الفهرس 
Smart Healthcare ParadigmsOnly 14 pages are availabe for public view
 |  | from | 143 |  |  |
| | | from | 143 | | |
Abstract
It was the aim of this thesis to contribute to the growing interest in smart health and to the ongoing research in this area by exploring the ICU health care system. The general purpose was to study different strategies to improve monitoring in the ICU health care system in order to identify some of the key issues and conditions in the creation of smart health.
For the purpose of the evaluation of smart healthcare and enhancing health care monitoring system, a combination of different methods was used, and the research was guided by a philosophy of triangulation, that is, an application and combination of several research methodologies in the study of the same phenomenon. Smart health has a high priority in most centuries’ health care system. However, regardless of the high official priority and several years of experience, smart health is making slow progress. Even so, it seems that smart health is here to stay. Smart health is an upgraded family- and community-oriented primary care, supported by flexible hospital services. Many politicians and policy makers are convinced that such an integrated system will assure both high quality and a cost-effective health care.
A comparative study was made in this thesis to analyze the recent ML approaches used in smart health, and the results showed that machine learning approaches can help improve smart health. Machine learning approaches are used in many smart health applications, for example, Glaucoma diagnosis, Alzheimer’s disease, bacterial sepsis diagnoses, ICU readmissions, and cataract detection. Also a smart health system development pipeline was introduced. That pipeline includes data acquisition, networking and computing technologies, data security and privacy, data processing, and data dissemination.
The study indicated that ICU equipment contains many sensors and devices that collect an enormous amount of data on each critically ill patient on daily basis, where the smart health pipeline can be applied. In the case of ICU, only the data processing part were handled in this thesis. Another comparative study was made in this thesis to study and analyze the recent ML and CI used in ICU data analytics, where ML techniques proven to be a pioneer method for ICU data analysis, and that many ML can help with ICU monitoring, ICU mortality prediction, sepsis prediction, and many other applications.
The thesis also showed that Sepsis prediction is one of the most life-threatening disease, which controls mortality rate inside the ICU. Early prediction of sepsis will help decrease that mortality rate. Two different datasets were used. First The MIMIC III dataset was used, where only the bedside data were selected; to help predict sepsis, and SVM was used as a supervised machine learning technique. An accuracy of 99% was achieved. Second The PhysioNet/CinC challenge 2019’s dataset A, and dataset B were used, where 14 features were selected; to help in sepsis prediction. A comparative study was made to compare between different ML techniques used for sepsis prediction with the use of the PhysioNet/CinC challenge 2019’s datasets.
Thus, Sepsis prediction is a challenging problem in spite of many years of research efforts because its symptoms is often unclear until later stages. In this study, a methodology is proposed for the sepsis predication. This methodology was conducted from different experiments to help decide which is a better ML approach. Random forest ensemble classifier proved to be the best ML approach with an accuracy of 99% on dataset A, and accuracy of 98% on dataset B.
7.2 Conclusion
Smart health is a developing and exceedingly critical research field with a possibly noteworthy effect on the conventional healthcare industry. A systematic pipeline of data processing is accommodated for conventional smart health, covering data acquisition, networking and computing technologies, data security and privacy, data processing, and data dissemination. In spite of numerous chances and methodologies for data analytics in healthcare presented in this work, there are numerous different bearings to be investigated concerning different aspects of healthcare data such as quality, privacy and so on. Sepsis prediction is a challenging problem and remains so despite many years of research and development efforts because its manifestation is often unclear until later stages. The system constructed consists of Acquisition phase which includes reading data from the ICU monitoring equipment, then that data is processed and analyzed using the sepsis prediction model (constructed using machine learning techniques), and finally a prediction is made.
The first methodology proposed for constructing a machine learning model in this research for real-time sepsis prediction in intensive care unit for the critically ill patients using vital signs only (Bedside data) showed a prediction performance with 99% as accuracy. The second methodology proposed in this research for real-time sepsis prediction in intensive care unit for the critically ill people showed a prediction performance with 98% as accuracy.
7.3 Future Work
In future work:
• The proposed methodology can be applied to different ICU datasets in order to find a generic model that can be used for real-time monitoring of the ICU patients.
• Build a dataset collected from the Egyptian hospitals, and then apply the proposed methodology on this dataset.
• Convert this system into a mobile based application.
• Work on corona virus dataset instead of sepsis.