Classification of Echocardiogram View using A Convolutional Neural Network

The standard views in echocardiography capture distinct slices of the heart which can be used to assess cardiac function. Determining the view of a given echocardiogram is the first step for analysis. To automate this step, a deep network of the ResNet-18 architecture was used to classify between six standard views. The network parameters were pre-trained with the ImageNet database and prediction quality was assessed with a visualization tool known as gradient-weighted class activation mapping (Grad-CAM). The network was able to distinguish between three parasternal short axis views and three apical views to ~99\% accuracy. 10-fold cross validation showed a 97\%-98\% accuracy for the apical view subcategories (which included apical two-, three-, and four- chamber views). Grad-CAM images of these views highlighted features that were similar to those used by experts in manual classification. Parasternal short axis subcategories (which included apex level, mitral valve level, and papillary muscle level) had accuracies of 54\%-73\%. Grad-CAM images illustrate that the network classifies most parasternal short axis views as belonging to the papillary muscle level. Likely more images and incorporating time-dependent features would increase the parasternal short axis view accuracy. Overall, a convolutional neural network can be used to reliably classify echocardiogram views.

Workplace and human resource safety monitoring using internet of things

The subject of safety and security of Human Resource (HR) of corporate organizations is a major concern due to the sudden rise in crimes, accidents and various hazards associated with workplaces and the society in recent times. This paper proposes a system for monitoring HR activities and movements in the workplace in real time. The proposed system employs the use of Internet of Things (IoT) wearable devices which are made up of Arduino Uno microcontroller, wireless Radio Frequency Sensors (RFS), Radio Frequency Identification (RFID) tags/readers and Global Positioning System (GPS) modules. The system aims at tracking, locating and keeping the log of the activities and movements of employees at any instant thereby providing information required by employers and security agencies to ensure timely intervention in case of emergency and urgent evacuation.

Estimation of the number of clusters on d-dimensional sphere

Spherical data is distributed on the sphere. The data appears in various fields such as meteorology, biology, and natural language processing. However, a method for analysis of spherical data does not develop enough yet. One of the important issues is an estimation of the number of clusters in spherical data. To address the issue, I propose a new method called the Spherical X-means (SX-means) that can estimate the number of clusters on d-dimensional sphere. The SX-means is the model-based method assuming that the data is generated from a mixture of von Mises-Fisher distributions. The present paper explains the proposed method and shows its performance of estimation of the number of clusters.

Development process of multiagent system for glycemic control of intensive care unit patients

This article presents the entire process of developing an agent-based system for the glycemic control of patients in the Intensive Care Unit (ICU). The agent’s goal is to monitor and recommend treatment to keep the patient’s blood glucose within the target range, avoiding complications in the health of patients and even decreasing rates of morbidity and mortality in the ICU. The process of developing the agent-based solution was presented, starting from the understanding of the problem, including a brief review of the literature, going through the pre-project and modelling through the Tropos methodology, until the implementation. The agent inference mechanism is based on production rules and intuitionistic fuzzy logic. An illustration of use, with the collaboration of a specialist intensive care physician, shows how agents behave in a real situation of monitoring and controlling the blood glucose of patients admitted to the ICU, interacting with all elements of the proposed architecture. Finally, feedback from health professionals indicate the system can assist in the glycemic control of patients in the ICU having advantages over traditional monitoring systems.

A study of quality prediction for large-scale open source software projects

Open source software (OSS) has seen remarkable progress in recent years. Moreover, OSS usage in corporate information systems has been increasing steadily; consequently, the overall impact of OSS on the society is increasing as well. While product quality of enterprise software is assured by the provider, the deliverables of an OSS are developed by the OSS developer community; therefore, their quality is not guaranteed. Thus, the objective of this study is to build an artificial-intelligence-based quality prediction model that corporate businesses could use for decision-making to determine whether a desired OSS should be adopted. We define the quality of an OSS as “the resolution rate of issues processed by OSS developers as well as the promptness and continuity of doing so.” We selected 44 large-scale OSS projects from GitHub for our quality analysis. First, we investigated the monthly changes in the status of issue creation and resolution for each project. It was found that there are three different patterns in the increase of issue creation, and three patterns in the relationship between the increase in issue creation and that of resolution. It was confirmed that there are multiple cases of each pattern that affect the final resolution rate. Next, we investigated the correlation between the final resolution rate and that for a relevant number of months after issue creation. We deduced that the correlation coefficient even between the resolution rate in the first month and the final rate exceeded 0.5. Based on these analysis results, we conclude that the issue resolution rate in the first month once an issue is created is applicable as knowledge for knowledge-based AI systems that can be used to assist in decision-making regarding OSS adoption in business projects.

Decision branch joint venture ex-fold T-z re-validation

Comparing classifier performances may seem a banal affair but makes a side show in machine learning. Usually the paired t-test is used. It requires that two classifiers were run simultaneously or this was simulated. This is not always possible and then entails creating a superstructure only for that purpose. However, the utility of t-test in the given context is altogether doubted. The literature on alternatives is much involved. This does not measure up to the scale of the issue. In this paper the topics in connection with accuracy calculation are surveyed once more, emphasizing the result variation. The known technique of multifold cross-validation is exemplified. A simplified methodology for comparison of classifier performances is proposed. It is based on the accuracy mean and variance and calculating differences between objects defined in these terms. It is being applied to the naive Bayesian and decision tree classifiers implemented on different platforms. The lazy learning approach, applicable to decision trees in discrete domains, is closely followed with an imposition of how it can be improved. Examples are given from the field of health diagnostics.

Comparison of Centralized and Distributed Intelligent Particle Multi-Swarm Optimization on Search Performance

In recent years, the technology of particle swarm optimization (PSO) is expanding remarkably. Especially, the technical development of particle multi-swarm optimization (PMSO) attracts attention, and it is expected to handle complex optimization problems. In this paper, we propose two kinds of search methods of PMSO for pattern classification. The crucial idea, here, is how to handle the given parity problems by using these search methods of centralized and distributed intelligent particle multi-swarm optimization (i.e., CIPMSO and DIPMSO). Due to accomplish the hard task of obtaining the high-performance and high-efficiency of PMSO technology, many computer experiments are carried out to handle the 2-bit, 3-bit and 4-bit parity problems under different search situations. Therefore, the obtained experimental results are analyzed and compared, respectively, the search performance and characteristics of the search methods of both CIPMSO and DIPMSO are clarified. Based on the obtained information and know-how, it will further improve the search efficiency and act in conformity of PMSO technology.

Investigation of differential evolution and particle swarm optimization in search performance

The purpose of this study is to clarify the search performance of differential evolution (DE) and particle swarm optimization (PSO) technologies for instinctively understanding the specificity of the used search methods. Due to achieve the task, here, the several search methods of both, i.e. DE/rand/1, DE/rand/2, DE/best/1, DE/best/2, the PSO, PSOIW, and CPSO, which are implemented in this paper. Therefore, many computer experiments are carried out for handling the given four benchmark problems. Through the analysis of the obtained experimental data, the detail search performance and characteristics of them are observed and compared, respectively. From the obtained results, it is found that the search methods of DE/best/1 and the PSO relatively have better search performance. Based on the findings and know-how, they can provide some important reference and key hint for encouraging development and improvement of both DE and PSO technologies in the near future. And as the applicative examples, the PSO is used to handle typical 2-bit and 3-bit parity problems for pattern classification.

Hybrid approaches to feature subset selection for data classification in high-dimensional feature space

This paper proposes two hybrid feature subset selection approaches based on the combination (union or intersection) of both supervised and unsupervised filter approaches before using a wrapper, aiming to obtain low-dimensional features with high accuracy and interpretability and low time consumption. Experiments with the proposed hybrid approaches have been conducted on seven high-dimensional feature datasets. The classifiers adopted are support vector machine (SVM), linear discriminant analysis (LDA), and K-nearest neighbour (KNN). Experimental results have demonstrated the advantages and usefulness of the proposed methods in feature subset selection in high-dimensional space in terms of the number of selected features and time spent to achieve the best classification accuracy.

Properly initialized Bayesian Network for decision making leveraging random forest

Bayesian network is one of major methods for probabilistic inference among items. But if it contains particular targeting node and other explanatory nodes for decision making, for example how to select suitable appealing keywords to make customers like a product, edges around the target should be counted with more importance than those among others while constructing the network. In order to achieve this adjustment, this study proposes to configure initial state consisting of a few nodes and their edges connected with the target. The initial state is obtained by leveraging Random forest which is a proven method for decision making. Initial nodes are extracted by measuring mean decrease of Gini coefficient calculated with decision trees of Random forest. Initial edges are designated by comparing Kullback-Leibler divergences of conditional probability distribution among nodes which are corresponding to edge directions. Through an actual experiment, this method is confirmed to be effective for adjusting Bayesian network in decision making. This approach is especially useful for business scenes, such as selecting preferable keywords for appealing products over SNS.