No abstract

No abstract

Accurate segmentation is a crucial step in medical image analysis and applying supervised machine learning to segment the organs or lesions has been substantiated effective. However, it is costly to perform data annotation that provides ground truth labels for training the supervised algorithms, and the high variance of data that comes from different domains tends to severely degrade system performance over cross-site or cross-modality datasets. To mitigate this problem, a novel unsupervised domain adaptation (UDA) method named dispensed Transformer network (DTNet) is introduced in this paper. Our novel DTNet contains three modules. First, a dispensed residual transformer block is designed, which realizes global attention by dispensed interleaving operation and deals with the excessive computational cost and GPU memory usage of the Transformer. Second, a multi-scale consistency regularization is proposed to alleviate the loss of details in the low-resolution output for better feature alignment. Finally, a feature ranking discriminator is introduced to automatically assign different weights to domaingap features to lessen the feature distribution distance, reducing the performance shift of two domains. The proposed method is evaluated on large fluorescein angiography (FA) retinal nonperfusion (RNP) cross-site dataset with 676 images and a wide used cross-modality dataset from the MM-WHS challenge. Extensive results demonstrate that our proposed network achieves the best performance in comparison with several state-of-the-art

Purpose The ongoing debate focuses on whether the freeze-all strategy is suitable for the general population or may be offered to specific subgroups of patients. This study aimed to compare the pregnancy and neonatal outcomes between FET and fresh ET and evaluate the effectiveness of the embryo freezing strategy for a specific group of patients undergoing single poor cleavage-stage embryo transfer. Methods A total of 1,819 ET cycles that underwent single poor cleavage-stage embryo transfer between January 2014 and December 2020 were enrolled in this study and categorized into two groups according to the embryo processing methods: fresh ET group (n = 1124) and frozen ET group (n = 695). Results We found that the clinical pregnancy and live birth rates were significantly higher in the fresh cycles than in the frozen cycles (32.38% vs. 22.30%, p = 0.000; 25.62% vs. 16.12%, p = 0.000, respectively). The multivariate logistic regression model showed that the cycle type (fresh or frozen) still had a significant impact on the live birth rate (OR 1.62, 95% CI: 1.19–2.21, p = 0.002) after adjusting for potential confounders. Discussion FET was associated with a significantly lower risk of clinical pregnancy and live birth rates. For patients who are more likely to develop poor-quality cleavage embryos, such as women with previous low response or reduced ovarian reserve, it is suggested that fresh ET should be given priority to achieve better pregnancy and neonatal outcomes than FET if there are no special circumstances.

Segmentation of brain tumors and their subregions remains a challenging task due to their weak features and deformable shapes. In this paper, three patterns (cross-skip, skip-1 and skip-2) of distributed dense connections (DDCs) are proposed to enhance feature reuse and propagation of CNNs by constructing tunnels between key layers of the network. For better detecting and segmenting brain tumors from multimodal 3D MR images, CNN-based models embedded with DDCs (DDU-Nets) are trained efficiently from pixel to pixel with a limited number of parameters. Postprocessing is then applied to refine the segmentation results by reducing the false-positive samples. The proposed method is evaluated on the BraTS 2019 dataset with results demonstrating the effectiveness of the DDU-Nets while requiring less computational cost.