ABSTRACT
convolutional neural networks (CNNs) show great potential in medical image segmentation tasks, and can provide reliable basis for disease diagnosis and clinical research. However, CNNs exhibit general limitations on modeling explicit long-range relation, and existing cures, resorting to building deep encoders along with aggressive downsampling operations, leads to loss of localized details. Transformer has naturally excellent ability to model the global features and long-range correlations of the input information, which is strongly complementary to the inductive bias of CNNs. In this paper, a novel Bi-directional Multi-scale Cascaded Segmentation Network, BMCS-Net, is proposed to improve the performance of medical segmentation tasks by aggregating these features obtained from Transformers and CNNs branches. Specifically, a novel feature integration technique, termed as Two-stream Cascaded Feature Aggregation (TCFA) module, is designed to fuse features in two-stream branches, and solve the problem of gradual dilution of global information in the network. Besides, a Multi-Scale Expansion-Aware (MSEA) module based on the convolution of feature perception and expansion is introduced to capture context information, and further compensate for the loss of details. Extensive experiments demonstrated that BMCS-Net has an excellent performance on both skin and Polyp segmentation datasets.
ABSTRACT
Neoatherosclerosis is a major cause of stent failure after percutaneous coronary intervention. Metabolism such as hyperuricemia is associated with in-stent restenosis (ISR). However, the association between serum uric acid (sUA) levels and in-stent neoatherosclerosis (ISNA) has never been validated.A total of 216 patients with 220 ISR lesions who had undergone optical coherence tomography (OCT) of culprit stents were included in this study. According to their sUA levels, eligible patients were divided into two groups [normal-sUA group: sUA < 7 mg/dL, n = 126, and high-sUA group: sUA ≥ 7 mg/dL, n = 90]. OCT findings were analyzed and compared between the normal- and high-sUA groups.The incidence of ISNA (63.0% versus 43.0%, P = 0.004) was significantly higher in the high-sUA group than in the normal-sUA group. Lipid plaques (66.3% versus 43.0%, P < 0.001) and thin-cap fibroatheroma (38.0% versus 18.0%, P = 0.001) were observed more frequently in the restenotic tissue structure in patients in the high-sUA group than in those in the normal-sUA group. Meanwhile, univariate (OR: 1.208, 95% CI: 1.037-1.407; P = 0.015) and multivariate (OR: 1.254, 95% CI: 1.048-1.501; P = 0.013) logistic regression analyses indicated that sUA levels were an independent risk factor for ISNA after adjusting for relevant risk factors.The high-sUA levels were an independent risk factor for the occurrence of neoatherosclerosis in patients with ISR via OCT.