pith. sign in

arxiv: 2512.14648 · v1 · submitted 2025-12-16 · 💻 cs.CV · eess.IV

Adaptable Segmentation Pipeline for Diverse Brain Tumors with Radiomic-guided Subtyping and Lesion-Wise Model Ensemble

Daniel Capell\'an-Mart\'in , Abhijeet Parida , Zhifan Jiang , Nishad Kulkarni , Krithika Iyer , Austin Tapp , Syed Muhammad Anwar , Mar\'ia J. Ledesma-Carbayo
show 1 more author
Marius George Linguraru
This is my paper

Pith reviewed 2026-05-16 21:28 UTC · model grok-4.3

classification 💻 cs.CV eess.IV
keywords brain tumor segmentationradiomicsmodel ensembleBraTS challengeMRI segmentationpediatric brain tumorsmeningiomabrain metastases
0
0 comments X

The pith

An adaptable pipeline uses radiomic subtyping and lesion-wise ensembles to segment diverse brain tumors at competitive accuracy.

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

The paper introduces a modular pipeline designed to handle the wide variation in brain tumor types on MRI scans. It detects tumor subtypes using radiomic features to balance training data and then ensembles multiple models with weights based on custom lesion-level performance metrics. Post-processing is also optimized per lesion. This setup is tested on several BraTS 2025 challenge datasets covering pediatric tumors, meningiomas, and brain metastases. The approach aims to deliver robust segmentations without being tied to a single model architecture.

Core claim

The central discovery is that a flexible segmentation pipeline, guided by radiomic features for tumor subtyping and using lesion-specific metrics to determine model contributions in an ensemble, achieves performance comparable to leading methods across multiple BraTS challenges for adult and pediatric brain tumors.

What carries the argument

The radiomic-guided subtyping combined with a lesion-wise model ensemble that dynamically weights predictions and refines post-processing based on per-lesion metrics.

If this is right

  • The method supports segmentation of multiple tumor categories including pediatric, meningioma, and metastatic tumors within one framework.
  • Lesion-level metrics allow for tailored post-processing that improves prediction accuracy for individual cases.
  • Without dependence on a fixed network, the pipeline can incorporate new state-of-the-art models as they emerge.
  • Quantitative tumor measurements become feasible for supporting diagnosis and prognosis in clinical settings.

Where Pith is reading between the lines

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

  • Extending radiomic subtyping to other medical imaging tasks could improve generalization in segmentation problems.
  • Testing the pipeline on real-world clinical data with varying scan quality would reveal its practical limits.
  • Combining this with longitudinal imaging might enable better tracking of tumor changes over time.

Load-bearing premise

That radiomic features extracted from MRI scans can accurately detect tumor subtypes and that performance metrics at the lesion level can reliably set ensemble weights and rules that apply to new, unseen cases.

What would settle it

Observing a dataset of diverse brain tumors where the radiomic subtype detection misclassifies tumors leading to unbalanced training or where the lesion-wise ensemble weights result in lower performance than individual models on test cases.

Figures

Figures reproduced from arXiv: 2512.14648 by Abhijeet Parida, Austin Tapp, Daniel Capell\'an-Mart\'in, Krithika Iyer, Mar\'ia J. Ledesma-Carbayo, Marius George Linguraru, Nishad Kulkarni, Syed Muhammad Anwar, Zhifan Jiang.

Figure 1
Figure 1. Figure 1: Overview of the segmentation pipeline. PP, CC, and LBL REDEF refer [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Clustering visualization across challenges using top two radiomic features. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Post-PP1 confusion matrix indicating labels that need to be redefined in [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Adaptive Post-processing: optimal thresholds per cluster and label across [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative results showing median lesion-wise Dice (L) or global Dice [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
read the original abstract

Robust and generalizable segmentation of brain tumors on multi-parametric magnetic resonance imaging (MRI) remains difficult because tumor types differ widely. The BraTS 2025 Lighthouse Challenge benchmarks segmentation methods on diverse high-quality datasets of adult and pediatric tumors: multi-consortium international pediatric brain tumor segmentation (PED), preoperative meningioma tumor segmentation (MEN), meningioma radiotherapy segmentation (MEN-RT), and segmentation of pre- and post-treatment brain metastases (MET). We present a flexible, modular, and adaptable pipeline that improves segmentation performance by selecting and combining state-of-the-art models and applying tumor- and lesion-specific processing before and after training. Radiomic features extracted from MRI help detect tumor subtype, ensuring a more balanced training. Custom lesion-level performance metrics determine the influence of each model in the ensemble and optimize post-processing that further refines the predictions, enabling the workflow to tailor every step to each case. On the BraTS testing sets, our pipeline achieved performance comparable to top-ranked algorithms across multiple challenges. These findings confirm that custom lesion-aware processing and model selection yield robust segmentations yet without locking the method to a specific network architecture. Our method has the potential for quantitative tumor measurement in clinical practice, supporting diagnosis and prognosis.

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 presents a modular, adaptable pipeline for segmenting diverse brain tumors (pediatric, meningioma, and metastases) on multi-parametric MRI from the BraTS 2025 challenges. It uses radiomic features extracted from MRI to detect tumor subtypes for balanced training, applies lesion-level performance metrics to weight models in an ensemble, and optimizes case-specific post-processing. The central claim is that this workflow produces segmentations comparable to top-ranked algorithms on the held-out test sets for PED, MEN, MEN-RT, and MET without locking to a single network architecture.

Significance. If the performance claims hold with supporting evidence, the work would demonstrate a practical, architecture-agnostic framework that tailors subtyping, ensembling, and post-processing to tumor and lesion variability. This addresses a core challenge in generalizable brain-tumor segmentation and could support quantitative clinical measurements for diagnosis and prognosis across heterogeneous datasets.

major comments (2)
  1. [Abstract] Abstract: The claim that 'our pipeline achieved performance comparable to top-ranked algorithms across multiple challenges' is stated without any Dice scores, HD95 values, ablation tables, error bars, or direct numerical comparisons to BraTS leaderboard entries. This absence prevents verification of the central performance assertion.
  2. [Results] Results (or equivalent evaluation section): No quantitative metrics, tables, or figures report test-set performance on PED, MEN, MEN-RT, or MET, nor any details on how radiomic subtype detection, lesion-level weighting, or post-processing rules were validated on held-out data. These omissions make it impossible to assess whether the described components actually deliver the claimed robustness.
minor comments (2)
  1. [Methods] Methods: Provide explicit definitions and pseudocode for how radiomic features are extracted and thresholded to assign subtypes, and how lesion-level metrics are aggregated to set ensemble weights.
  2. [Figure 1] Figure 1 (pipeline diagram): Ensure all steps (radiomic subtyping, lesion-wise weighting, post-processing) are clearly labeled with input/output arrows and that the diagram distinguishes training from inference paths.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We agree that the submitted manuscript lacks explicit quantitative metrics and direct comparisons, which are required to substantiate the performance claims. We will revise the manuscript to address these points by adding the necessary results, tables, and details.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The claim that 'our pipeline achieved performance comparable to top-ranked algorithms across multiple challenges' is stated without any Dice scores, HD95 values, ablation tables, error bars, or direct numerical comparisons to BraTS leaderboard entries. This absence prevents verification of the central performance assertion.

    Authors: We acknowledge that the abstract in the current version states the comparability claim without supporting numerical values. In the revised manuscript we will expand the abstract to report specific Dice and HD95 scores (with error bars) for PED, MEN, MEN-RT, and MET, together with direct numerical comparisons against the top-ranked BraTS 2025 leaderboard entries. We will also reference the ablation studies that support the contribution of each pipeline component. revision: yes

  2. Referee: [Results] Results (or equivalent evaluation section): No quantitative metrics, tables, or figures report test-set performance on PED, MEN, MEN-RT, or MET, nor any details on how radiomic subtype detection, lesion-level weighting, or post-processing rules were validated on held-out data. These omissions make it impossible to assess whether the described components actually deliver the claimed robustness.

    Authors: We agree that the evaluation section must be strengthened. The revised manuscript will contain a dedicated Results section that reports full test-set metrics (Dice, HD95, and additional overlap measures) for each of the four BraTS 2025 challenges. We will also add tables and text detailing the validation of radiomic subtype detection accuracy, the lesion-level weighting scheme, and the case-specific post-processing rules, all evaluated on held-out data. Ablation experiments quantifying the contribution of each module will be included. revision: yes

Circularity Check

0 steps flagged

No significant circularity in pipeline derivation or performance claims

full rationale

The paper describes a modular pipeline that selects models and applies radiomic subtyping plus lesion-level metrics for ensemble weighting and post-processing. The performance claim references external BraTS testing sets and top-ranked algorithms without presenting any equations, fitted parameters, or self-citations that reduce the reported results to quantities defined by the paper's own inputs. The derivation relies on empirical selection against independent benchmarks and is therefore self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The pipeline rests on the assumption that radiomic features carry sufficient subtype signal and that lesion-level validation metrics transfer to test performance; no new physical entities or mathematical axioms are introduced.

pith-pipeline@v0.9.0 · 5569 in / 1078 out tokens · 15109 ms · 2026-05-16T21:28:06.923572+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

29 extracted references · 29 canonical work pages · 2 internal anchors

  1. [1]

    The RSNA-ASNR-MICCAI BraTS 2021 Benchmark on Brain Tumor Segmentation and Radiogenomic Classification

    Baid, U., Ghodasara, S., Mohan, S., et al.: The RSNA-ASNR-MICCAI BraTS 2021 benchmark on brain tumor segmentation and radiogenomic classification. arXiv preprint arXiv:2107.02314 (2021)

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  2. [2]

    Scientific Data4(1), 170117 (sep 2017).https://doi.org/10.1038/sdata.2017

    Bakas, S., Akbari, H., Sotiras, A., et al.: Advancing The Cancer Genome Atlas glioma MRI collections with expert segmentation labels and radiomic features. Scientific Data4(1), 170117 (sep 2017).https://doi.org/10.1038/sdata.2017. 117,https://doi.org/10.1038/sdata.2017.117

    work page doi:10.1038/sdata.2017 2017

  3. [3]

    The Cancer Imag- ing Archive (2017).https://doi.org/10.7937/K9/TCIA.2017.KLXWJJ1Q

    Bakas, S., Akbari, H., Sotiras, A., et al.: Segmentation Labels and Radiomic Fea- tures for the Pre-operative Scans of the TCGA-GBM collection. The Cancer Imag- ing Archive (2017).https://doi.org/10.7937/K9/TCIA.2017.KLXWJJ1Q

    work page doi:10.7937/k9/tcia.2017.klxwjj1q 2017

  4. [4]

    The Cancer Imag- ing Archive (2017).https://doi.org/10.7937/K9/TCIA.2017.GJQ7R0EF

    Bakas, S., Akbari, H., Sotiras, A., et al.: Segmentation Labels and Radiomic Fea- tures for the Pre-operative Scans of the TCGA-LGG collection. The Cancer Imag- ing Archive (2017).https://doi.org/10.7937/K9/TCIA.2017.GJQ7R0EF

    work page doi:10.7937/k9/tcia.2017.gjq7r0ef 2017

  5. [5]

    In: International ChallengeonCross-ModalityDomainAdaptationforMedicalImageSegmentation, pp

    Capellán-Martín, D., Jiang, Z., Parida, A., Liu, X., Lam, V., Nisar, H., Tapp, A., Elsharkawi, S., Ledesma-Carbayo, M.J., Anwar, S.M., et al.: Model Ensemble for Brain Tumor Segmentation in Magnetic Resonance Imaging. In: International ChallengeonCross-ModalityDomainAdaptationforMedicalImageSegmentation, pp. 221–232. Springer (2023)

    work page 2023

  6. [6]

    Fathi Kazerooni, A., Khalili, N., Liu, X., Gandhi, D., Jiang, Z., et al.: The Brain Tumor Segmentation in Pediatrics (BraTS-PEDs) Challenge: Focus on Pedi- atrics (CBTN-CONNECT-DIPGR-ASNR-MICCAI BraTS-PEDs) (2024),https: //arxiv.org/abs/2404.15009

    work page arXiv 2024

  7. [7]

    Hatamizadeh, A., Nath, V., Tang, Y., et al.: Swin UNETR: Swin transformers for semanticsegmentationofbraintumorsinmriimages.In:Crimi,A.,Bakas,S.(eds.) Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries. pp. 272–284. Springer, Cham (2022)

    work page 2022

  8. [8]

    Nature methods18(2), 203–211 (2021)

    Isensee, F., Jaeger, P.F., Kohl, S.A., et al.: nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nature methods18(2), 203–211 (2021)

    work page 2021

  9. [9]

    In: Medical Image Computing and Computer Assisted Intervention – MICCAI 2024

    Isensee, F., Wald, T., Ulrich, C., Baumgartner, M., Roy, S., et al.: nnU-Net Re- visited: A Call for Rigorous Validation in 3D Medical Image Segmentation. In: Medical Image Computing and Computer Assisted Intervention – MICCAI 2024. pp. 488–498. Springer Nature Switzerland, Cham (2024)

    work page 2024

  10. [10]

    In: 2024 IEEE International Symposium on Biomedical Imaging (ISBI)

    Jiang, Z., Capellán-Martín, D., Parida, A., Liu, X., Ledesma-Carbayo, M.J., An- war, S.M., Linguraru, M.G.: Enhancing Generalizability in Brain Tumor Segmenta- tion: Model Ensemble with Adaptive Post-Processing. In: 2024 IEEE International Symposium on Biomedical Imaging (ISBI). pp. 1–4. IEEE (2024)

    work page 2024

  11. [11]

    arXiv preprint arXiv:2412.04094 (2024)

    Jiang, Z., Capellán-Martín, D., Parida, A., Tapp, A., Liu, X., Ledesma-Carbayo, M.J., Anwar, S.M., Linguraru, M.G.: Magnetic Resonance Imaging Feature-Based Subtyping and Model Ensemble for Enhanced Brain Tumor Segmentation. arXiv preprint arXiv:2412.04094 (2024)

    work page arXiv 2024

  12. [12]

    In: 2023 45th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC)

    Jiang, Z., Parida, A., Anwar, S.M., Tang, Y., Roth, H.R., Fisher, M.J., Packer, R.J., Avery, R.A., Linguraru, M.G.: Automatic visual acuity loss prediction in children with optic pathway gliomas using magnetic resonance imaging. In: 2023 45th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC). pp. 1–5. IEEE (2023)

    work page 2023

  13. [13]

    Nat Mach Intell5, 799–810 (2023) 12 D

    Karargyris, A., Umeton, R., Sheller, M., et al.: Federated benchmarking of medical artificial intelligence with medperf. Nat Mach Intell5, 799–810 (2023) 12 D. Capellán-Martín, A. Parida, Z. Jiang et al

    work page 2023

  14. [14]

    Kazerooni, A.F., Khalili, N., Liu, X., et al.: The Brain Tumor Segmentation (BraTS) Challenge 2023: Focus on Pediatrics (CBTN-CONNECT-DIPGR-ASNR- MICCAI BraTS-PEDs) (2023)

    work page 2023

  15. [15]

    LaBella, D., Adewole, M., Alonso-Basanta, M., et al.: The ASNR-MICCAI Brain Tumor Segmentation (BraTS) Challenge 2023: Intracranial Meningioma (2023)

    work page 2023

  16. [16]

    LaBella, D., Baid, U., Khanna, O., McBurney-Lin, S., McLean, R., et al.: Analy- sis of the BraTS 2023 intracranial meningioma segmentation challenge. J. Mach. Learn. Biomed. Imaging3(March 2025), 38–58 (Mar 2025)

    work page 2023

  17. [17]

    LaBella, D., Schumacher, K., Mix, M., Leu, K., McBurney-Lin, S., et al.: Brain Tu- mor Segmentation (BraTS) Challenge 2024: Meningioma Radiotherapy Planning Automated Segmentation (2024),https://arxiv.org/abs/2405.18383

    work page arXiv 2024

  18. [18]

    U-Mamba: Enhancing Long-range Dependency for Biomedical Image Segmentation

    Ma, J., Li, F., Wang, B.: U-Mamba: Enhancing Long-range Dependency for Biomedical Image Segmentation. arXiv preprint arXiv:2401.04722 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  19. [19]

    In: International Challenge on Cross-Modality Domain Adaptation for Medical Image Segmentation, pp

    Maani, F., Hashmi, A.U.R., Aljuboory, M., Saeed, N., Sobirov, I., Yaqub, M.: Advanced tumor segmentation in medical imaging: An ensemble approach for BraTS 2023 adult glioma and pediatric tumor tasks. In: International Challenge on Cross-Modality Domain Adaptation for Medical Image Segmentation, pp. 264–277. Springer (2023)

    work page 2023

  20. [20]

    The Multimodal Brain Tumor Image Segmentation Benchmark (BRATS),

    Menze, B.H., Jakab, A., Bauer, S., et al.: The Multimodal Brain Tumor Image Seg- mentation Benchmark (BRATS). IEEE Transactions on Medical Imaging34(10), 1993–2024 (2015).https://doi.org/10.1109/TMI.2014.2377694

    work page doi:10.1109/tmi.2014.2377694 1993

  21. [21]

    Moawad,A.W.,Janas,A.,Baid,U.,etal.:TheBrainTumorSegmentation(BraTS- METS) Challenge 2023: Brain Metastasis Segmentation on Pre-treatment MRI (2023)

    work page 2023

  22. [22]

    arXiv preprint arXiv:2412.04111 (2024)

    Parida, A., Capellán-Martín, D., Jiang, Z., Tapp, A., Liu, X., Anwar, S.M., Ledesma-Carbayo, M.J., Linguraru, M.G.: Adult Glioma Segmentation in Sub- Saharan Africa using Transfer Learning on Stratified Finetuning Data. arXiv preprint arXiv:2412.04111 (2024)

    work page arXiv 2024

  23. [23]

    In: International Conference on Medical Image Computing and Computer-Assisted Intervention

    Roy, S., Koehler, G., Ulrich, C., Baumgartner, M., Petersen, J., Isensee, F., Jaeger, P.F., Maier-Hein, K.H.: Mednext: Transformer-driven scaling of convnets for medi- cal image segmentation. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 405–415. Springer (2023)

    work page 2023

  24. [24]

    Siegel, R.L., Kratzer, T.B., Giaquinto, A.N., Sung, H., Jemal, A.: Cancer statistics,

    work page

  25. [25]

    Cancer statistics, 2025,

    CA: A Cancer Journal for Clinicians75(1), 10–45 (2025).https://doi.org/ https://doi.org/10.3322/caac.21871,https://acsjournals.onlinelibrary. wiley.com/doi/abs/10.3322/caac.21871

    work page doi:10.3322/caac.21871 2025

  26. [26]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Tang, Y., Yang, D., Li, W., Roth, H.R., et al.: Self-Supervised Pre-Training of Swin Transformers for 3D Medical Image Analysis. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 20730–20740 (2022)

    work page 2022

  27. [27]

    Cancer research77(21), e104–e107 (2017)

    Van Griethuysen, J.J., Fedorov, A., Parmar, C., Hosny, A., Aucoin, N., Narayan, V., Beets-Tan, R.G., Fillion-Robin, J.C., Pieper, S., Aerts, H.J.: Computational radiomics system to decode the radiographic phenotype. Cancer research77(21), e104–e107 (2017)

    work page 2017

  28. [28]

    IEEE Transactions on Medical Imaging23(7), 903–921 (2004).https://doi.org/10

    Warfield, S., Zou, K., Wells, W.: Simultaneous truth and performance level esti- mation (STAPLE): an algorithm for the validation of image segmentation. IEEE Transactions on Medical Imaging23(7), 903–921 (2004).https://doi.org/10. 1109/TMI.2004.828354

    work page arXiv 2004

  29. [29]

    IEEE Transactions on Medical Imaging pp

    Xing, Z., Ye, T., Yang, Y., Cai, D., Gai, B., Wu, X.J., Gao, F., Zhu, L.: Segmamba-v2: Long-range sequential modeling mamba for general 3d medical image segmentation. IEEE Transactions on Medical Imaging pp. 1–1 (2025). https://doi.org/10.1109/TMI.2025.3589797

    work page doi:10.1109/tmi.2025.3589797 2025