Search

Computer-aided disease diagnosis and prognosis based on medical images is a rapidly emerging field. Many Convolutional Neural Network (CNN) architectures have been developed by researchers for disease classification and localization from chest X-ray images. It is known that different thoracic disease lesions are more likely to occur in specific anatomical regions compared to others. Based on this knowledge, we first estimate a disease-dependent spatial probability, i.e., an anatomical prior, that indicates the probability of occurrence of a disease in a specific region in a chest X-ray image. Next, we develop a novel attention-based classification model that combines information from the estimated anatomical prior and automatically extracted chest region of interest (ROI) masks to provide attention to the feature maps generated from a deep convolution network. Unlike previous works that utilize various self-attention mechanisms, the proposed method leverages the extracted chest ROI masks along with the probabilistic anatomical prior information, which selects the region of interest for different diseases to provide attention. The proposed method shows superior performance in disease classification on the NIH ChestX-ray14 dataset compared to existing state-of-the-art methods while reaching an area under the ROC curve (AUC) of 0.8427. Regarding disease localization, the proposed method shows competitive performance compared to state-of-the-art methods, achieving an accuracy of 61% with an Intersection over Union (IoU) threshold of 0.3. The proposed method can also be generalized to other medical image-based disease classification and localization tasks where the probability of occurrence of the lesion is dependent on specific anatomical sites.

Overlapping community detection is a key problem in graph mining. Some research has considered applying graph convolutional networks (GCN) to tackle the problem. However, it is still challenging to incorporate deep graph convolutional networks in the case of general irregular graphs. In this study, we design a deep dynamic residual graph convolutional network (DynaResGCN) based on our novel dynamic dilated aggregation mechanisms and a unified end-to-end encoder-decoder-based framework to detect overlapping communities in networks. The deep DynaResGCN model is used as the encoder, whereas we incorporate the Bernoulli-Poisson (BP) model as the decoder. Consequently, we apply our overlapping community detection framework in a research topics dataset without having ground truth, a set of networks from Facebook having a reliable (hand-labeled) ground truth, and in a set of very large co-authorship networks having empirical (not hand-labeled) ground truth. Our experimentation on these datasets shows significantly superior performance over many state-of-the-art methods for the detection of overlapping communities in networks.

Background Breastfeeding is known to reduce risk of enteropathogen infections, but protection from specific enteropathogens is not well characterized. Objective To estimate the association between full breastfeeding (days fed breast milk exclusively or with non-nutritive liquids) and enteropathogen detection. Design 2,145 newborns were enrolled in eight sites, of whom 1,712 had breastfeeding and key enteropathogen data through 6 months. We focused on eleven enteropathogens: adenovirus 40/41, norovirus, sapovirus, astrovirus, and rotavirus, enterotoxigenic Escherichia coli (ETEC), Campylobacter spp, and typical enteropathogenic E. coli as well as entero-aggregative E. coli, Shigella and Cryptosporidium. Logistic regression was used to estimate the risk of enteropathogen detection in stools and survival analysis to estimate the timing of first detection of an enteropathogen. Results Infants with 10% more days of full breastfeeding within the preceding 30 days of a stool sample were less likely to have the three E. Coli and Campylobacter spp detected in their stool (mean odds 0.92���0.99) but equally likely (0.99���1.02) to have the viral pathogens detected in their stool. A 10% longer period of full breastfeeding from birth was associated with later first detection of the three E. Coli, Campylobacter, adenovirus, astrovirus, and rotavirus (mean hazard ratios of 0.52���0.75). The hazards declined and point estimates were not statistically significant at 3 months. Conclusions In this large multi-center cohort study, full breastfeeding was associated with lower likelihood of detecting four important enteric pathogens in the first six months of life. These results also show that full breastfeeding is related to delays in the first detection of some bacterial and viral pathogens in the stool. As several of these pathogens are risk factors for poor growth during childhood, this work underscores the importance of exclusive or full breastfeeding during the first six months of life to optimize early health., Accepted for publication in AJCN

Additional file 1: Supplemental Tables and figures. Supplemental Table 1. Composition of diets with calorie values for children with severe acute malnutrition (SAM) during rehabilitation. Supplemental Table 1A. Composition of liquid diets during acute phase of treatment of children with SAM. Supplemental Table 1B. Composition of semi-solid diets during rehabilitation phase of treatment for SAM. Supplemental Table 2. Specification of emollient (sunflower seed oil). Supplemental Table 3. 24-h Food Intake Chart. Supplemental Table 4. Lower and upper limits of detection provided by Metabolon for their free fatty acid analysis platform in ug/mL. Supplemental Table 5. Fatty acid measurements where an analyte was not detected. Suppplemental Table 6. Fatty acid measurements where an analyte was below the limit of quantitation (BLOQ). Supplemental Table 7. Fatty acid measurements where an analyte was above the limit of quantitation (ALOQ). Supplemental Table 8. Summary of out-of-range measurements of fatty acid levels. Supplemental Table 9. Metadata used for statistical analysis of fatty acid levels in children with severe acute malnutrition. Supplemental Table 10. Summary statistics on fatty acid concentrations. Supplemental Figure 1. t-distributed stochastic neighbor embedding (t-SNE) of absolute concentrations performed on samples from both study timepoints (i.e., day 0 and day 10), plotting data points (each representing a sample) as scatter plots shown for patients ages 2 to 24 months (left column), 2 to

Symmetry is a key feature observed in nature (from flowers and leaves, to butterflies and birds) and in human-made objects (from paintings and sculptures, to manufactured objects and architectural design). Rotational, translational, and especially reflectional symmetries, are also important in drawings of graphs. Detecting and classifying symmetries can be very useful in algorithms that aim to create symmetric graph drawings and in this paper we present a machine learning approach for these tasks. Specifically, we show that deep neural networks can be used to detect reflectional symmetries with 92% accuracy. We also build a multi-class classifier to distinguish between reflectional horizontal, reflectional vertical, rotational, and translational symmetries. Finally, we make available a collection of images of graph drawings with specific symmetric features that can be used in machine learning systems for training, testing and validation purposes. Our datasets, best trained ML models, source code are available online., Comment: Appears in the Proceedings of the 27th International Symposium on Graph Drawing and Network Visualization (GD 2019)

In an $\mathsf{L}$-embedding of a graph, each vertex is represented by an $\mathsf{L}$-segment, and two segments intersect each other if and only if the corresponding vertices are adjacent in the graph. If the corner of each $\mathsf{L}$-segment in an $\mathsf{L}$-embedding lies on a straight line, we call it a monotone $\mathsf{L}$-embedding. In this paper we give a full characterization of monotone $\mathsf{L}$-embeddings by introducing a new class of graphs which we call "non-jumping" graphs. We show that a graph admits a monotone $\mathsf{L}$-embedding if and only if the graph is a non-jumping graph. Further, we show that outerplanar graphs, convex bipartite graphs, interval graphs, 3-leaf power graphs, and complete graphs are subclasses of non-jumping graphs. Finally, we show that distance-hereditary graphs and $k$-leaf power graphs ($k\le 4$) admit $\mathsf{L}$-embeddings.

In this paper we describe a system for visualizing and analyzing worldwide research topics, {\tt rtopmap}. We gather data from google scholar academic research profiles, putting together a weighted topics graph, consisting of over 35,000 nodes and 646,000 edges. The nodes correspond to self-reported research topics, and edges correspond to co-occurring topics in google scholar profiles. The {\tt rtopmap} system supports zooming/panning/searching and other google-maps-based interactive features. With the help of map overlays, we also visualize the strengths and weaknesses of different academic institutions in terms of human resources (e.g., number of researchers in different areas), as well as scholarly output (e.g., citation counts in different areas). Finally, we also visualize what parts of the map are associated with different academic departments, or with specific documents (such as research papers, or calls for proposals). The system itself is available at \url{http://rtopmap.arl.arizona.edu/}.

A monotone drawing of a planar graph $G$ is a planar straight-line drawing of $G$ where a monotone path exists between every pair of vertices of $G$ in some direction. Recently monotone drawings of planar graphs have been proposed as a new standard for visualizing graphs. A monotone drawing of a planar graph is a monotone grid drawing if every vertex in the drawing is drawn on a grid point. In this paper we study monotone grid drawings of planar graphs in a variable embedding setting. We show that every connected planar graph of $n$ vertices has a monotone grid drawing on a grid of size $O(n)\times O(n^2)$, and such a drawing can be found in O(n) time.

Input use efficiency and productivity of different sizes of farms producing FIYV Boro in some selected areas of Brahman Baria district were investigated by employing primarily a Cobb-Douglas production function. Returns to scale and farmer's capability of producing at the least cost level were statistically tested. Farm size and productivity relationships were found to be positive. Boro production was characterized by increasing return to scale for only the medium farms. Few inputs were used in Boro production at the least cost combination level. Adequate extension services including application of right quantity of inputs at right times may help achieve efficiency in input use and improve profitability.

The study was conducted to investigate the effect of different levels of protein supplements on fertility and assess the fertility gene (BMP15 and GDP9) in sheep of Bangladesh. Thirty six ewes fed two protein level (Treatment 1, T1 and treatment 2, T2) and no protein supplement ration (control, T0). Polymerase chain reactions (PCR) were performed from blood cell DNA following standard protocols. PCR products were sequenced using Sanger sequencing and MEGA6 software were used for molecular analysis. The conception rate of ewes was significantly higher (P < 0.05) in the treated group than the control. Gestation period and lambing interval were significantly higher (worse) in control (154.00 ± 0.57 and 186.00 ± 1.24 days) than treated groups. Birth weight of lamb ranged from 1.07 to 1.65 kg and higher birth weight was found in T2 than other groups. Sequence alignment of BMP15 and GDF9 gene in indigenous sheep showed 100% homology with other sheep breed. As no mutation was observed and apparently these genes have no effect on ewe fertility. However, nutritional supplements had a positive impact on reproductive performance. The phylogenetic inferences have close relations of Bangladeshi sheep with other breed elsewhere Refereed/Peer-reviewed