ABSTRACT
Despite the recent success of deep learning models for text generation, generating clinically accurate reports remains challenging. More precisely modeling the relationships of the abnormalities revealed in an X-ray image has been found promising to enhance the clinical accuracy. In this paper, we first introduce a novel knowledge graph structure called an attributed abnormality graph (ATAG). It consists of interconnected abnormality nodes and attribute nodes for better capturing more fine-grained abnormality details. In contrast to the existing methods where the abnormality graph are constructed manually, we propose a methodology to automatically construct the fine-grained graph structure based on annotated X-ray reports and the RadLex radiology lexicon. We then learn the ATAG embeddings as part of a deep model with an encoder-decoder architecture for the report generation. In particular, graph attention networks are explored to encode the relationships among the abnormalities and their attributes. A hierarchical attention attention and a gating mechanism are specifically designed to further enhance the generation quality. We carry out extensive experiments based on the benchmark datasets, and show that the proposed ATAG-based deep model outperforms the SOTA methods by a large margin in ensuring the clinical accuracy of the generated reports.
Subject(s)
X-RaysABSTRACT
Breakfast is an important contributor to the daily dietary intake of children. This study investigated the nutritional composition of ready to eat (RTE) children’s breakfast cereals, which display fictional cartoon characters and themes, compared to other cereals available in Australia. Nutrient content claims on packaging were also examined. Data were collected from RTE breakfast cereal packages (N = 347) from four major supermarkets in Sydney. Cereals were classified based on product type and promotional information displayed. Overall, 46% of children’s cereals were classified as “less healthy” as per nutrient profiling score criteria. Children’s cereals had a similar energy and sodium content per 100 g compared to other cereals but contained significantly higher levels of total sugar and lower levels of protein and dietary fibre compared to other varieties. Children’s cereals with nutrient content claims had improved (lower) nutrient profiling scores than those that did not (2 vs. 13, p = 0.021), but total sugar per 100 g was similar: 25 g (interquartile range (IQR) 14 g) vs. 32 g (IQR 19 g). In conclusion, RTE children’s breakfast cereals were found to be less healthy compared to other cereals on the market and the use of nutrient content claims on children’s cereals may mislead consumers regarding their overall nutrient profile.
ABSTRACT
The Flaviviridae family consists of several medically important pathogens such as West Nile virus (WNV) and Dengue virus (DENV). Flavivirus capsid (C) protein is a key structural component of virus particles. However, the role of C protein in the pathogenesis of arthropod-borne flaviviruses is poorly understood. To examine whether flavivirus C protein can associate with cellular proteins, and contribute to viral pathogenesis, WNV/DENV C protein was screened against a human brain/liver cDNA yeast two-hybrid library. This study identified human Sec3 exocyst protein (hSec3p) as a novel interacting partner of WNV and DENV C protein. Mutagenesis studies showed that the SH2 domain-binding motif of hSec3p binds to the first 15 amino acids of C protein. We report for the first time that hSec3p can modulate virus production by affecting viral RNA transcription and translation through the sequestration of elongation factor 1alpha (EF1alpha). This molecular discovery shed light on the protective role of hSec3p during flavivirus infection. This study also highlighted the antagonistic mechanism adopted by flavivirus C protein that can negatively regulate the formation of hSec3p-EF1alpha complex by sequestering hSec3p to establish successful infection.
