Keywords: Medical Image Segmentation, AutoML, Neural Architecture Search, Model Scaling
Abstract: The U-Net and its variants are proved as the most successful architectures in the medical image segmentation domain. However, the optimal configuration of the hyperparameters in U-Net structure such as depth and width remain challenging to adjust manually due to the diversity of medical image segmentation tasks. In this paper, we propose AdwU-Net, which is an efficient neural architecture search framework to search the optimal task-specific depth and width in the U-Net backbone. Specifically, an adaptive depth and width block is designed and applied hierarchically in U-Net. In each block, the optimal number of convolutional layers and channels in each layer are directly learned from data. To reduce the computational costs and alleviate the memory pressure, we conduct an efficient architecture search and reuse the network weights of different depth and width options in a differentiable manner. Extensive experiments on the Medical Segmentation Decathlon (MSD) dataset show that our method outperforms not only the manually scaled U-Net but also other state-of-the-art architectures. Our code is publicly available at https://github.com/Ziyan-Huang/AdwU-Net.
Registration: I acknowledge that publication of this at MIDL and in the proceedings requires at least one of the authors to register and present the work during the conference.
Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.
Paper Type: both
Primary Subject Area: Segmentation
Secondary Subject Area: Segmentation
Confidentiality And Author Instructions: I read the call for papers and author instructions. I acknowledge that exceeding the page limit and/or altering the latex template can result in desk rejection.
Code And Data: Code: https://github.com/Ziyan-Huang/AdwU-Net. Data: http://medicaldecathlon.com/