English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Large-scale models pre-trained on large-scale datasets have profoundlyadvanced the development of deep learning. However, the state-of-the-art modelsfor medical image segmentation are still small-scale, with their parametersonly in the tens of millions. Further scaling them up to higher orders ofmagnitude is rarely explored. An overarching goal of exploring large-scalemodels is to train them on large-scale medical segmentation datasets for bettertransfer capacities. In this work, we design a series of Scalable andTransferable U-Net (STU-Net) models, with parameter sizes ranging from 14million to 1.4 billion. Notably, the 1.4B STU-Net is the largest medical imagesegmentation model to date. Our STU-Net is based on nnU-Net framework due toits popularity and impressive performance. We first refine the defaultconvolutional blocks in nnU-Net to make them scalable. Then, we empiricallyevaluate different scaling combinations of network depth and width, discoveringthat it is optimal to scale model depth and width together. We train ourscalable STU-Net models on a large-scale TotalSegmentator dataset and find thatincreasing model size brings a stronger performance gain. This observationreveals that a large model is promising in medical image segmentation.Furthermore, we evaluate the transferability of our model on 14 downstreamdatasets for direct inference and 3 datasets for further fine-tuning, coveringvarious modalities and segmentation targets. We observe good performance of ourpre-trained model in both direct inference and fine-tuning. The code andpre-trained models are available at https://github.com/Ziyan-Huang/STU-Net.

STU-Net: Scalable and Transferable Medical Image Segmentation Models Empowered by Large-Scale Supervised Pre-training