Do fusion modules and compound losses really help? A multi-seed ablation of a hybrid cnn–transformer model for 3D brain tumor segmentation

Abdukadirova, G. X., Abduqodirova, M. I.

Innovative Multidisciplinary Journal of Applied Technology · 2026-yil

Annotatsiya

Hybrid CNN–Transformer models for brain tumor segmentation are usually justified by ablation studies in which each variant is trained once, although run-to-run variation can be as large as the reported gains. This study re-examines the components of the HyCTSeg model - a parallel CNN and Swin Transformer encoder with bidirectional cross-attention fusion (CAFF) and a four-component loss (Dice, cross-entropy, focal, Tversky) - with a statistically powered design. Nine variants were each trained with four random seeds on BraTS 2021 (32 independent runs, 150 epochs) on an identical data split, and compared by paired t-tests with Benjamini–Hochberg correction. The multi-seed design reduced the standard deviation to 0.0006–0.0018 Dice. The full model ranked third of five architectures (0.8593 ± 0.0015); the Swin-only encoder was best (0.8645 ± 0.0016) and exceeded the full model by 0.0052 (p = 0.0096), and simple additive fusion exceeded it by 0.0032 (p = 0.0107), both significant after correction. The differences were concentrated in the enhancing tumor region. The compound loss did not outperform Dice loss alone (−0.0004, p = 0.38). These negative results show that single-run ablations can overstate the value of architectural components and argue for multi-seed evaluation as standard practice.

Maqola ma’lumotlari
MualliflarAbdukadirova, G. X., Abduqodirova, M. I.
JurnalInnovative Multidisciplinary Journal of Applied Technology
NashriyotТДТУ Қўқон филиали
ISSN2995-486X, 2995-486x
Nashr sanasi2026-10-12
Jild4
Son10
Betlar72-77
TilIngliz
DOI10.51699/4typtc28
Indeksga qo‘shilgan2026-10-13

DOI: 10.51699/4typtc28 · Maqolaning asl sahifasi

Kalit so‘zlar

brain tumor segmentation, hybrid CNN–Transformer, feature fusion, cross-attention, compound loss, ablation study, multi-seed evaluation, reproducibility, negative results, BraTS 2021

Iqtibos keltirish

APAAbdukadirova, G. X., & Abduqodirova, M. I. (2026). DO FUSION MODULES AND COMPOUND LOSSES REALLY HELP? A MULTI-SEED ABLATION OF A HYBRID CNN–TRANSFORMER MODEL FOR 3D BRAIN TUMOR SEGMENTATION. Innovative Multidisciplinary Journal of Applied Technology, 4(10), 72-77. https://doi.org/10.51699/4typtc28
GOST (OAK)Abdukadirova, G. X., Abduqodirova, M. I. DO FUSION MODULES AND COMPOUND LOSSES REALLY HELP? A MULTI-SEED ABLATION OF A HYBRID CNN–TRANSFORMER MODEL FOR 3D BRAIN TUMOR SEGMENTATION // Innovative Multidisciplinary Journal of Applied Technology. – 2026. – T. 4, № 10. – P. 72-77. – DOI: 10.51699/4typtc28.
MLAAbdukadirova, G. X., and Abduqodirova, M. I. “DO FUSION MODULES AND COMPOUND LOSSES REALLY HELP? A MULTI-SEED ABLATION OF A HYBRID CNN–TRANSFORMER MODEL FOR 3D BRAIN TUMOR SEGMENTATION.” Innovative Multidisciplinary Journal of Applied Technology, 4(10), 72-77, 2026. https://doi.org/10.51699/4typtc28
ChicagoAbdukadirova, G. X., and Abduqodirova, M. I. “DO FUSION MODULES AND COMPOUND LOSSES REALLY HELP? A MULTI-SEED ABLATION OF A HYBRID CNN–TRANSFORMER MODEL FOR 3D BRAIN TUMOR SEGMENTATION.” Innovative Multidisciplinary Journal of Applied Technology, 4(10), 72-77 (2026). https://doi.org/10.51699/4typtc28
HarvardAbdukadirova, G. X., and Abduqodirova, M. I (2026) ‘DO FUSION MODULES AND COMPOUND LOSSES REALLY HELP? A MULTI-SEED ABLATION OF A HYBRID CNN–TRANSFORMER MODEL FOR 3D BRAIN TUMOR SEGMENTATION’, Innovative Multidisciplinary Journal of Applied Technology, 4(10), 72-77. https://doi.org/10.51699/4typtc28
BibTeX
@article{abdukadirovagx2026,
  title = {DO FUSION MODULES AND COMPOUND LOSSES REALLY HELP? A MULTI-SEED ABLATION OF A HYBRID CNN–TRANSFORMER MODEL FOR 3D BRAIN TUMOR SEGMENTATION},
  author = {Abdukadirova, G. X. and Abduqodirova, M. I.},
  journal = {Innovative Multidisciplinary Journal of Applied Technology},
  year = {2026},
  volume = {4},
  number = {10},
  pages = {72-77},
  doi = {10.51699/4typtc28},
  url = {https://doi.org/10.51699/4typtc28}
}

Boshqa maqola uchun iqtibos tayyorlash

Innovative Multidisciplinary Journal of Applied Technology jurnalidan boshqa maqolalar

Innovative Multidisciplinary Journal of Applied Technology — barcha maqolalar · 2026-yil arxivi