Skip to content
DraconiticResearch
Search Sources API
203,234 studies31 sourcesUpdated 20 min ago
Peer-reviewedCC BYDiagnostics (Basel, Switzerland)

Detector-Stage Failures Account for Most of the Increase in the External Deficit of a Box-Prompted Glioma Segmentation Pipeline on BraTS-Africa: A Three-Arm Decomposition Across Two External Cohorts

  • Aykut Kaşkaya1
  • Baha Şen2
  1. 1Department of Computer Engineering, , Ankara 06010, Türkiye
  2. 2Department of Computer Engineering, Turkish Military Academy, National Defence University, Ankara 06420, Türkiye;
Read full textPDFDOICite

Abstract

Background/Objectives: When a box-prompted segmentation pipeline loses accuracy on new data, the detector proposing the box or a later stage may be responsible; this study measured which. Methods: A real-time detection transformer (RT-DETR) feeding the Segment Anything Model 2.1 was compared with a detector-free model (MedSAM3 with a low-rank adapter) on whole-tumor Dice in 158 patients with glioma: held-out Brain Tumor Segmentation (BraTS) 2020 cases (n = 24), the RHUH-GBM glioblastoma cohort (n = 39), and BraTS-Africa (n = 95, glioma presumed). Giving the segmenter the ground-truth box splits the Dice gap between the systems into a detector-stage term and a residual. Results: The detector-stage term was +0.006 Dice on held-out cases, +0.018 on RHUH-GBM, and +0.091 on BraTS-Africa. Its increase over the held-out value was +0.0854 [95% CI +0.0450, +0.1303] on BraTS-Africa (Holm-adjusted p < 0.001) and, after correction of a preprocessing fault, +0.0120 [−0.0062, +0.0295] on RHUH-GBM, which was not resolved. On BraTS-Africa, it came mainly from slices without a detector box. No residual change was resolved. Conclusions: On BraTS-Africa, the detector stage accounted for most of the increase in the pipeline’s gap; on RHUH-GBM, no increase was resolved. The decomposition should be measured per cohort.

Related

  1. A Case Report Series on Supplementation with 1100 mg Myo-Inositol/ 300 mg D-Chiro-Inositol (3.6:1 Ratio), Antioxidants and Probiotics to Improve Oocyte Quality and Increase Pregnancy Chances in Women With History of ART Failures Without PCOS
    LI Devesa Otero MD, A Forgiarini MD, M Ruíz Batalla MD +1
    Clinical Medicine Insights: Case ReportsPeer-reviewedCC BYFull textEpidemiology & public healthDOAJ
  2. Convolutional, State-Space and Transformer Liver-Tumor Segmentation Models Trained on Hepatocellular Carcinoma Fail on the Same External Cases: Lesion Size Outweighs Architecture in Cross-Dataset Generalization
    Hüseyin Kutlu, Cemil Çolak, Sami Akbulut
    CancersPeer-reviewedCC BY 4.0Full textOncologyCrossrefEurope PMC
  3. Failures and lessons learned from the medical school quota expansion policy under the Yoon Suk Yeol administration
    Yu Jin Kim, So Young Kim, Jong Hyock Park
    Frontiers in Public HealthPeer-reviewedCC BY 4.0Full textEpidemiology & public healthPMCCrossrefEurope PMC
  4. Structured reasoning failures compromise LLM interpretation of clinical oncology notes
    Matthew W. Kenaston, Umair Ayub, Mihir Parmar +13
    npj Digital MedicinePeer-reviewedCC BYFull text2 citedEpidemiology & public healthDOAJ
  5. When conviction outpaces evidence: diagnostic boundary failures of large language models in adult ADHD self-assessment
    Jiajia Qin, Lu Bai, Chujie Qu +3
    Frontiers in PsychiatryPeer-reviewedCC BY 4.0Psychiatry & mental healthCrossref

Cite

BibTeXRIS
BibTeX
@article{kaskaya2026detectorstage,
  title = {{Detector-Stage Failures Account for Most of the Increase in the External Deficit of a Box-Prompted Glioma Segmentation Pipeline on BraTS-Africa: A Three-Arm Decomposition Across Two External Cohorts}},
  author = {Aykut Kaşkaya and Baha Şen},
  journal = {Diagnostics (Basel, Switzerland)},
  year = {2026},
  month = oct,
  doi = {10.3390/diagnostics16203391},
  url = {https://europepmc.org/article/PMC/PMC13703227},
  note = {CC BY},
}
RIS
TY  - JOUR
TI  - Detector-Stage Failures Account for Most of the Increase in the External Deficit of a Box-Prompted Glioma Segmentation Pipeline on BraTS-Africa: A Three-Arm Decomposition Across Two External Cohorts
AU  - Aykut Kaşkaya
AU  - Baha Şen
PY  - 2026
DA  - 2026/10/01
JO  - Diagnostics (Basel, Switzerland)
DO  - 10.3390/diagnostics16203391
UR  - https://europepmc.org/article/PMC/PMC13703227
AB  - Background/Objectives: When a box-prompted segmentation pipeline loses accuracy on new data, the detector proposing the box or a later stage may be responsible; this study measured which. Methods: A real-time detection transformer (RT-DETR) feeding the Segment Anything Model 2.1 was compared with a detector-free model (MedSAM3 with a low-rank adapter) on whole-tumor Dice in 158 patients with glioma: held-out Brain Tumor Segmentation (BraTS) 2020 cases (n = 24), the RHUH-GBM glioblastoma cohort (n = 39), and BraTS-Africa (n = 95, glioma presumed). Giving the segmenter the ground-truth box splits the Dice gap between the systems into a detector-stage term and a residual. Results: The detector-stage term was +0.006 Dice on held-out cases, +0.018 on RHUH-GBM, and +0.091 on BraTS-Africa. Its increase over the held-out value was +0.0854 [95% CI +0.0450, +0.1303] on BraTS-Africa (Holm-adjusted p < 0.001) and, after correction of a preprocessing fault, +0.0120 [−0.0062, +0.0295] on RHUH-GBM, which was not resolved. On BraTS-Africa, it came mainly from slices without a detector box. No residual change was resolved. Conclusions: On BraTS-Africa, the detector stage accounted for most of the increase in the pipeline’s gap; on RHUH-GBM, no increase was resolved. The decomposition should be measured per cohort.
SN  - 2075-4418
ER  -