Time filter

Source Type

Ngo T.A.,Vietnam National University of Agriculture | Lu Z.,University of South Australia | Carneiro G.,University of Adelaide
Medical Image Analysis | Year: 2017

We introduce a new methodology that combines deep learning and level set for the automated segmentation of the left ventricle of the heart from cardiac cine magnetic resonance (MR) data. This combination is relevant for segmentation problems, where the visual object of interest presents large shape and appearance variations, but the annotated training set is small, which is the case for various medical image analysis applications, including the one considered in this paper. In particular, level set methods are based on shape and appearance terms that use small training sets, but present limitations for modelling the visual object variations. Deep learning methods can model such variations using relatively small amounts of annotated training, but they often need to be regularised to produce good generalisation. Therefore, the combination of these methods brings together the advantages of both approaches, producing a methodology that needs small training sets and produces accurate segmentation results. We test our methodology on the MICCAI 2009 left ventricle segmentation challenge database (containing 15 sequences for training, 15 for validation and 15 for testing), where our approach achieves the most accurate results in the semi-automated problem and state-of-the-art results for the fully automated challenge. © 2016


PubMed | Vietnam National University of Agriculture, University of South Australia and University of Adelaide
Type: | Journal: Medical image analysis | Year: 2016

We introduce a new methodology that combines deep learning and level set for the automated segmentation of the left ventricle of the heart from cardiac cine magnetic resonance (MR) data. This combination is relevant for segmentation problems, where the visual object of interest presents large shape and appearance variations, but the annotated training set is small, which is the case for various medical image analysis applications, including the one considered in this paper. In particular, level set methods are based on shape and appearance terms that use small training sets, but present limitations for modelling the visual object variations. Deep learning methods can model such variations using relatively small amounts of annotated training, but they often need to be regularised to produce good generalisation. Therefore, the combination of these methods brings together the advantages of both approaches, producing a methodology that needs small training sets and produces accurate segmentation results. We test our methodology on the MICCAI 2009 left ventricle segmentation challenge database (containing 15 sequences for training, 15 for validation and 15 for testing), where our approach achieves the most accurate results in the semi-automated problem and state-of-the-art results for the fully automated challenge.


Tran T.T.H.,University of Hohenheim | Tran T.T.H.,Vietnam National University of Agriculture | Zeller M.,University of Hohenheim | Suhardiman D.,International Water Management Institute
Ecosystem Services | Year: 2016

This study examines the institutional design and actual performance, of payments for ecosystem services (PES) in Vietnam. Taking Payments for Forest Environmental Services Program (PFES Program) implementation in Da Bac district, Hoa Binh province as a case study, it brings to light how PES program design and implementation contributed to the central government's objectives to: (1) involve stakeholders in forest management; (2) reduce the government's budget burden for forest protection; and (3) maintain political control over forest resources. In Vietnam, the PFES Program is implemented in a top-down manner. Participating households act as government-induced forest guards rather than forest owners. Incomplete design at the central-level results in poorer performance at lower levels and, the lack of strategic management makes it difficult to know whether the program has actually improved ecosystem services and forest management. While the PFES Program complements other institutions at the national- and local-levels, some institutional incompatibilities exist in terms of customary practices. It is unlikely, however, that these will develop into an institutional conflict. © 2016 Elsevier B.V.


Luziga C.,Sokoine University of Agriculture | Nga B.T.T.,Vietnam National University of Agriculture | Mbassa G.,Sokoine University of Agriculture | Yamamoto Y.,Yamaguchi UniversityYamaguchi
Acta Histochemica | Year: 2016

Cathepsins B and L are two prominent members of cystein proteases with broad substrate specificity and are known to be involved in the process of intra- and extra-cellular protein degradation and turnover. The propeptide region of cathepsin L is identical to Cytotoxic T-lymphocyte antigen-2α (CTLA-2α) discovered in mouse activated T-cells and mast cells. CTLA-2α exhibits selective inhibitory activities against papain and cathepsin L. We previously demonstrated the distribution pattern of the CTLA-2α protein in mouse brain by immunohistochemistry, describing that it is preferentially localized within nerve fibre bundles than neuronal cell bodies. In the present study we report colocalization of cathepsin L and CTLA-2α by double labeling immunofluorescence analysis in the mouse brain. In the telencephalon, immunoreactivity was identified in cerebral cortex and subcortical structures, hippocampus and amygdala. Within the diencephalon intense colocalization was detected in stria medullaris of thalamus, mammillothalamic tract, medial habenular nucleus and choroid plexus. Colocalization signals in the mesencephalon were strong in the hypothalamus within supramammillary nucleus and lateroanterior hypothalamic nucleus while in the cerebellum was in the deep white matter, granule cell layer and Purkinje neurons but moderately in stellate, and basket cells of cerebellar cortex. The distribution pattern indicates that the fine equilibrium between synthesis and secretion of cathespin L and CTLA-2α is part of the brain processes to maintain normal growth and development. The functional implication of cathespin L coexistence with CTLA-2α in relation to learning, memory and disease mechanisms is discussed. © 2016 Elsevier GmbH


PubMed | Posts and Telecommunications Institute of Technology PTIT Hanoi, Vietnam National University of Agriculture, Hanoi University of Science and Technology and Hanoi University
Type: | Journal: Waste management (New York, N.Y.) | Year: 2016

The amount of municipal solid waste (MSW) has been increasing steadily over the last decade by reason of population rising and waste generation rate. In most of the urban areas, disposal sites are usually located outside of the urban areas due to the scarcity of land. There is no fixed route map for transportation. The current waste collection and transportation are already overloaded arising from the lack of facilities and insufficient resources. In this paper, a model for optimizing municipal solid waste collection will be proposed. Firstly, the optimized plan is developed in a static context, and then it is integrated into a dynamic context using multi-agent based modelling and simulation. A case study related to Hagiang City, Vietnam, is presented to show the efficiency of the proposed model. From the optimized results, it has been found that the cost of the MSW collection is reduced by 11.3%.


PubMed | Vietnam National University of Agriculture, Yamaguchi University and Sokoine University of Agriculture
Type: Journal Article | Journal: Acta histochemica | Year: 2016

Cathepsins B and L are two prominent members of cystein proteases with broad substrate specificity and are known to be involved in the process of intra- and extra-cellular protein degradation and turnover. The propeptide region of cathepsin L is identical to Cytotoxic T-lymphocyte antigen-2 (CTLA-2) discovered in mouse activated T-cells and mast cells. CTLA-2 exhibits selective inhibitory activities against papain and cathepsin L. We previously demonstrated the distribution pattern of the CTLA-2 protein in mouse brain by immunohistochemistry, describing that it is preferentially localized within nerve fibre bundles than neuronal cell bodies. In the present study we report colocalization of cathepsin L and CTLA-2 by double labeling immunofluorescence analysis in the mouse brain. In the telencephalon, immunoreactivity was identified in cerebral cortex and subcortical structures, hippocampus and amygdala. Within the diencephalon intense colocalization was detected in stria medullaris of thalamus, mammillothalamic tract, medial habenular nucleus and choroid plexus. Colocalization signals in the mesencephalon were strong in the hypothalamus within supramammillary nucleus and lateroanterior hypothalamic nucleus while in the cerebellum was in the deep white matter, granule cell layer and Purkinje neurons but moderately in stellate, and basket cells of cerebellar cortex. The distribution pattern indicates that the fine equilibrium between synthesis and secretion of cathespin L and CTLA-2 is part of the brain processes to maintain normal growth and development. The functional implication of cathespin L coexistence with CTLA-2 in relation to learning, memory and disease mechanisms is discussed.


Tran T.-H.,Hanoi University of Science and Technology | Vo T.-H.,Hanoi University of Science and Technology | Tran D.-T.,Hanoi University of Science and Technology | Le T.-L.,Hanoi University of Science and Technology | Nguyen T.T.,Vietnam National University of Agriculture
International Conference on Advanced Technologies for Communications | Year: 2015

Gesture recognition has important applications in sign language and human-machine interfaces. In recent years, recognizing dynamic hand gesture using multi-modal data has become an emerging research topic. The problem is challenging due to the complex movements of hands and the limitations of data acquisition. In this work, we present a new approach for recognizing hand gesture using motion history images (MHI) [1] and a kernel descriptor (KDES) [2]. We propose to use an improved version of MHI for modeling movements of hand gesture, where MHI is computed on both RGB and depth data. We propose some improvements in patch-level feature extraction for KDES, which is then applied to MHI to represent gesture features. Then SVM classifier is trained for recognizing gestures. Experiments have been conducted on challenging hand gesture data set of CHALEARN contest [3]. An extensive investigation has been done to analyze the performance of both improved MHI and KDES on multi-modal data. Experimental results show the state-of-the-art of our approach in comparison to the results of the contest. © 2014 IEEE.


Tuan N.D.,Hanoi University of Science and Technology | Manh N.Q.,Hanoi University of Science and Technology | Sang D.V.,Hanoi University of Science and Technology | Binh H.T.T.,Hanoi University of Science and Technology | Thuy N.T.,Vietnam National University of Agriculture
Proceedings - 2015 IEEE International Conference on Knowledge and Systems Engineering, KSE 2015 | Year: 2015

Dictionary learning (DL) approach has been successfully applied to many pattern classification problems. Sparse property has played an important role in the success of DL-based classification models. However, the sparsity constraints make the learning problem expensive. Recently, there has been an emerged trend in relaxing the sparsity constraints by using L2-norm constraint. The new approach has shown its advantages in both accuracy and classification time. However, the relationship between the quality of the data and the dictionary learning issues that affect the performance of the system has not been investigated. In this paper, we present a comparative study on non-sparse coding dictionary learning for pattern classification. We then propose a dictionary learning model with a non-sparsity constraint on representation coefficients using L2-norm. Our experimental results on three popular benchmark datasets for image classification show that our proposed model can outperform state-of-the-art models and be a promising approach for dictionary learning based classification. © 2015 IEEE.


Quang N.T.,Hanoi University of Science and Technology | Sang D.V.,Hanoi University of Science and Technology | Thuy N.T.,Vietnam National University of Agriculture | Binh H.T.T.,Hanoi University of Science and Technology
ACM International Conference Proceeding Series | Year: 2015

Detection of buildings in aerial images is an important and challenging task in computer vision and aerial image interpretation. This paper presents an efficient approach that combines Random forest (RF) and a fully connected conditional random field (CRF) on various features for the detection and segmentation of buildings at pixel level. RF allows one to learn extremely fast on big aerial image data. The unary potentials given by RF are then combined in a fully connected conditional random field model for pixelwise classification. The use of high dimensional Gaussian filter for pairwise potentials makes the inference tractable while obtaining high classification accuracy. Experiments have been conducted on a challenging aerial image dataset from a recent ISPRS Semantic Labeling Contest [9]. We obtained state-of-the-art accuracy with a reasonable computation time. © 2015 ACM.


Nguyen L.V.,Vietnam National University of Agriculture
International Journal on Advanced Science, Engineering and Information Technology | Year: 2016

The objective of this study was to understand the change in response of quinoa genotypes to divers salinity stress conditions e.g in controlled (net-house) and in the different saline fields. The pot experiment was conducted in a net-house at Vietnam National University of Agriculture, Hanoi, Vietnam in spring cropping season to characterize the growth and yield of six quinoa genotypes under four NaCl concentrations (0, 10, 20 and 30 dS m-1). At the same time, in Nam Dinh and Hai Phong provinces, two coastal provinces that are most affected by seawater intrusion in the North of Vietnam, same genotypes were studied under two plant densities (20 × 5cm and 50 × 5cm). The results showed that salinity stresses reduced growth and yield characteristics of quinoa plant and varying due to different saline conditions. Plant density of quinoa grown under saline fields was not associated with difference in morphological traits, but might relate to the change in yield characteristics. Salinity stresses reduced plant height, the number of leaves on main stem, the number of branches on plant, head panicle length, dry matter accumulation, 1000-seed weight, individual and grain yield of all quinoa genotypes. However, most of quinoa genotypes produced acceptable yield even under high salt conditions in the field. Among quinoa genotypes, Moradas and Verde adapted well to salt stress conditions with high potential for the number of leaves on main stem, the number of branches on plant, dry matter accumulation and yield than others. These should be recommended varieties for cultivation in saline areas in Vietnam as well as be useful to improve genetic resources in breeding program for salt tolerant quinoa varieties.

Loading Vietnam National University of Agriculture collaborators
Loading Vietnam National University of Agriculture collaborators