Your search

The use of learning analytics (LA) in real-world educational applications is growing very fast as academic institutions realize the positive potential that is possible if LA is integrated in decision making. Education in schools on public health need to evolve in response to the new knowledge and th...

Association Rule Mining by Aprior method has been one of the popular data mining techniques for decades, where knowledge in the form of item-association rules is harvested from a dataset. The quality of item-association rules nevertheless depends on the concentration of frequent items from the input dataset. When the dataset becomes large, the items are scattered far apart. It is known from previous literature that clustering helps produce some data groups which are concentrated with frequent items. Among all the data clusters generated by a clustering algorithm, there must be one or more clusters which contain suitable and frequent items. In turn, the association rules that are mined from such clusters would be assured of better qualities in terms of high confidence than those mined from the whole dataset. However, it is not known in advance which cluster is the suitable one until all the clusters are tried by association rule mining. It is time consuming if they were to be tested by brute-force. In this paper, a statistical property called prior probability is investigated with respect to selecting the best out of many clusters by a clustering algorithm as a pre-processing step before association rule mining. Experiment results indicate that there is correlation between prior probability of the best cluster and the relatively high quality of association rules generated from that cluster. The results are significant as it is possible to know which cluster should be best used for association rule mining instead of testing them all out exhaustively.

Nowadays, the increasing number of medical diagnostic data and clinical data provide more complementary references for doctors to make diagnosis to patients. For example, with medical data, such as electrocardiography (ECG), machine learning algorithms can be used to identify and diagnose heart disease to reduce the workload of doctors. However, ECG data is always exposed to various kinds of noise and interference in reality, and medical diagnostics only based on one-dimensional ECG data is not trustable enough. By extracting new features from other types of medical data, we can implement enhanced recognition methods, called multimodal learning. Multimodal learning helps models to process data from a range of different sources, eliminate the requirement for training each single learning modality, and improve the robustness of models with the diversity of data. Growing number of articles in recent years have been devoted to investigating how to extract data from different sources and build accurate multimodal machine learning models, or deep learning models for medical diagnostics. This paper reviews and summarizes several recent papers that dealing with multimodal machine learning in disease detection, and identify topics for future research.

The adoption of computer-aided diagnosis and treatment systems based on different types of artificial neural networks (ANNs) is already a reality in several hospital and ambulatory premises. This chapter aims to present a discussion focused on the challenges and trends of adopting these computerized systems, highlighting solutions based on different types and approaches of ANN, more specifically, feed-forward, recurrent, and deep convolutional architectures. One section is focused on the application of AI/ANN solutions to support cardiology in different applications, such as the classification of the heart structure and functional behavior based on echocardiography images; the automatic analysis of the heart electric activity based on ECG signals; and the diagnosis support of angiogram images during surgical interventions. Finally, a case study is presented based on the application of a deep learning convolutional network together with a recent technique called transfer learning to detect brain tumors using an MRI images data set. According to the findings, the model has a high degree of specificity (precision of 0.93 and recall of 0.94 for images with no brain tumor) and can be used as a screening tool for images that do not contain a brain tumor. The f1-score for images with brain tumor was 0.93. The results achieved are very promising and the proposed solution may be considered to be used as a computer-aided diagnosis tool based on deep learning convolutional neural networks. Future works will consider other techniques and compare them with the one presented here. With the comprehensive approach and overview of multiple applications, it is valid to conclude that computer-aided diagnosis and treatment systems are important tools to be considered today and will be an essential part of the trend of personalized medicine over the coming years.

The area of clinical decision support systems (CDSS) is facing a boost in research and development with the increasing amount of data in clinical analysis together with new tools to support patient care. This creates a vibrant and challenging environment for the medical and technical staff. This chapter presents a discussion about the challenges and trends of CDSS considering big data and patient-centered constraints. Two case studies are presented in detail. The first presents the development of a big data and AI classification system for maternal and fetal ambulatory monitoring, composed by different solutions such as the implementation of an Internet of Things sensors and devices network, a fuzzy inference system for emergency alarms, a feature extraction model based on signal processing of the fetal and maternal data, and finally a deep learning classifier with six convolutional layers achieving an F1-score of 0.89 for the case of both maternal and fetal as harmful. The system was designed to support maternal–fetal ambulatory premises in developing countries, where the demand is extremely high and the number of medical specialists is very low. The second case study considered two artificial intelligence approaches to providing efficient prediction of infections for clinical decision support during the COVID-19 pandemic in Brazil. First, LSTM recurrent neural networks were considered with the model achieving R2=0.93 and MAE=40,604.4 in average, while the best, R2=0.9939, was achieved for the time series 3. Second, an open-source framework called H2O AutoML was considered with the “stacked ensemble” approach and presented the best performance followed by XGBoost. Brazil has been one of the most challenging environments during the pandemic and where efficient predictions may be the difference in saving lives. The presentation of such different approaches (ambulatory monitoring and epidemiology data) is important to illustrate the large spectrum of AI tools to support clinical decision-making.