الفهرس 
Chapter 1 - IntroductionOnly 14 pages are availabe for public view
 |  | from | 202 |  |  |
| | | from | 202 | | |
Abstract
In the last decades, there is a high demand for monitoring system that help clinicians as well as patients in detection, prediction and monitoring of various diseases remotely. from the other hand, most hospitals are moving towards replacing the traditional infrastructure with smart systems to maximize the utilization of information and communication technology. Smart health is expected to significantly improve the quality of service in the healthcare sector. The intensive care unit (ICU) is a special department in the health care sector that typically helps people recover from life-threatening injuries and illnesses. Patients in the ICU require consistent supervision from medical staff and caregivers to ensure the stability of their health. Therefore, early- and reliable-prediction tools for sensitive medical conditions would be useful caregiving aids. This thesis concentrates on diseases prediction and detection for various problems that related to ICU patients include (1) mortality prediction, (2) COVID-19 detection and monitoring, (3) Sepsis prediction. In the first section our thesis, the framework focused on utilizing the machine learning models in predicting mortality among patients inside ICU. Early ICU mortality prediction is crucial for identifying patients who are at a greater risk of dying and for providing suitable interventions to save the lives of these patients. Accordingly, if it is predicted early enough that a patient is likely to die, they will receive appropriate and timely medical services. In recent decades, various severity scores and machine-learning models have been developed for early mortality prediction; however, it remains to be an open challenge. In this thesis, we propose the addition of an advanced step in the mortality prediction process for ICU patients; it presents an accurate, medically intuitive, and robust stacking ensemble model. This model fuses the decisions of five different classifiers, including linear discriminant analysis, decision tree, multilayer perceptron, k-nearest neighbor, and logistic regression. The framework has been evaluated using 10,664 patients from the medical information mart for intensive care (MIMIC III) benchmark database by collecting the first 6, 12, and 24 hours of ICU data of each patient. The results of an analysis and statistical tests indicate that the performance of our model was comparable to those of state-of-the-art models in terms of accuracy (94.4%), F1 score (93.7%), precision (96.4%), recall (91.1%), and area under the receiver operator characteristic (ROC) curve (93.3%). The experimental results demonstrate the ability and efficiency of the proposed method to improve the ICU mortality prediction performance. In the second section, deep learning models and multitask learning have been utilized in building medically intuitive model for Sepsis prediction before the onset time. Sepsis is a life-threaten disease that associated with organ dysfunction. It occurs due to dysregulated body response to infection. Although it is difficult to identify sepsis at early stage, the delay in identification result in high mortality rate. Developing prognostic tools for sepsis prediction has been the focus of various studies in the last decades. However, most of them relied on a few numbers of features, which may not be sufficient to predict sepsis in much cases. Therefore, in this section, we concentrate on building predictive model for sepsis. Initially, both multi-objective genetic approach (NSGA II) and the artificial neural network (ANN) are run concurrently to extract the optimal feature subset from patient’s features. Next, multilayer deep learning model build based on the optimal feature set. The first layer (classification layer) is a predictive model for sepsis, an ensemble stacking neural network model used to predict patient with sepsis. ICD-9 code used to define patients with sepsis in the Medical Information Mart for Intensive Care (MIMIC)-III dataset. The second layer (regression layer) is multitask deep learning model used to identify the onset time for sepsis, in addition to the blood pressure in the onset time. This model utilizes the generalization ability of MTL in the prediction of clinical related tasks. . Our proposed model achieved superior performance over the state of the art, AUC= 0.901 for classification AUC= 0.831 for regression. The third section concentrated on Coronavirus (COVID-19) prediction and monitoring as it is a new virus of viral pneumonia. It can outbreak in the world through person-to-person transmission. Although several medical companies provide cooperative monitoring healthcare systems, these solutions lack offering of the end-to-end management of the disease. In this section, we aim to develop new approach that could bridge the current gap between current technologies and healthcare systems. The wireless body area network, cloud computing, fog computing, and clinical decision support system are integrated to provide a comprehensive and complete model for disease detection and monitoring. By monitoring a person with COVID-19 in real time, physicians can guide patients with the right decisions. The proposed framework has three main layers (i.e. patient layer, cloud layer, and hospital layer). In the patient layer, patient is tracked through a set of wearable sensors and mobile app. In the cloud layer, a fog network architecture is proposed to solve the issues of storage and data transmission. In the hospital layer, we propose a convolutional neural network-based deep learning model and transfer learning to build classification model for COVID-19 detection based on patient’s X-ray scan images. The proposed model achieved promising results compared to the state of the art (i.e. accuracy of 97.95% and specificity of 98.85%). Our framework is a useful application where we expect significant effects on COVID-19 proliferation and considerable lowering in healthcare expenses