nnU-Net: Self-adapting Framework for U-Net-Based Medical Image Segmentation

TL;DR

The paper tackles the generalization gap in medical image segmentation by introducing nnU-Net, a self-adapting framework built on vanilla U-Nets that automatically configures architectures and all surrounding steps (preprocessing, training, inference, postprocessing, and ensembling) to fit each dataset. It argues that non-architectural factors drive much of the performance and demonstrates this via the Medical Segmentation Decathlon, where nnU-Net achieves top mean Dice across most phase-1 tasks without manual tuning. The approach systematically automates dataset-specific decisions, including geometry-adaptive networks (2D, 3D, cascade), resampling, normalization, data augmentation, and test-time strategies, enabling robust cross-dataset performance. The results suggest that carefully automated preprocessing and training pipelines can rival task-specific architectural innovations, with implications for generalizable medical image segmentation pipelines and future ablation-focused studies.

Abstract

The U-Net was presented in 2015. With its straight-forward and successful architecture it quickly evolved to a commonly used benchmark in medical image segmentation. The adaptation of the U-Net to novel problems, however, comprises several degrees of freedom regarding the exact architecture, preprocessing, training and inference. These choices are not independent of each other and substantially impact the overall performance. The present paper introduces the nnU-Net ('no-new-Net'), which refers to a robust and self-adapting framework on the basis of 2D and 3D vanilla U-Nets. We argue the strong case for taking away superfluous bells and whistles of many proposed network designs and instead focus on the remaining aspects that make out the performance and generalizability of a method. We evaluate the nnU-Net in the context of the Medical Segmentation Decathlon challenge, which measures segmentation performance in ten disciplines comprising distinct entities, image modalities, image geometries and dataset sizes, with no manual adjustments between datasets allowed. At the time of manuscript submission, nnU-Net achieves the highest mean dice scores across all classes and seven phase 1 tasks (except class 1 in BrainTumour) in the online leaderboard of the challenge.

Figures (1)

• Figure 1: U-Net Cascade (on applicable datasets only). Stage 1 (left): a 3D U-Net processes downsampled data, the resulting segmentation maps are upsampled to the original resolution. Stage 2 (right): these segmentations are concatenated as one-hot encodings to the full resolution data and refined by a second 3D U-Net.