MedNeXt-v2: Scaling 3D ConvNeXts for Large-Scale Supervised Representation Learning in Medical Image Segmentation

TL;DR

MedNeXt-v2 introduces a scaled 3D ConvNeXt architecture with improved micro-architecture and data scaling, achieving state-of-the-art results in large-scale supervised 3D medical image segmentation.

Contribution

This work revisits and enhances ConvNeXt-based architectures for volumetric segmentation, demonstrating the importance of backbone selection and scaling strategies for optimal performance.

Findings

Stronger backbones predict better downstream performance.

Representation scaling benefits pathological segmentation.

Modality-specific pretraining offers negligible benefits after full finetuning.

Abstract

Large-scale supervised pretraining is rapidly reshaping 3D medical image segmentation. However, existing efforts focus primarily on increasing dataset size and overlook the question of whether the backbone network is an effective representation learner at scale. In this work, we address this gap by revisiting ConvNeXt-based architectures for volumetric segmentation and introducing MedNeXt-v2, a compound-scaled 3D ConvNeXt that leverages improved micro-architecture and data scaling to deliver state-of-the-art performance. First, we show that routinely used backbones in large-scale pretraining pipelines are often suboptimal. Subsequently, we use comprehensive backbone benchmarking prior to scaling and demonstrate that stronger from scratch performance reliably predicts stronger downstream performance after pretraining. Guided by these findings, we incorporate a 3D Global Response Normalization module and use depth, width, and context scaling to improve our architecture for effective representation learning. We pretrain MedNeXt-v2 on 18k CT volumes and demonstrate state-of-the-art performance when fine-tuning across six challenging CT and MR benchmarks (144 structures), showing consistent gains over seven publicly released pretrained models. Beyond improvements, our benchmarking of these models also reveals that stronger backbones yield better results on similar data, representation scaling disproportionately benefits pathological segmentation, and that modality-specific pretraining offers negligible benefit once full finetuning is applied. In conclusion, our results establish MedNeXt-v2 as a strong backbone for large-scale supervised representation learning in 3D Medical Image Segmentation. Our code and pretrained models are made available with the official nnUNet repository at: https://www.github.com/MIC-DKFZ/nnUNet