A Novel 3D Unsupervised Domain Adaptation Framework for Cross-Modality Medical Image Segmentation

Abstract

We consider the problem of volumetric (3D) unsupervised domain adaptation (UDA) in cross-modality medical image segmentation, aiming to perform segmentation on the unannotated target domain (e.g. MRI) with the help of labeled source domain (e.g. CT). Previous UDA methods in medical image analysis usually suffer from two challenges: 1) they focus on processing and analyzing data at 2D level only, thus missing semantic information from the depth level; 2) one-to-one mapping is adopted during the style-transfer process, leading to insufficient alignment in the target domain. Different from the existing methods, in our work, we conduct a first of its kind investigation on multi-style image translation for complete image alignment to alleviate the domain shift problem, and also introduce 3D segmentation in domain adaptation tasks to maintain semantic consistency at the depth level. In particular, we develop an unsupervised domain adaptation framework incorporating a novel quartet self-attention module to efficiently enhance relationships between widely separated features in spatial regions on a higher dimension, leading to a substantial improvement in segmentation accuracy in the unlabeled target domain. In two challenging crossmodality tasks, specifically brain structures and multiorgan abdominal segmentation, our model is shown to outperform current state-of-the-art methods by a significant margin, demonstrating its potential as a benchmark resource for the biomedical and health informatics research community. 1

A novel internet of things based on deep neural network framework using soft-attention convolutional neural networks for COVID-19 detection

“…The proposed model is very efficient in differentiating successfully between COVID-19 images with accuracy ranging from 96.43% to 98.96%. In the future, the REGATT model will be improved using new optimal strategies [55] and employed in other challenging detection problems, such as 3D image analysis [56]. This model will also be exploited to detect different kinds of diseases (such as [57]) and predict their severity.…”

Automated liver tissues delineation techniques: A systematic survey on machine learning current trends and future orientations

“…By contrast, DA is accomplished to tackle domain shifts often occuring when the medical images are recorded by different equipment or in different environments (Hong et al, 2022a). DA is mainly based on different approaches: (i) divergence-based DA; (ii) adversarial-based DA using GANs (Hong et al, 2022b); and (iii) reconstruction-based DA using stacked autoencoders (SAEs) or GANs (Yao et al, 2022).…”

“…On the other hand, labeling liver/tumor segmentation images in the target domain is a crucial challenge, which necessitates the intervention of experienced radiologists (Yao et al, 2022). Fortunately, unsupervised DA techniques help transfer knowledge across domains without the need for annotated data in the target domain.…”

3D Brain and Heart Volume Generative Models: A Survey

“…Specifically, they proposed a GAN-regularized 3D U-Net structure, which uses a pixel-topixel scheme to concatenate the input with labels for effective training. DAR-UNet [217] modeled the cross-modality medical image segmentation (e.g., brain structure segmentation) within an Unsupervised Domain Adaptation (UDA) framework. The first step was to train multiple GANs to disentangle the style and content of both the source and target images at the same time.…”