In this paper we tackle the problem of unsupervised domain adaptation for the task of semantic segmentation, where we attempt to transfer the knowledge learned upon synthetic datasets with ground-truth labels to real-world images without any annotation. With the hypothesis that the structural content of images is the most informative and decisive factor to semantic segmentation and can be readily shared across domains, we propose a Domain Invariant Structure Extraction (DISE) framework to disentangle images into domain-invariant structure and domain-specific texture representations, which can further realize imagetranslation across domains and enable label transfer to improve segmentation performance. Extensive experiments verify the effectiveness of our proposed DISE model and demonstrate its superiority over several state-of-the-art approaches.

No abstract

Stroke survivors have high likelihood of readmission within 1 year following discharge, with infections and recurrent vascular events being the most common reasons. Identification of high-risk subgroups might foster preventive interventions.

No abstract

The transcriptional network of the SRY (sex determining region Y)-box 17 (SOX17) and the prognostic impact of SOX17 protein expression in human cancers remain largely unclear. In this study, we evaluated the prognostic effect of low SOX17 protein expression and its dysregulation of transcriptional network in esophageal squamous cell carcinoma (ESCC). Low SOX17 protein expression was found in 47.4% (73 of 154) of ESCC patients with predicted poor prognosis. Re-expression of SOX17 in ESCC cells caused reduced foci formation, cell motility, decreased ESCC xenograft growth and metastasis in animals. Knockdown of SOX17 increased foci formation in ESCC and normal esophageal cells. Notably, 489 significantly differential genes involved in cell growth and motility controls were identified by expression array upon SOX17 overexpression and 47 genes contained putative SRY element in their promoters. Using quantitative chromatin immunoprecipitation-PCR and promoter activity assays, we confirmed that MACC1, MALAT1, NBN, NFAT5, CSNK1A1, FN1 and SERBP1 genes were suppressed by SOX17 via the SRY binding-mediated transcriptional regulation. Overexpression of FN1 and MACC1 abolished SOX17-mediated migration and invasion suppression. The inverse correlation between SOX17 and FN1 protein expression in ESCC clinical samples further strengthened our conclusion that FN1 is a transcriptional repression target gene of SOX17. This study provides compelling clinical evidence that low SOX17 protein expression is a prognostic biomarker and novel cell and animal data of SOX17-mediated suppression of ESCC metastasis. We establish the first transcriptional network and identify new suppressive downstream genes of SOX17 which can be potential therapeutic targets for ESCC.