Abstract:Purpose: In neonatal brain Magnetic Resonance Image(MRI) segmentation, the model we trained on the training set(source domain) often performs poorly in clinical practice(target domain). Since the label of target-domain images is unavailable, this cross-domain segmentation needs unsupervised domain adaptation(UDA) to make the model adapt to the target domain. However, the shape and intensity distribution of neonatal brain MRI images across the domains are largely different from adults'. Current UDA methods aim to make synthesized images similar to the target domain as a whole. But it is impossible to synthesize images with intra-class similarity, because of the regional misalignment caused by the cross-domain difference. This will result in generating intra-classly incorrect intensity information from target-domain images. To address this issue, we propose an IAS-NET(joint Intra-classly Adaptive GAN and Segmentation) framework to bridge the gap between the two domains for intra-class alignment.Methods: Our proposed IAS-NET is an elegant learning framework that transfers the appearance of images across the domains from both image and feature perspectives. It consists of the proposed intra-classly adaptive GAN(IA-NET) and a segmentation network(S-NET). The proposed IA-NET is a GAN-based adaptive network, which contains one generator(including two encoders and one shared decoder) and four discriminators for cross-domain transfer. The two encoders are implemented to extract original image, mean, and variance features from source and target domains. The proposed Local Adaptive Instance Normalization (LAdaIN) algorithm is used to performs intra-class feature alignment to the target domain in the feature-map level. S-NET is a U-net structure network, which is used to provide semantic constraint by a segmentation loss for the training of IA-NET. Meanwhile it offers pseudo-label images for calculating intra-class features of the target domain. Source code (in Tensorflow) is available at: https://github.com/lb-whu/RAS-NET/.Results: Extensive experiments are carried out on two different data sets(NeoBrainS12 and dHCP) respectively. There exist great differences in the shape, size, and intensity distribution of MR images in the two databases. Compared to baseline, we improve the average dice score of all tissues on NeoBrains12 by 6% through adaptive training with unlabeled dHCP images. Besides, we also conduct experiments on dHCP and improved the average dice score by 4%. The quantitative analysis of the mean and variance of the synthesized images shows that the synthesized image by the proposed is closer to the target domain both in the full brain or within each class than that of the compared methods.Conclusions: In this paper, the proposed IAS-NET can improve the performance of the segmentation network effectively by its intra-class feature alignment in the target domain. Compared to the current UDA methods, the synthesized images by IAS-NET are more intra-classly similar to the target domain for neonatal brain MR images. Therefore, it achieves state-of-the-art results in the compared UDA models for the segmentation task.This article is protected by copyright. All rights reserved

Harmonized neonatal brain MR image segmentation model for cross-site datasets

Automated 3D Fetal Brain Segmentation Using an Optimized Deep Learning Approach

CNS - CycleGAN-Assisted Neonatal Segmentation Model for Cross-Datasets.

Harmonized Segmentation of Neonatal Brain MRI

IAS‐NET: Joint intraclassly adaptive GAN and segmentation network for unsupervised cross‐domain in neonatal brain MRI segmentation

Unified Model for Children's Brain Image Segmentation with Co-Registration Framework Guided by Longitudinal MRI

Multi-task Learning for Neonatal Brain Segmentation Using 3D Dense-Unet with Dense Attention Guided by Geodesic Distance

Deep Learning-Based Multiclass Brain Tissue Segmentation in Fetal MRIs

Multi-Site Infant Brain Segmentation Algorithms: The iSeg-2019 Challenge

MG-Net: A fetal brain tissue segmentation method based on multiscale feature fusion and graph convolution attention mechanisms

SUSAN: Segment Unannotated image Structure using Adversarial Network

Improving segmentation reliability of multi-scanner brain images using a generative adversarial network

Segmentation-Renormalized Deep Feature Modulation for Unpaired Image Harmonization

Automated cardiac segmentation of cross-modal medical images using unsupervised multi-domain adaptation and spatial neural attention structure

A Deep Learning Harmonization of Multi-Vendor MRI for Robust Intervertebral Disc Segmentation

Simultaneous Segmentation of Fetal Hearts and Lungs for Medical Ultrasound Images Via an Efficient Multi-scale Model Integrated with Attention Mechanism

Inter-site harmonization based on dual generative adversarial networks for diffusion tensor imaging: application to neonatal white matter development

Improving the generalizability of convolutional neural network-based segmentation on CMR images

3D‐MASNet: 3D mixed‐scale asymmetric convolutional segmentation network for 6‐month‐old infant brain MR images

A foundation model for enhancing magnetic resonance images and downstream segmentation, registration and diagnostic tasks

Automated neonatal nnU-Net brain MRI extractor trained on a large multi-institutional dataset