Abstract

A matched-condition comparison of four U-Net family decoder architectures (U-Net, UNet++, MAnet, LinkNet) crossed with four standard segmentation losses (Dice, Focal, Tversky, Focal-Tversky) is conducted on the BraTS 2020 dataset, producing a 16-configuration matrix. All configurations share an identical ResNet34 encoder pre-trained on ImageNet, identical preprocessing, augmentation, optimizer schedule, and held-out 38-case test set. Statistical significance is assessed via paired bootstrap resampling with 10,000 iterations on per-case Dice. Mean Dice across all 16 configurations falls in the narrow range 0.792 to 0.826. UNet++ paired with Dice loss ranks first (mean 0.826; WT 0.851, TC 0.799, ET 0.828), but the top five configurations are statistically indistinguishable (all p greater than 0.32, with 95 percent confidence intervals spanning zero). LinkNet with Dice attains mean Dice 0.815, within 0.011 of the top, while requiring 2.7 times less inference time (53.9 versus 145.4 ms per case) and 18 percent fewer parameters than UNet++ with Dice. Worst-case analysis reveals three failure patterns: edema over-segmentation in atypical presentations, necrotic-core and edema confusion, and missed multi-focal disease. The findings suggest that improvements claimed in prior literature may be partly attributable to unreported variation in experimental conditions, and that LinkNet offers a favorable accuracy-efficiency trade-off for clinical deployment.

References

• Z. U. Abidin, R. A. Naqvi, A. Haider, H. S. Kim, D. Jeong, and S. W. Lee, "Recent deep learning-based brain tumor segmentation models using multi-modality magnetic resonance imaging: A prospective survey," Frontiers in Bioengineering and Biotechnology, vol. 12, art. 1392807, 2024.
• N. Abraham and N. M. Khan, "A novel focal Tversky loss function with improved attention U-Net for lesion segmentation," in IEEE International Symposium on Biomedical Imaging (ISBI), pp. 683-687, 2019.
• U. Baid, S. Ghodasara, S. Mohan, M. Bilello, E. Calabrese, et al., "The RSNA-ASNR-MICCAI BraTS 2021 benchmark on brain tumor segmentation and radiogenomic classification," arXiv:2107.02314, 2021.
• S. Bakas, H. Akbari, A. Sotiras, M. Bilello, M. Rozycki, J. S. Kirby, et al., "Advancing the Cancer Genome Atlas glioma MRI collections with expert segmentation labels and radiomic features," Scientific Data, vol. 4, art. 170117, 2017.
• A. Chaurasia and E. Culurciello, "LinkNet: Exploiting encoder representations for efficient semantic segmentation," in IEEE Visual Communications and Image Processing (VCIP), 2017.
• J. Chen, Y. Lu, Q. Yu, X. Luo, E. Adeli, et al., "TransUNet: Transformers make strong encoders for medical image segmentation," arXiv:2102.04306, 2021.
• O. Cicek, A. Abdulkadir, S. S. Lienkamp, T. Brox, and O. Ronneberger, "3D U-Net: Learning dense volumetric segmentation from sparse annotation," in MICCAI, pp. 424-432, Springer, 2016.
• A. Hatamizadeh, V. Nath, Y. Tang, D. Yang, H. R. Roth, and D. Xu, "Swin UNETR: Swin transformers for semantic segmentation of brain tumors in MRI images," in Brainlesion (BrainLes 2021), vol. 12962, pp. 272-284, Springer, 2022.
• K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition," in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 770-778, 2016.
• F. Isensee, P. F. Jaeger, S. A. A. Kohl, J. Petersen, and K. H. Maier-Hein, "nnU-Net: A self-configuring method for deep learning-based biomedical image segmentation," Nature Methods, vol. 18, no. 2, pp. 203-211, 2021.
• F. Isensee, P. F. Jaeger, P. M. Full, P. Vollmuth, and K. H. Maier-Hein, "nnU-Net for brain tumor segmentation," arXiv:2011.00848, 2020.
• F. Isensee, T. Wald, C. Ulrich, M. Baumgartner, S. Roy, K. Maier-Hein, and P. F. Jaeger, "nnU-Net Revisited: A call for rigorous validation in 3D medical image segmentation," in MICCAI, 2024. arXiv:2404.09556.
• T.-Y. Lin, P. Goyal, R. Girshick, K. He, and P. Dollar, "Focal loss for dense object detection," in IEEE International Conference on Computer Vision (ICCV), pp. 2980-2988, 2017.
• Z. Liu, L. Chen, L. Tong, F. Zhou, Z. Jiang, et al., "Deep learning based brain tumor segmentation: A survey," arXiv:2007.09479, 2020.
• T. Magadza and S. Viriri, "Deep learning for brain tumor segmentation: A survey of state-of-the-art," Journal of Imaging, vol. 7, no. 2, art. 19, 2021.
• B. H. Menze, A. Jakab, S. Bauer, J. Kalpathy-Cramer, K. Farahani, et al., "The multimodal brain tumor image segmentation benchmark (BRATS)," IEEE Transactions on Medical Imaging, vol. 34, no. 10, pp. 1993-2024, 2015.
• F. Milletari, N. Navab, and S.-A. Ahmadi, "V-Net: Fully convolutional neural networks for volumetric medical image segmentation," in 4th International Conference on 3D Vision (3DV), pp. 565-571, IEEE, 2016.
• O. Oktay, J. Schlemper, L. Le Folgoc, M. Lee, M. Heinrich, et al., "Attention U-Net: Learning where to look for the pancreas," in Medical Imaging with Deep Learning (MIDL), 2018.
• Q. T. Ostrom, G. Cioffi, K. Waite, C. Kruchko, and J. S. Barnholtz-Sloan, "CBTRUS statistical report: Primary brain and other central nervous system tumors diagnosed in the United States in 2014-2018," Neuro-Oncology, vol. 23, Supplement 3, pp. iii1-iii105, 2021.
• O. Ronneberger, P. Fischer, and T. Brox, "U-Net: Convolutional networks for biomedical image segmentation," in Medical Image Computing and Computer-Assisted Intervention (MICCAI), pp. 234-241, Springer, 2015.
• S. S. M. Salehi, D. Erdogmus, and A. Gholipour, "Tversky loss function for image segmentation using 3D fully convolutional deep networks," in Machine Learning in Medical Imaging (MLMI), vol. 10541, pp. 379-387, Springer, 2017.
• S. L. Smith, P.-J. Kindermans, C. Ying, and Q. V. Le, "Don't decay the learning rate, increase the batch size," in International Conference on Learning Representations (ICLR), 2018.
• Z. Zhou, M. M. R. Siddiquee, N. Tajbakhsh, and J. Liang, "UNet++: A nested U-Net architecture for medical image segmentation," in Deep Learning in Medical Image Analysis (DLMIA), vol. 11045, pp. 3-11, Springer, 2018.