# An attempt at beating the 3D U-Net

arXiv: https://arxiv.org/abs/1908.02182 · HTML: https://mlsanity.com/paper/1908.02182/

## Key facts

- Cited by 21 papers in the map (data through 7 Oct 2026).
- First posted on arXiv 6 Aug 2019.

## Built on by (most cited first, 21 of 21)

- [UNETR: Transformers for 3D Medical Image Segmentation](https://mlsanity.com/paper/2103.10504/) (2021)
- [nnU-Net Revisited: A Call for Rigorous Validation in 3D Medical Image Segmentation](https://mlsanity.com/paper/2404.09556/) (2024)
- [A Deep-Learning Approach For Direct Whole-Heart Mesh Reconstruction](https://mlsanity.com/paper/2102.07899/) (2021)
- [MultiTalent: A Multi-Dataset Approach to Medical Image Segmentation](https://mlsanity.com/paper/2303.14444/) (2023)
- [Learning Whole Heart Mesh Generation From Patient Images For Computational Simulations](https://mlsanity.com/paper/2203.10517/) (2022)
- [Using Out-of-the-Box Frameworks for Contrastive Unpaired Image Translation for Vestibular Schwannoma and Cochlea Segmentation: An approach for the crossMoDA Challenge](https://mlsanity.com/paper/2110.01607/) (2021)
- [LNQ 2023 challenge: Benchmark of weakly-supervised techniques for mediastinal lymph node quantification](https://mlsanity.com/paper/2408.10069/) (2024)
- [Understanding the Tricks of Deep Learning in Medical Image Segmentation: Challenges and Future Directions](https://mlsanity.com/paper/2209.10307/) (2022)
- [Self-Supervised Learning for Segmentation](https://mlsanity.com/paper/2101.05456/) (2021)
- [Look Ma, no code: fine tuning nnU-Net for the AutoPET II challenge by only adjusting its JSON plans](https://mlsanity.com/paper/2309.13747/) (2023)
- [BeSt-LeS: Benchmarking Stroke Lesion Segmentation using Deep Supervision](https://mlsanity.com/paper/2310.07060/) (2023)
- [SE3D: A Framework For Saliency Method Evaluation In 3D Imaging](https://mlsanity.com/paper/2405.14584/) (2024)
- [Automated Segmentation of Ischemic Stroke Lesions in Non-Contrast Computed Tomography Images for Enhanced Treatment and Prognosis](https://mlsanity.com/paper/2411.09402/) (2024)
- [Decomposition Sampling for Efficient Region Annotations in Active Learning](https://mlsanity.com/paper/2512.07606/) (2025)
- [Crosslink-Net: Double-branch Encoder Segmentation Network via Fusing Vertical and Horizontal Convolutions](https://mlsanity.com/paper/2107.11517/) (2021)
- [Knowledge Distillation from Cross Teaching Teachers for Efficient Semi-Supervised Abdominal Organ Segmentation in CT](https://mlsanity.com/paper/2211.05942/) (2022)
- [Kidney and Kidney Tumour Segmentation in CT Images](https://mlsanity.com/paper/2212.13034/) (2022)
- [TransCC: Transformer Network for Coronary Artery CCTA Segmentation](https://mlsanity.com/paper/2310.04779/) (2023)
- [Segmentation of Kidney Tumors on Non-Contrast CT Images using Protuberance Detection Network](https://mlsanity.com/paper/2312.04796/) (2023)
- [From Convolution to Transformer: A Comparative Study of U-Net Variants for Brain Tumor and Retinal Vessel Segmentation](https://mlsanity.com/paper/2606.22168/) (2026)
- [Deep Learning Segmentation of Diffusion-Weighted MRI Acute Ischaemic Stroke: A Pragmatic Evaluation Across Three Datasets](https://mlsanity.com/paper/2608.25675/) (2026)

## Abstract (arXiv)

The U-Net is arguably the most successful segmentation architecture in the medical domain. Here we apply a 3D U-Net to the 2019 Kidney and Kidney Tumor Segmentation Challenge and attempt to improve upon it by augmenting it with residual and pre-activation residual blocks. Cross-validation results on the training cases suggest only very minor, barely measurable improvements. Due to marginally higher dice scores, the residual 3D U-Net is chosen for test set prediction. With a Composite Dice score of 91.23 on the test set, our method outperformed all 105 competing teams and won the KiTS2019 challenge by a small margin.