pith. sign in

arxiv: 1907.08303 · v1 · pith:OHRGFMB6new · submitted 2019-07-18 · 📡 eess.IV · cs.CV

Fully-automated deep learning-powered system for DCE-MRI analysis of brain tumors

Jakub Nalepa , Pablo Ribalta Lorenzo , Michal Marcinkiewicz , Barbara Bobek-Billewicz , Pawel Wawrzyniak , Maksym Walczak , Michal Kawulok , Wojciech Dudzik
show 3 more authors
Grzegorz Mrukwa Pawel Ulrych Michael P. Hayball
This is my paper

Pith reviewed 2026-05-24 19:12 UTC · model grok-4.3

classification 📡 eess.IV cs.CV
keywords DCE-MRIbrain tumorsdeep learningimage segmentationpharmacokinetic modelingvascular input functionautomated analysislow-grade glioma
0
0 comments X

The pith

A deep learning system fully automates DCE-MRI brain tumor analysis with state-of-the-art accuracy in under three minutes on one GPU.

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

The paper shows that deep learning can replace manual steps in extracting quantitative biomarkers from dynamic contrast-enhanced MRI of brain tumors. It builds an end-to-end pipeline that segments tumors and fits contrast concentration curves, introducing a cubic model for the vascular input function that lowers fitting error and a fast method to locate the input region. Validation covers standard benchmarks plus scans from 44 low-grade glioma patients, with the whole study processed in less than three minutes. If correct, quantitative DCE-MRI information becomes available without expert time or operator variability.

Core claim

Our deep learning-powered technique does not require any user interaction, it yields reproducible results, and it is rigorously validated against benchmark (BraTS'17 for tumor segmentation, and a test dataset released by the Quantitative Imaging Biomarkers Alliance for the contrast-concentration fitting) and clinical (44 low-grade glioma patients) data. Also, we introduce a cubic model of the vascular input function used for pharmacokinetic modeling which significantly decreases the fitting error when compared with the state of the art, alongside a real-time algorithm for determination of the vascular input region. An extensive experimental study, backed up with statistical tests, showed our

What carries the argument

End-to-end deep learning pipeline that performs tumor segmentation, real-time vascular input region detection, and pharmacokinetic modeling with a cubic vascular input function.

If this is right

  • Quantitative contrast-concentration fitting and tumor segmentation become available without manual contouring or parameter tuning.
  • An entire DCE-MRI study finishes in less than three minutes on one GPU instead of the longer times required by manual methods.
  • The cubic vascular input function model produces lower fitting error than earlier models on the same data.
  • Results stay consistent across public benchmarks and the clinical low-grade glioma cohort.

Where Pith is reading between the lines

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

  • The same pipeline could be retrained on data from other tumor grades or body sites to test whether automation generalizes.
  • Embedding the system into scanner consoles might allow immediate biomarker maps during the scan session.
  • Combining the output maps with diffusion or perfusion sequences could tighten diagnostic criteria for tumor grading.

Load-bearing premise

Models tuned on BraTS'17 and the QIBA test set will keep the same segmentation accuracy and fitting performance on the separate set of 44 low-grade glioma patients.

What would settle it

Performance on a new independent collection of DCE-MRI studies from different scanners or patient groups falls below the reported segmentation and fitting scores.

Figures

Figures reproduced from arXiv: 1907.08303 by Barbara Bobek-Billewicz, Grzegorz Mrukwa, Jakub Nalepa, Maksym Walczak, Michael P. Hayball, Michal Kawulok, Michal Marcinkiewicz, Pablo Ribalta Lorenzo, Pawel Ulrych, Pawel Wawrzyniak, Wojciech Dudzik.

Figure 1
Figure 1. Figure 1: Our deep learning-powered approach for assessing brain tumor perfusion does [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Our U-Net-based DNN with its blocks and connections. [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Determination of the vascular input region: a) a slice in the axial plane from an [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Fits of the linear model and our cubic model of the VIF to the QIBA phantom [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Examples of the BraTS’17 images segmented using our DNN: a), b), and c) are [PITH_FULL_IMAGE:figures/full_fig_p019_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Examples from our DCE-MRI set segmented using our DNN: a), b), and c) are [PITH_FULL_IMAGE:figures/full_fig_p020_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: QIBA (version 14) phantom set: a) a visualization of a single DICOM file (brown [PITH_FULL_IMAGE:figures/full_fig_p021_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Artifacts generated at the pivotal steps of ECONIB: a) T2-weighted image, b) [PITH_FULL_IMAGE:figures/full_fig_p024_8.png] view at source ↗
read the original abstract

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) plays an important role in diagnosis and grading of brain tumor. Although manual DCE biomarker extraction algorithms boost the diagnostic yield of DCE-MRI by providing quantitative information on tumor prognosis and prediction, they are time-consuming and prone to human error. In this paper, we propose a fully-automated, end-to-end system for DCE-MRI analysis of brain tumors. Our deep learning-powered technique does not require any user interaction, it yields reproducible results, and it is rigorously validated against benchmark (BraTS'17 for tumor segmentation, and a test dataset released by the Quantitative Imaging Biomarkers Alliance for the contrast-concentration fitting) and clinical (44 low-grade glioma patients) data. Also, we introduce a cubic model of the vascular input function used for pharmacokinetic modeling which significantly decreases the fitting error when compared with the state of the art, alongside a real-time algorithm for determination of the vascular input region. An extensive experimental study, backed up with statistical tests, showed that our system delivers state-of-the-art results (in terms of segmentation accuracy and contrast-concentration fitting) while requiring less than 3 minutes to process an entire input DCE-MRI study using a single GPU.

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

2 major / 2 minor

Summary. The manuscript describes a fully-automated end-to-end deep-learning system for DCE-MRI analysis of brain tumors that performs tumor segmentation, real-time vascular input function (VIF) region detection, and pharmacokinetic modeling with a newly introduced cubic VIF model. It reports validation against BraTS'17 (segmentation), the QIBA phantom test set (contrast-concentration fitting), and an independent clinical cohort of 44 low-grade glioma patients, claiming state-of-the-art accuracy, reduced fitting error relative to prior VIF models, and end-to-end runtime under 3 minutes on a single GPU, with supporting statistical tests.

Significance. If the reported performance generalizes, the work would deliver a practical, reproducible tool that removes manual steps from quantitative DCE-MRI biomarker extraction, directly addressing time and variability barriers in clinical tumor grading and prognosis. The cubic VIF model, if shown to be robust rather than over-parameterized, would constitute a concrete methodological advance in pharmacokinetic fitting.

major comments (2)
  1. [Abstract / clinical-validation section] Abstract and clinical-validation section: the headline SOTA claim on the 44-patient low-grade glioma cohort rests on the assumption that BraTS'17-trained segmentation and QIBA-tuned fitting models transfer without measurable domain shift. The manuscript must supply explicit numerical metrics (Dice, Hausdorff, fitting RMSE or R²) for the clinical cases side-by-side with the benchmark numbers, together with any domain-adaptation steps or statistical tests for distribution shift; without these, the independent-validation assertion cannot be evaluated.
  2. [VIF-model section] VIF-model section: the cubic model is introduced as an invented functional form that 'significantly decreases the fitting error.' The manuscript must state the exact equation (including the three free coefficients), compare it to the standard biexponential or other QIBA reference models on identical data, and demonstrate that the error reduction is not an artifact of the added degrees of freedom; otherwise the improvement claim is not load-bearing.
minor comments (2)
  1. Clarify whether the 44 clinical cases were used only for final testing or whether any hyper-parameter tuning occurred on them; if the latter, the independence of the validation set is compromised.
  2. Provide error bars or confidence intervals on all reported metrics (Dice, fitting error, runtime) and state the exact statistical tests used for the SOTA comparisons.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We address each major point below and will revise the manuscript to strengthen the presentation of results and methods.

read point-by-point responses

  1. Referee: [Abstract / clinical-validation section] Abstract and clinical-validation section: the headline SOTA claim on the 44-patient low-grade glioma cohort rests on the assumption that BraTS'17-trained segmentation and QIBA-tuned fitting models transfer without measurable domain shift. The manuscript must supply explicit numerical metrics (Dice, Hausdorff, fitting RMSE or R²) for the clinical cases side-by-side with the benchmark numbers, together with any domain-adaptation steps or statistical tests for distribution shift; without these, the independent-validation assertion cannot be evaluated.

    Authors: We agree that a direct side-by-side comparison would make the transfer performance clearer. The manuscript already reports segmentation and fitting metrics on the 44 clinical cases, but we will add an explicit comparative table in the clinical-validation section showing Dice, Hausdorff, RMSE and R² values for both the BraTS/QIBA benchmarks and the clinical cohort. We will also include statistical tests (e.g., Kolmogorov-Smirnov or Mann-Whitney) for distribution shift between the benchmark and clinical data distributions. No domain-adaptation steps were applied; the models were used in a direct transfer setting to assess generalization. revision: yes

  2. Referee: [VIF-model section] VIF-model section: the cubic model is introduced as an invented functional form that 'significantly decreases the fitting error.' The manuscript must state the exact equation (including the three free coefficients), compare it to the standard biexponential or other QIBA reference models on identical data, and demonstrate that the error reduction is not an artifact of the added degrees of freedom; otherwise the improvement claim is not load-bearing.

    Authors: We will state the exact cubic VIF equation, including the three free coefficients, in the revised VIF-model section. Direct numerical comparisons of fitting error (RMSE) between the cubic model and the standard biexponential (and other QIBA reference) models on the identical QIBA phantom data will be added. To address the degrees-of-freedom concern, we will include an AIC-based model-selection analysis demonstrating that the error reduction is not explained by the additional parameters alone. revision: yes

Circularity Check

0 steps flagged

No circularity; claims rest on external benchmarks and separate clinical cohort

full rationale

The paper's core claims concern segmentation accuracy on BraTS'17, contrast-concentration fitting on the QIBA test set, and performance on a distinct 44-patient low-grade glioma clinical cohort, with no equations, fitted parameters, or self-citations presented as load-bearing derivations. The introduced cubic vascular input function model is described as reducing fitting error relative to prior art without reducing to a self-referential fit. All reported results are positioned as comparisons against independent public benchmarks and a held-out clinical set, satisfying the criteria for a self-contained derivation with no reductions by construction.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 1 invented entities

The central claims rest on standard deep-learning assumptions for medical image segmentation plus the introduction of a new parametric model for the vascular input function whose superiority is asserted via reduced fitting error on the cited test sets.

free parameters (1)
  • cubic VIF model coefficients
    The newly introduced cubic model for the vascular input function necessarily introduces coefficients that are fitted or chosen to minimize error on the contrast-concentration data.
axioms (1)
  • domain assumption Deep learning models trained on BraTS'17 generalize sufficiently to produce state-of-the-art tumor segmentation on the authors' clinical DCE-MRI data.
    Invoked when claiming segmentation accuracy on the 44-patient cohort.
invented entities (1)
  • cubic model of the vascular input function no independent evidence
    purpose: To replace standard VIF models in pharmacokinetic fitting and thereby reduce fitting error for contrast-concentration curves.
    New functional form proposed in the paper; independent evidence outside the reported fitting improvement is not stated in the abstract.

pith-pipeline@v0.9.0 · 5802 in / 1357 out tokens · 57471 ms · 2026-05-24T19:12:28.958950+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

49 extracted references · 49 canonical work pages · 3 internal anchors

  1. [1]

    Abe, et al., Diagnosis of brain tumors using dynamic contrast-enhanced perfusion imaging with short acquisition time, Springer 4 (1) (2015) 88

    T. Abe, et al., Diagnosis of brain tumors using dynamic contrast-enhanced perfusion imaging with short acquisition time, Springer 4 (1) (2015) 88

    work page 2015

  2. [2]

    Aljabar, et al., Multi-atlas based segmentation of brain images: Atlas selection and its effect on accuracy, NeuroImage 46 (3) (2009) 726 – 738

    P. Aljabar, et al., Multi-atlas based segmentation of brain images: Atlas selection and its effect on accuracy, NeuroImage 46 (3) (2009) 726 – 738

    work page 2009

  3. [3]

    Bakas, et al., Advancing the cancer genome atlas glioma MRI collections with expert segmentation labels and radiomic features, Scientific data 4 (2017) 1–13

    S. Bakas, et al., Advancing the cancer genome atlas glioma MRI collections with expert segmentation labels and radiomic features, Scientific data 4 (2017) 1–13

    work page 2017

  4. [4]

    Barata, M

    C. Barata, M. E. Celebi, J. S. Marques, Development of a clinically oriented system for melanoma diagnosis, Pattern Recognition 69 (2017) 270 – 285

    work page 2017

  5. [5]

    Bauer, et al., Atlas-based segmentation of brain tumor images using a Markov Random Field-based tumor growth model and non-rigid registra- tion, in: Proc

    S. Bauer, et al., Atlas-based segmentation of brain tumor images using a Markov Random Field-based tumor growth model and non-rigid registra- tion, in: Proc. IEEE EMBC, 2010, pp. 4080–4083

    work page 2010

  6. [6]

    Bjrnerud, K

    A. Bjrnerud, K. E. Emblem, A fully automated method for quantitative cerebral hemodynamic analysis using dscmri, Journal of Cerebral Blood Flow & Metabolism 30 (5) (2010) 1066–1078

    work page 2010

  7. [7]

    Chander, A

    A. Chander, A. Chatterjee, P. Siarry, A new social and momentum com- ponent adaptive PSO algorithm for image segmentation, Expert Systems with Applications 38 (5) (2011) 4998 – 5004

    work page 2011

  8. [8]

    S.-L. Chao, T. Metens, M. Lemort, Tumourmetrics: a comprehensive clin- ical solution for the standardization of dce-mri analysis in research and routine use, J. QIMS 7 (5)

    work page

  9. [9]

    Cuenod, D

    C. Cuenod, D. Balvay, Perfusion and vascular permeability: Basic concepts and measurement in DCE-CT and DCE-MRI, Diagnostic and Interven- tional Imaging 94 (12) (2013) 1187 – 1204

    work page 2013

  10. [10]

    Dozat, Incorporating Nesterov Momentum into Adam, in: Proc

    T. Dozat, Incorporating Nesterov Momentum into Adam, in: Proc. Work- shop track - ICLR 2016, 2015, pp. 1–6. 27

    work page 2016

  11. [11]

    Fan, et al., A novel unsupervised segmentation method for MR brain images based on fuzzy methods, in: Proc

    X. Fan, et al., A novel unsupervised segmentation method for MR brain images based on fuzzy methods, in: Proc. CVBIA, Springer, Berlin, 2005, pp. 160–169

    work page 2005

  12. [12]

    Geremia, et al., Spatial decision forests for MS lesion segmentation in multi-channel magnetic resonance images, NeuroImage 57 (2) (2011) 378 – 390

    E. Geremia, et al., Spatial decision forests for MS lesion segmentation in multi-channel magnetic resonance images, NeuroImage 57 (2) (2011) 378 – 390

    work page 2011

  13. [13]

    Location Sensitive Deep Convolutional Neural Networks for Segmentation of White Matter Hyperintensities

    M. Ghafoorian, et al., Location sensitive deep convolutional neu- ral networks for segmentation of white matter hyperintensities, CoRR abs/1610.04834

    work page internal anchor Pith review Pith/arXiv arXiv

  14. [14]

    Ghafoorian, et al., Transfer learning for domain adaptation in MRI: application in brain lesion segmentation, in: Proc

    M. Ghafoorian, et al., Transfer learning for domain adaptation in MRI: application in brain lesion segmentation, in: Proc. MICCAI, 2017, pp. 516–524

    work page 2017

  15. [15]

    S. Ji, B. Wei, Z. Yu, G. Yang, Y. Yin, A new multistage medical segmen- tation method based on superpixel and fuzzy clustering, Comp. and Math. Meth. in Med. 2014 (2014) 747549:1–747549:13

    work page 2014

  16. [16]

    K. Kamnitsas, et al., Ensembles of multiple models and architectures for robust brain tumour segmentation, in: Brainlesion: Glioma, Multiple Scle- rosis, Stroke and Traumatic Brain Inj., Springer, 2018, pp. 450–462

    work page 2018

  17. [17]

    Korfiatis, T

    P. Korfiatis, T. L. Kline, B. J. Erickson, Automated segmentation of hyper- intense regions in FLAIR MRI using deep learning, Tomography: a journal for imaging research 2 (4) (2016) 334340

    work page 2016

  18. [18]

    Ladgham, G

    A. Ladgham, G. Torkhani, A. Sakly, A. Mtibaa, Modified support vector machines for MR brain images recognition, in: Proc. CoDIT, 2013, pp. 032–035

    work page 2013

  19. [19]

    Liskowski, K

    P. Liskowski, K. Krawiec, Segmenting retinal blood vessels with deep neural networks, IEEE Transactions on Medical Imaging 35 (11) (2016) 2369– 2380. 28

    work page 2016

  20. [20]

    P. R. Lorenzo, J. Nalepa, Memetic evolution of deep neural networks, in: Proc. GECCO, ACM, 2018, pp. 505–512

    work page 2018

  21. [21]

    P. A. Mei, et al., Analysis of neoplastic lesions in magnetic resonance imag- ing using self-organizing maps, Journal of the Neurological Sciences 359 (1-

    work page

  22. [22]

    B. H. Menze, et al., The multimodal brain tumor image segmentation benchmark (brats), IEEE Transactions on Medical Imaging 34 (10) (2015) 1993–2024

    work page 2015

  23. [23]

    K. A. Miles, B. Ganeshan, M. P. Hayball, CT texture analysis using the filtration-histogram method: what do the measurements mean?, Cancer Imaging 13 (4) (2013) 400–406

    work page 2013

  24. [24]

    Moeskops, et al., Automatic segmentation of MR brain images with a convolutional neural network, IEEE Transactions on Medical Imaging 35 (5) (2016) 1252–1261

    P. Moeskops, et al., Automatic segmentation of MR brain images with a convolutional neural network, IEEE Transactions on Medical Imaging 35 (5) (2016) 1252–1261

    work page 2016

  25. [25]

    Nalepa, M

    J. Nalepa, M. Kawulok, Adaptive genetic algorithm to select training data for support vector machines, in: Proc. Applications of Evolutionary Com- putation, LNCS, Springer, 2014, pp. 514–525

    work page 2014

  26. [26]

    Nalepa, M

    J. Nalepa, M. Kawulok, Adaptive memetic algorithm enhanced with data geometry analysis to select training data for SVMs, Neurocomputing 185 (2016) 113 – 132

    work page 2016

  27. [27]

    M. R. Orton, J. A. d’Arcy, S. Walker-Samuel, D. J. Hawkes, D. Atkinson, D. J. Collins, M. O. Leach, Computationally efficient vascular input func- tion models for quantitative kinetic modelling using DCE-MRI, Physics in Medicine & Biology 53 (5) (2008) 1225

    work page 2008

  28. [28]

    M. T. M. Park, et al., Derivation of high-resolution MRI atlases of the human cerebellum at 3T and segmentation using multiple automatically generated templates, NeuroImage 95 (2014) 217 – 231. 29

    work page 2014

  29. [29]

    Passat, et al., Watershed and multimodal data for brain vessel seg- mentation: Application to the superior sagittal sinus, Image and Vision Computing 25 (4) (2007) 512 – 521

    N. Passat, et al., Watershed and multimodal data for brain vessel seg- mentation: Application to the superior sagittal sinus, Image and Vision Computing 25 (4) (2007) 512 – 521

    work page 2007

  30. [30]

    Pinto, S

    A. Pinto, S. Pereira, H. Correia, J. Oliveira, D. M. L. D. Rasteiro, C. A. Silva, Brain tumour segmentation based on extremely rand. forest with high-level features, in: Proc. IEEE EMBC, 2015, pp. 3037–3040

    work page 2015

  31. [31]

    Pipitone, et al., Multi-atlas segmentation of the whole hippocampus and subfields using multiple automatically generated templates, NeuroImage 101 (2014) 494 – 512

    J. Pipitone, et al., Multi-atlas segmentation of the whole hippocampus and subfields using multiple automatically generated templates, NeuroImage 101 (2014) 494 – 512

    work page 2014

  32. [32]

    R. Port, M. Knopp, G. Brix, Dynamic contrast-enhanced MRI using Gd- DTPA: Interindividual variability of the arterial input function and conse- quences for the assessment of kinetics in tumors, Magn. Res. in Med. 45 (6) (2001) 1030–1038

    work page 2001

  33. [33]

    Rajendran, R

    A. Rajendran, R. Dhanasekaran, Fuzzy clustering and deformable model for tumor segmentation on MRI brain image: A combined approach, Procedia Engineering 30 (2012) 327 – 333

    work page 2012

  34. [34]

    Rezaei, K

    M. Rezaei, K. Harmuth, W. Gierke, T. Kellermeier, M. Fischer, H. Yang, C. Meinel, A conditional adversarial network for semantic segmentation of brain tumor, in: Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Inj., Springer, 2018, pp. 241–252

    work page 2018

  35. [35]

    R. G. Rodr´ ıguez Colmeiro, C. A. Verrastro, T. Grosges, Multimodal brain tumor segmentation using 3D convolutional networks, in: Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Inj., Springer, 2018, pp. 226–240

    work page 2018

  36. [36]

    U-Net: Convolutional Networks for Biomedical Image Segmentation

    O. Ronneberger, P. Fischer, T. Brox, U-Net: Convolutional networks for biomedical image segmentation, CoRR abs/1505.04597. 30

    work page internal anchor Pith review Pith/arXiv arXiv

  37. [37]

    S. Saha, S. Bandyopadhyay, MRI brain image segmentation by fuzzy sym- metry based genetic clustering technique, in: Proc. IEEE CEC, 2007, pp. 4417–4424

    work page 2007

  38. [38]

    V. Simi, J. Joseph, Segmentation of glioblastoma multiforme from MR images - a comprehensive review, The Egyptian Journal of Radiology and Nuclear Medicine 46 (4) (2015) 1105–1110

    work page 2015

  39. [39]

    D. S. Smith, X. Li, L. R. Arlinghaus, T. E. Yankeelov, E. B. Welch, DCEMRI.jl: a fast, validated, open source toolkit for dynamic contrast enhanced MRI analysis, PeerJ 3 (2015) e909

    work page 2015

  40. [40]

    Soltaninejad, et al., Automated brain tumour detection and segmen- tation using superpixel-based extremely randomized trees in FLAIR MRI, Int

    M. Soltaninejad, et al., Automated brain tumour detection and segmen- tation using superpixel-based extremely randomized trees in FLAIR MRI, Int. J. of Comp. Assist. Radiol. and Surgery 12 (2) (2017) 183–203

    work page 2017

  41. [41]

    C. Sun, S. Guo, H. Zhang, J. Li, M. Chen, S. Ma, L. Jin, X. Liu, X. Li, X. Qian, Automatic segmentation of liver tumors from multiphase contrast- enhanced ct images based on fcns, Artificial Intelligence in Medicine 83 (2017) 58 – 66

    work page 2017

  42. [42]

    Taherdangkoo, et al., An effective method for segmentation of MR brain images using the ant colony optimization algorithm, Journal of Digital Imaging 26 (6) (2013) 1116–1123

    M. Taherdangkoo, et al., An effective method for segmentation of MR brain images using the ant colony optimization algorithm, Journal of Digital Imaging 26 (6) (2013) 1116–1123

    work page 2013

  43. [43]

    P. S. Tofts, A. G. Kermode, Measurement of the bloodbrain barrier perme- ability and leakage space using dynamic MR imaging, Magn. Res. in Med. 17 (2) (1991) 357–367

    work page 1991

  44. [44]

    Verma, M

    N. Verma, M. C. Cowperthwaite, M. K. Markey, Superpixels in brain MR image analysis, in: Proc. IEEE EMBC, 2013, pp. 1077–1080

    work page 2013

  45. [45]

    Wu, et al., Brain tumor detection and segmentation in a CRF frame- work with pixel-pairwise affinity and superpixel-level features, Int

    W. Wu, et al., Brain tumor detection and segmentation in a CRF frame- work with pixel-pairwise affinity and superpixel-level features, Int. J. CARS 9 (2) (2014) 241–253. 31

    work page 2014

  46. [46]

    J. Yin, J. Yang, Q. Guo, Automatic determination of the arterial input function in dynamic susceptibility contrast MRI: comparison of different reproducible clustering algorithms, Neuroradiology 57 (5) (2015) 535–543

    work page 2015

  47. [47]

    X. Zhao, Y. Wu, G. Song, Z. Li, Y. Zhang, Y. Fan, A deep learning model integrating FCNNs and CRFs for brain tumor segmentation, CoRR abs/1702.04528

    work page internal anchor Pith review Pith/arXiv arXiv

  48. [48]

    Zhuge, A

    Y. Zhuge, A. V. Krauze, H. Ning, J. Y. Cheng, B. C. Arora, K. Cam- phausen, R. W. Miller, Brain tumor segmentation using holistically nested neural networks in MRI images, Med. Phys. (2017) 1–10

    work page 2017

  49. [49]

    Zikic, B

    D. Zikic, B. Glocker, E. Konukoglu, A. Criminisi, C. Demiralp, J. Shotton, O. M. Thomas, T. Das, R. Jena, S. J. Price, Decision forests for tissue- specific segmentation of high-grade gliomas in multi-channel MR, in: Proc. MICCAI, Springer, 2012, pp. 369–376. 32

    work page 2012