pith. machine review for the scientific record. sign in

arxiv: 2512.22331 · v4 · submitted 2025-12-26 · 💻 cs.CV · cs.AI

The Multi-View Paradigm Shift in MRI Radiomics: Predicting MGMT Methylation in Glioblastoma

Mariya Miteva , Maria Nisheva-Pavlova This is my paper

Pith reviewed 2026-05-16 18:57 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords radiomicsglioblastomaMGMT methylationmulti-view VAEMRIvariational autoencoderradiogenomicsnecrotc core
0
0 comments X

The pith

A multi-view variational autoencoder on T1Gd and FLAIR radiomics predicts MGMT methylation status in glioblastoma tumors with an AUC of 0.77.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces a multi-view latent representation learning framework based on variational autoencoders to integrate radiomic features from two MRI modalities while preserving modality-specific structure. It claims this late-fusion approach in a probabilistic latent space reduces redundancy that limits conventional unimodal or early-fusion radiomics models. A reader would care because MGMT promoter methylation status guides prognosis and treatment choices in glioblastoma, and non-invasive imaging-based prediction could reduce reliance on invasive biopsies. The method is applied specifically to features from the necrotic tumor core in post-contrast T1-weighted and FLAIR scans. On held-out test data the combined multi-view VAE plus random forest model reaches an AUC of 0.77, exceeding both a plain radiomics baseline and a hyperparameter-tuned single-model version.

Core claim

The central claim is that the proposed multi-view VAE combined with a random forest classifier achieves a test AUC of 0.77 for predicting MGMT promoter methylation from radiomic features extracted from the necrotic tumor core in T1Gd and FLAIR MRI, substantially outperforming both a baseline radiomics model (AUC 0.54) and a hyperparameter-tuned model (AUC 0.64). The framework enables late fusion by encoding each modality into its own probabilistic latent distribution before combining them, thereby capturing complementary information without forcing early concatenation of high-dimensional feature sets.

What carries the argument

Multi-view variational autoencoder that encodes each MRI modality into a separate probabilistic latent distribution before late fusion, preserving modality-specific radiomic structure.

If this is right

  • Late fusion in a shared probabilistic latent space integrates complementary modality information more effectively than early concatenation or single-modality models.
  • Reduced feature redundancy allows the downstream classifier to focus on predictive signals rather than duplicated measurements across T1Gd and FLAIR.
  • The resulting AUC improvement supports more accurate non-invasive inference of MGMT methylation status, which carries direct prognostic and therapeutic weight in glioblastoma.
  • The same multi-view encoding strategy can be reused for other radiogenomic targets once paired modality-specific radiomic features are available.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Extending the framework to additional MRI sequences such as T2 or diffusion-weighted images could further enrich the latent representation without retraining the entire pipeline from scratch.
  • The probabilistic latent codes naturally provide uncertainty estimates around each prediction, which could be used to flag low-confidence cases for biopsy confirmation.
  • Similar multi-view latent fusion may improve prediction of other glioblastoma molecular markers such as IDH mutation or EGFR amplification when corresponding radiomic feature sets are extracted.
  • Replacing handcrafted radiomics with end-to-end learned features inside the same multi-view VAE structure could test whether the performance gain is tied to the specific feature extraction step or to the fusion mechanism itself.

Load-bearing premise

Radiomic features from the necrotic tumor core in T1Gd and FLAIR MRI contain complementary non-redundant information about MGMT methylation status that the multi-view VAE latent space can capture effectively.

What would settle it

An independent validation cohort where the multi-view VAE plus random forest model fails to exceed the performance of a single-view or early-fusion radiomics model on the same necrotic-core features would falsify the central performance claim.

Figures

Figures reproduced from arXiv: 2512.22331 by Maria Nisheva-Pavlova, Mariya Miteva.

Figure 1
Figure 1. Figure 1: Stepwise improvement in test-set AUC across baseline, tuned radiomics, and multi-view VAE-based model. Impact of Multi-View Latent Representation Learning We observe a substantially larger improvement in discriminative performance when classification is performed on the latent representations learned by the proposed multi-view VAE. Training a RF classifier on the fused 12-dimensional latent space resulted … view at source ↗
Figure 3
Figure 3. Figure 3: Two-dimensional projections of the 12-dimensional latent space for the Baseline, Tuned, and MultiViewVAE models. Points are colored by predicted probability of the positive class (purple = low, green/yellow = high), with smoothed probability contours (levels 0.3, 0.5, and 0.7). From left to right, the latent space exhibits a progressive shift toward higher predicted probabilities, consistent with improved … view at source ↗
read the original abstract

Non-invasive inference of molecular tumor characteristics from medical imaging is a central goal of radiogenomics, particularly in glioblastoma (GBM), where O6-methylguanine-DNA methyltransferase (MGMT) promoter methylation carries important prognostic and therapeutic significance. Although radiomics-based machine learning methods have shown promise for this task, conventional unimodal and early-fusion approaches are often limited by high feature redundancy and incomplete modeling of modality-specific information. In this work, we introduce a multi-view latent representation learning framework based on variational autoencoders (VAE) that preserves modality-specific radiomic structure while enabling late fusion in a compact probabilistic latent space. The approach is evaluated on radiomic features extracted from the necrotic tumor core in post-contrast T1-weighted (T1Gd) and Fluid-Attenuated Inversion Re-covery (FLAIR) Magnetic Resonance Imaging (MRI). Experimental results demonstrate that the proposed multi-view VAE combined with a random forest classifier achieves a test Area Under the Receiver Operating Characteristic (ROC) Curve (AUC) of 0.77 (95% confidence interval: 0.71-0.83), substantially outperforming both a baseline radiomics model (AUC = 0.54) and a hyperparameter-tuned model (AUC = 0.64). These findings indicate that multi-view probabilistic encoding enables more effective integration of complementary MRI information and significantly improves predictive performance for MGMT promoter methylation status.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript proposes a multi-view variational autoencoder (VAE) framework to learn latent representations from radiomic features extracted from the necrotic tumor core in T1Gd and FLAIR MRI sequences, enabling late fusion for predicting MGMT promoter methylation status in glioblastoma. It reports that a random forest classifier trained on these representations achieves a test AUC of 0.77 (95% CI 0.71-0.83), outperforming a baseline radiomics model (AUC 0.54) and a hyperparameter-tuned model (AUC 0.64).

Significance. If the performance gains are reproducible, the work would demonstrate that multi-view probabilistic encoding can integrate complementary modality-specific radiomic information more effectively than unimodal or early-fusion baselines, supporting advances in non-invasive radiogenomics for GBM. The explicit reporting of a 95% CI and two distinct baselines is a strength that allows direct assessment of the claimed lift; however, the modest absolute AUC and lack of supporting experimental details limit the immediate clinical or methodological impact.

major comments (3)
  1. [Experimental results] Experimental results and methods sections: the manuscript provides no dataset size, patient cohort details, MGMT class distribution, or train/test split ratios, which are load-bearing for interpreting whether the AUC 0.77 and its outperformance over the 0.64 baseline reflect genuine generalization or are artifacts of small-sample or imbalanced data.
  2. [Methods] Methods section: no description of the cross-validation strategy, feature selection procedure, or safeguards against data leakage between VAE training on radiomic features and downstream random forest evaluation is given, leaving the central empirical claim only partially supported.
  3. [Results] Results section: while the 95% CI is reported, the number of test samples and the exact procedure used to compute the interval (e.g., bootstrap) are omitted, preventing assessment of whether the improvement from 0.64 to 0.77 is statistically meaningful.
minor comments (2)
  1. [Abstract] Abstract: the phrase 'hyperparameter-tuned model' is underspecified; clarify whether this baseline uses the same single-view VAE architecture or a different radiomics pipeline.
  2. [Results] The weakest assumption (complementary non-redundant information captured in the necrotic-core latent space) is directly tested by the AUC comparison, but the manuscript would benefit from an ablation showing performance when each modality is used alone.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments, which highlight important aspects of transparency and reproducibility. We have revised the manuscript to incorporate the requested details on the dataset, experimental procedures, and statistical methods, thereby strengthening the support for our central claims.

read point-by-point responses

  1. Referee: [Experimental results] Experimental results and methods sections: the manuscript provides no dataset size, patient cohort details, MGMT class distribution, or train/test split ratios, which are load-bearing for interpreting whether the AUC 0.77 and its outperformance over the 0.64 baseline reflect genuine generalization or are artifacts of small-sample or imbalanced data.

    Authors: We agree that these details are necessary for interpreting the results. In the revised manuscript we have added a dedicated 'Dataset Description' subsection in Methods that reports the total cohort size, patient demographics, MGMT methylation class balance, and the exact train/test split ratios used. These additions allow direct assessment of whether the observed performance lift reflects genuine generalization. revision: yes

  2. Referee: [Methods] Methods section: no description of the cross-validation strategy, feature selection procedure, or safeguards against data leakage between VAE training on radiomic features and downstream random forest evaluation is given, leaving the central empirical claim only partially supported.

    Authors: We accept that the original Methods section lacked sufficient procedural detail. The revised version now explicitly describes the cross-validation strategy (5-fold CV performed exclusively on the training partition), the feature selection approach (applied only to training data), and the data-leakage safeguards (VAE training and latent representation learning confined to the training set, with the random forest evaluated on a completely held-out test set). These clarifications fully support the reported empirical claims. revision: yes

  3. Referee: [Results] Results section: while the 95% CI is reported, the number of test samples and the exact procedure used to compute the interval (e.g., bootstrap) are omitted, preventing assessment of whether the improvement from 0.64 to 0.77 is statistically meaningful.

    Authors: We have updated the Results section to state the exact number of test samples and to specify that the 95% CI was obtained via bootstrap resampling of the test predictions. We have also added a statistical comparison (DeLong test) between the AUCs to quantify the significance of the observed improvement. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper introduces a multi-view VAE for integrating radiomic features from T1Gd and FLAIR MRI sequences in the necrotic core, followed by random forest classification for MGMT methylation prediction. The reported result is an empirical test AUC of 0.77 (95% CI 0.71-0.83) outperforming baselines at 0.54 and 0.64. No derivation chain, equations, or parameter fittings are described that reduce the AUC metric to a fitted input by construction. The approach relies on standard VAE latent encoding and late fusion without self-definitional loops, uniqueness theorems, or load-bearing self-citations. The performance comparison is externally falsifiable on held-out data and does not collapse to renaming or ansatz smuggling. The framework is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard variational autoencoder assumptions plus domain assumptions about radiomic features; no new physical entities are postulated.

free parameters (2)
  • VAE latent dimension and regularization weights
    Hyperparameters chosen to balance reconstruction fidelity and modality-specific preservation in the multi-view model.
  • Random forest hyperparameters
    Tuned parameters for the downstream classifier.
axioms (2)
  • domain assumption Radiomic features from necrotic tumor core in T1Gd and FLAIR capture complementary biological signals relevant to MGMT methylation
    Invoked when selecting the necrotic core as the region of interest for feature extraction.
  • domain assumption Variational autoencoder can learn a compact latent space that preserves modality-specific structure while enabling effective late fusion
    Core modeling assumption of the multi-view framework.

pith-pipeline@v0.9.0 · 5562 in / 1530 out tokens · 24267 ms · 2026-05-16T18:57:35.000621+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

37 extracted references · 37 canonical work pages · 3 internal anchors

  1. [1]

    A Survey on Multi-view Learning

    Xu, Chang, Tao Dacheng, Chao Xu. A Survey on Multi-view Learning. arXiv:1304.5634 (cs), 2013. https://arxiv.org/abs/1304.5634

    work page internal anchor Pith review Pith/arXiv arXiv 2013

  2. [2]

    Popular Ensemble Methods: An Empirical Study

    Opitz, D.; Maclin, R. (1999). "Popular ensemble methods: An empirical study". Journal of Artificial Intelligence Research. 11: 169–198. arXiv:1106.0257. doi:10.1613/jair.614

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1613/jair.614 1999

  3. [3]

    Ensemble based systems in decision making

    Polikar, R. (2006). "Ensemble based systems in decision making". IEEE Circuits and Systems Magazine. 6 (3): 21–45. doi:10.1109/MCAS.2006.1688199. S2CID 18032543

    work page doi:10.1109/mcas.2006.1688199 2006

  4. [4]

    Ensemble-based classifiers

    Rokach, L. (2010). "Ensemble-based classifiers". Artificial Intelligence Review. 33 (1–2): 1–39. doi:10.1007/s10462-009-9124-7. hdl:11323/1748. S2CID 11149239

    work page doi:10.1007/s10462-009-9124-7 2010

  5. [5]

    Multimodal Machine Learning: A Survey and Taxonomy,

    T. Baltrušaitis, C. Ahuja and L. -P . Morency, "Multimodal Machine Learning: A Survey and Taxonomy," in IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 41, no. 2, pp. 423-443, 1 Feb. 2019, doi: 10.1109/TPAMI.2018.2798607

    work page doi:10.1109/tpami.2018.2798607 2019

  6. [6]

    Zheng, Lecheng & Cheng, Yu & He, Jingrui. (2019). Deep Multimodality Model for Multi-task Multi-view Learning. 10.48550/arXiv.1901.08723

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1901.08723 2019

  7. [7]

    P ., & Reyes, M

    Bauer, S., Wiest, R., Nolte, L. P ., & Reyes, M. (2013). A survey of MRI-based medical image analysis for brain tumor studies. Physics in Medicine & Biology, 58(13), R97–R129. https://doi.org/10.1088/0031-9155/58/13/R97

    work page doi:10.1088/0031-9155/58/13/r97 2013

  8. [8]

    Yihao Li, Mostafa El Habib Daho, Pierre-Henri Conze, Rachid Zeghlache, Hugo Le Boité, Ramin Tadayoni, Béatrice Cochener, Mathieu Lamard, Gwenolé Quellec, A review of deep learning-based information fusion techniques for multimodal medical image classification, Computers in Biology and Medicine, V olume 177, 2024, 108635, ISSN 0010-4825, https://doi.org/10...

    work page doi:10.1016/j.compbiomed.2024.108635 2024

  9. [9]

    Warner, Elisa & Lee, Joonsang & Hsu, William & Syeda-Mahmood, Tanveer & Jr, Charles & Gevaert, Olivier & Rao, Arvind. (2024). Multimodal Machine Learning in Image-Based and Clinical Biomedicine: Survey and Prospects. International Journal of Computer Vision

    work page 2024

  10. [10]

    10.1007/s11263-024-02032-8

    1-17. 10.1007/s11263-024-02032-8

    work page doi:10.1007/s11263-024-02032-8

  11. [11]

    Advancements in Medical Radiology Through Multimodal Machine Learning: A Comprehensive Overview

    Haq IU, Mhamed M, Al-Harbi M, Osman H, Hamd ZY , Liu Z. Advancements in Medical Radiology Through Multimodal Machine Learning: A Comprehensive Overview. Bioengineering (Basel). 2025 Apr 30;12(5):477. doi: 10.3390/bioengineering12050477. PMID: 40428096; PMCID: PMC12108733

    work page doi:10.3390/bioengineering12050477 2025

  12. [12]

    Felix Krones, Umar Marikkar, Guy Parsons, Adam Szmul, Adam Mahdi, Review of multimodal machine learning approaches in healthcare, Information Fusion, V olume 114, 2025, 102690, ISSN 1566-2535, https://doi.org/10.1016/j.inffus.2024.102690

    work page doi:10.1016/j.inffus.2024.102690 2025

  13. [13]

    Liu, Z., Duan, T., Zhang, Y . et al. Radiogenomics: a key component of precision cancer medicine. Br J Cancer 129, 741–753 (2023). https://doi.org/10.1038/s41416-023-02317-8

    work page doi:10.1038/s41416-023-02317-8 2023

  14. [14]

    Story, Michael & Durante, Marco. (2018). Radiogenomics. Medical Physics. 45. e1111- e1122. 10.1002/mp.13064

    work page doi:10.1002/mp.13064 2018

  15. [15]

    Oldrini, B., Vaquero-Siguero, N., Mu, Q. et al. MGMT genomic rearrangements contribute to chemotherapy resistance in gliomas. Nat Commun 11, 3883 (2020). https://doi.org/10.1038/s41467-020-17717-0

    work page doi:10.1038/s41467-020-17717-0 2020

  16. [16]

    V ., & Lucke-Wold, B

    Shah, S., Nag, A., Sachithanandam, S. V ., & Lucke-Wold, B. (2024). Predictive and Prognostic Significance of Molecular Biomarkers in Glioblastoma. Biomedicines, 12(12),

    work page 2024

  17. [17]

    https://doi.org/10.3390/biomedicines12122664

    work page doi:10.3390/biomedicines12122664

  18. [18]

    The 2021 WHO Classification of Tumors of the Central Nervous System: a summary

    Louis DN, Perry A, Wesseling P, Brat DJ, Cree IA, Figarella-Branger D, Hawkins C, Ng HK, Pfister SM, Reifenberger G, Soffietti R, von Deimling A, Ellison DW. The 2021 WHO Classification of Tumors of the Central Nervous System: a summary. Neuro Oncol. 2021 Aug 2;23(8):1231-1251. doi: 10.1093/neuonc/noab106. PMID: 34185076; PMCID: PMC8328013

    work page doi:10.1093/neuonc/noab106 2021

  19. [19]

    Saeed, Numan & Hardan, Shahad & Abutalip, Kudaibergen & Yaqub, Mohammad. (2022). Is it Possible to Predict MGMT Promoter Methylation from Brain Tumor MRI Scans using Deep Learning Models?. Proceedings of Machine Learning Research 172:1–14, 2022

    work page 2022

  20. [20]

    Mariya Miteva, Maria Nisheva-Pavlova, The power of integrating multiple data sources in medical imaging: A study of MGMT methylation status, Procedia Computer Science, V olume 239, 2024, Pages 1196-1203, ISSN 1877-0509, https://doi.org/10.1016/j.procs.2024.06.287

    work page doi:10.1016/j.procs.2024.06.287 2024

  21. [21]

    Radiomics and AI-Based Prediction of MGMTMethylation Status in Glioblastoma Using Multiparametric MRI: A Hybrid FeatureWeighting Approach

    Tasci, E.; Zhuge, Y .; Zhang, L.; Ning, H.; Cheng, J.Y .; Miller, R.W.; Camphausen, K.; Krauze, A.V . Radiomics and AI-Based Prediction of MGMTMethylation Status in Glioblastoma Using Multiparametric MRI: A Hybrid FeatureWeighting Approach. Diagnostics 2025, 15, 1292. https://doi.org/10.3390/diagnostics15101292

    work page doi:10.3390/diagnostics15101292 2025

  22. [22]

    Deep learning classification of MGMT status of glioblastomas using multiparametric MRI with a novel domain knowledge augmented mask fusion approach

    Koska İÖ, Koska Ç. Deep learning classification of MGMT status of glioblastomas using multiparametric MRI with a novel domain knowledge augmented mask fusion approach. Sci Rep. 2025 Jan 25;15(1):3273. doi: 10.1038/s41598-025-87803-0. PMID: 39863759; PMCID: PMC11762293

    work page doi:10.1038/s41598-025-87803-0 2025

  23. [23]

    Wilkinson, M., Dumontier, M., Aalbersberg, I. et al. The FAIR Guiding Principles for scientific data management and stewardship. Sci Data 3, 160018 (2016). https://doi.org/10.1038/sdata.2016.18

    work page doi:10.1038/sdata.2016.18 2016

  24. [24]

    Bakas, S., Sako, C., Akbari, H. et al. The University of Pennsylvania glioblastoma (UPenn-GBM) cohort: advanced MRI, clinical, genomics, & radiomics. Sci Data 9, 453 (2022). https://doi.org/10.1038/s41597-022-01560-7

    work page doi:10.1038/s41597-022-01560-7 2022

  25. [25]

    Chung, Relax and focus on brain tumor segmentation, Medical Image Analysis, V olume 75, 2022, 102259, ISSN 1361-8415, https://doi.org/10.1016/j.media.2021.102259

    Pei Wang, Albert C.S. Chung, Relax and focus on brain tumor segmentation, Medical Image Analysis, V olume 75, 2022, 102259, ISSN 1361-8415, https://doi.org/10.1016/j.media.2021.102259

    work page doi:10.1016/j.media.2021.102259 2022

  26. [26]

    Enhanced Segmentation of Glioma Subregions via Modality-Aware Encoding and Channel-Wise Attention in Multimodal MRI

    Cariola A, Sibilano E, Brunetti A, Buongiorno D, Guerriero A, Bevilacqua V . Enhanced Segmentation of Glioma Subregions via Modality-Aware Encoding and Channel-Wise Attention in Multimodal MRI. Applied Sciences. 2025; 15(14):8061. https://doi.org/10.3390/app15148061

    work page doi:10.3390/app15148061 2025

  27. [27]

    van Timmeren, J., Cester, D., Tanadini-Lang, S. et al. Radiomics in medical imaging— “how-to” guide and critical reflection. Insights Imaging 11, 91 (2020). https://doi.org/10.1186/s13244-020-00887-2

    work page doi:10.1186/s13244-020-00887-2 2020

  28. [28]

    Li, L., Xiao, F., Wang, S. et al. Preoperative prediction of MGMT promoter methylation in glioblastoma based on multiregional and multi-sequence MRI radiomics analysis. Sci Rep 14, 16031 (2024). https://doi.org/10.1038/s41598-024-66653-2

    work page doi:10.1038/s41598-024-66653-2 2024

  29. [29]

    Feldman L (2024) Hypoxia within the glioblastoma tumor microenvironment: a master saboteur of novel treatments. Front. Immunol. 15:1384249. doi: 10.3389/fimmu.2024.1384249

    work page doi:10.3389/fimmu.2024.1384249 2024

  30. [30]

    The Image Biomarker Standardization Initiative: Standardized Quantitative Radiomics for High-Throughput Image-based Phenotyping

    Zwanenburg A, Vallières M, Abdalah MA, et al. The Image Biomarker Standardization Initiative: Standardized Quantitative Radiomics for High-Throughput Image-based Phenotyping. Radiology. 2020;295(2):328-338. doi:10.1148/radiol.2020191145

    work page doi:10.1148/radiol.2020191145 2020

  31. [31]

    Zwanenburg, Alex & Leger, Stefan & Vallières, Martin & Löck, Steffen & Initiative, for. (2016). Image biomarker standardisation initiative - feature definitions. 10.48550/arXiv.1612.07003

    work page doi:10.48550/arxiv.1612.07003 2016

  32. [32]

    Computational Radiomics System to Decode the Radiographic Phenotype

    van Griethuysen JJM, Fedorov A, Parmar C, et al. Computational Radiomics System to Decode the Radiographic Phenotype. Cancer Res. 2017;77(21):e104-e107. doi:10.1158/0008- 5472.CAN-17-0339

    work page doi:10.1158/0008- 2017

  33. [33]

    Conventional and advanced magnetic resonance imaging in patients with high-grade glioma

    Pope WB, Brandal G. Conventional and advanced magnetic resonance imaging in patients with high-grade glioma. Q J Nucl Med Mol Imaging. 2018;62(3):239-253. doi:10.23736/S1824-4785.18.03086-8

    work page doi:10.23736/s1824-4785.18.03086-8 2018

  34. [34]

    Consensus recommendations for a standardized Brain Tumor Imaging Protocol in clinical trials

    Ellingson BM, Bendszus M, Boxerman J, et al. Consensus recommendations for a standardized Brain Tumor Imaging Protocol in clinical trials. Neuro Oncol. 2015;17(9):1188-

    work page 2015

  35. [35]

    doi:10.1093/neuonc/nov095

    work page doi:10.1093/neuonc/nov095

  36. [36]

    Chen, Yankun & Liu, Jingxuan & Peng, Lingyun & Wu, Yiqi & Xu, Yige & Zhang, Zhanhao. (2024). Auto-Encoding Variational Bayes. Cambridge Explorations in Arts and Sciences. 2. 10.61603/ceas.v2i1.33

    work page doi:10.61603/ceas.v2i1.33 2024

  37. [37]

    Jimenez Rezende, Danilo & Mohamed, Shakir & Wierstra, Daan. (2014). Stochastic Backpropagation and Approximate Inference in Deep Generative Models. Proceedings of the 31st International Conference on Machine Learning, PMLR 32(2):1278-1286, 2014

    work page 2014