, N.; Uematsu, Y.; Okita, Y.; Nonaka, M.; Moriuchi, S.; Uda, T.; et al. Radiomics and MGMT promoter methylation for prognostication of newly diagnosed glioblastoma. Sci. Rep. 2019, 9, 14435.
  39. Vils, A.; Bogowicz, M.; Tanadini-Lang, S.; Vuong, D.; Saltybaeva, N.; Kraft, J.; Wirsching, H.G.; Gramatzki, D.; Wick, W.; Rushing, E.; et al. Radiomic Analysis to Predict Outcome in Recurrent Glioblastoma Based on Multi-Center MR Imaging from the Prospective DIRECTOR Trial. Front. Oncol. 2021, 11, 636672.
  40. Kickingereder, P.; Götz, M.; Muschelli, J.; Wick, A.; Neuberger, U.; Shinohara, R.T.; Sill, M.; Nowosielski, M.; Schlemmer, H.P.; Radbruch, A.; et al. Large-scale Radiomic Profiling of Recurrent Glioblastoma Identifies an Imaging Predictor for Stratifying Anti-Angiogenic Treatment Response. Clin. Cancer Res. 2016, 22, 5765–5771.
  41. Wen, P.Y.; Chang, S.M.; Van Den Bent, M.J.; Vogelbaum, M.A.; Macdonald, D.R.; Lee, E.Q. Response assessment in neuro-oncology clinical trials. J. Clin. Oncol. 2017, 35, 2439–2449.
  42. Hamstra, D.A.; Galbán, C.J.; Meyer, C.R.; Johnson, T.D.; Sundgren, P.C.; Tsien, C.; Lawrence, T.S.; Junck, L.; Ross, D.J.; Rehemtulla, A.; et al. Functional diffusion map as an early imaging biomarker for high-grade glioma: Correlation with conventional radiologic response and overall survival. J. Clin. Oncol. 2008, 26, 3387–3394.
  43. Qiu, Q.; Duan, J.; Yin, Y. Radiomics in radiotherapy: Applications and future challenges. Precis. Radiat. Oncol. 2020, 4, 29–33.
  44. Sloan, A.E.; Shukla, G.; Rathore, S.; Akbari, H.; Gondi, V.; Davatzikos, C. Radiomics-based identification of peritumoral infiltration in de novo glioblastoma imaging presents targets amenable for potential targeted extended resection: A neurosurgical survey. J. Clin. Oncol. 2019, 37, e13573.
  45. Forghani, R.; Gupta, R.; Maleki, F.; Chiu, Y.; Yen, Y. Efficient Radiomics-Based Classification of Multi-Parametric MR Images to Identify Volumetric Habitats and Signatures in Glioblastoma: A Machine Learning Approach. Cancers 2022, 14, 1475.
  46. Lao, J.; Chen, Y.; Li, Z.C.; Li, Q.; Zhang, J.; Liu, J.; Zhai, G. A Deep Learning-Based Radiomics Model for Prediction of Survival in Glioblastoma Multiforme. Sci. Rep. 2017, 7, 10353.
  47. Suter, Y.; Knecht, U.; Alão, M.; Valenzuela, W.; Hewer, E.; Schucht, P.; Wiest, R.; Reyes, M. Radiomics for glioblastoma survival analysis in pre-operative MRI: Exploring feature robustness, class boundaries, and machine learning techniques. Cancer Imaging 2020, 20, 55.
  48. Carré, A.; Klausner, G.; Edjlali, M.; Lerousseau, M.; Briend-Diop, J.; Sun, R.; Ammari, S.; Reuzé, S.; Alvarez Andres, E.; Estienne, T.; et al. Standardization of brain MR images across machines and protocols: Bridging the gap for MRI-based radiomics. Sci. Rep. 2020, 10, 12340.
  49. Um, H.; Tixier, F.; Bermudez, D.; Deasy, J.O.; Young, R.J.; Veeraraghavan, H. Impact of image preprocessing on the scanner dependence of multi-parametric MRI radiomic features and covariate shift in multi-institutional glioblastoma datasets. Phys. Med. Biol. 2019, 64, 165011.
  50. Park, J.E.; Kim, H.S.; Kim, D.; Park, S.Y.; Kim, J.Y.; Cho, S.J.; Kim, J.H. A systematic review reporting quality of radiomics research in neuro-oncology: Toward clinical utility and quality improvement using high-dimensional imaging features. BMC Cancer 2020, 20, 29.
  51. Riley, R.D.; Snell, K.I.; Ensor, J.; Burke, D.L.; Harrell, F.E.; Moons, K.G.; Collins, G.S. Minimum sample size for developing a multivariable prediction model: PART II—Binary and time-to-event outcomes. Stat. Med. 2019, 38, 1276–1296.
  52. Hoebel, K.V.; Patel, J.B.; Beers, A.L.; Chang, K.; Singh, P.; Brown, J.M.; Pinho, M.C.; Batchelor, T.T.; Gerstner, E.R.; Rosen, B.R.; et al. Radiomics repeatability pitfalls in a scan-rescan mri study of glioblastoma. Radiol. Artif. Intell. 2021, 3, e190199.
  53. Mohyeldin, A.; Lonser, R.R.; Elder, J.B. Real-time magnetic resonance imaging-guided frameless stereotactic brain biopsy: Technical note. J. Neurosurg. 2016, 124, 1039–1046.
  54. Rathore, S.; Akbari, H.; Doshi, J. Radiomic signature of infiltration in peritumoral edema predicts subsequent recurrence in glioblastoma: Implications for personalized radiotherapy planning. J. Med. Imaging 2018, 5, 1.
More
© Text is available under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/)
Academic Video Service
Feedback