Machine learning approaches for early prediction of hypertension [thesis]

Heba Elsayed Mohamed Kandil
Medical imaging is a vital emerging technology that has been studied and used tremendously in the last few decades. Medical imaging techniques enable physicians to capture non-invasive images of structures inside the body (e.g., Blood vessels, tissue, bones) as well as their function (e.g., brain activity). Medical images comprise massive amounts of information that are very useful for understanding the structures and functions of different body organs. Yet, these huge amounts of data are very
more » ... ifficult to be processed by radiologists' or physicians' eyes. Therefore, the need for computerized systems that could store, analyze, and make use of these data is of great importance. The most important value of medical images is to help physicians in predicting, detecting, and diagnosing diseases. Specifically, there are many progressive diseases that may take many years before they are discovered or diagnosed. While some of these diseases are asymptomatic, they could usually start to affect internal human body organs by changing or damaging their structures and/or functionalities. Hypertension is a progressive disease that may take a decade or two before it is discovered or diagnosed. It is called the "Silent Killer" because it may not have any apparent symptoms. Hypertension afflicts one in every three adults and is a leading cause of mortality in 516, 955 patients in the USA. Many factors such as renal dysfunction, high sodium intake, and chronic stress contribute to the development v of hypertension. The chronic elevation of cerebral perfusion pressure (CPP) changes the cerebrovasculature of the brain and disrupts its vasoregulation mechanisms. This cerebral vascular alteration has a severe effect on the human body organs and is a leading cause of cognitive impairment, strokes, dementia, ischemic cerebral injury, and brain lesions. Reported correlations between changes in smaller cerebrovascular vessels and hypertension may be used to diagnose hypertension in its early stages, 10-15 years before the appearance of symptoms such as cognitive impairment and memory loss. Specifically, recent studies hypothesized that changes in the cerebrovasculature and CPP precede the systemic elevation of blood pressure (BP). Currently, sphygmomanometers are used to measure repeated brachial artery pressure to diagnose systemic hypertension after its onset. However, this method cannot detect cerebrovascular alterations that lead to adverse events which may occur prior to the onset of hypertension. The early detection and quantification of these cerebral vascular structural changes could help in predicting patients who are at a high risk of developing hypertension as well as other cerebral adverse events. This may enable early medical intervention prior to the onset of systemic hypertension, potentially mitigating vascular-initiated end-organ damage. The ultimate goal of this dissertation is to develop a novel efficient noninvasive computer-aided diagnosis (CAD) system for the early prediction of hypertension. The developed CAD system analyzes magnetic resonance angiography (MRA) data of human brains gathered over years to detect and track cerebral vascular alterations correlated with hypertension development. This CAD system is able to make decisions based on available data to help physicians on predicting potential hypertensive patients before the onset of the disease. Thus, taking appropriate actions to stop the progress of the disease and mitigating adverse events. To achieve this goal, a set of milestones have been targeted in this project: Milestone 1: Validating the clinical hypothesis that claims the correlation between specific cerebral vascular alterations/changes and the development of hypertension. Milestone 2: Developing an efficient, fast, and accurate segmentation algorithm that is able vi to segment smaller blood vessels of human brains accurately and automatically. Milestone 3: Selecting cerebral features that correlate with hypertension development and can be used to predict the disease before its onset. Milestone 4: Developing, testing, and validating a CAD system that can process MRA data and make decisions to help in the early prediction of hypertension. These milestones have been successfully achieved and a CAD system was developed to predict hypertension before its systematic onset. The function of the CAD system starts with preprocessing the MRA data to eliminate noise effects, correct the bias field effect, reduce the contrast inhomogeneity, and normalize the data, then the cerebral vascular tree of each MRA volume is segmented using a novel 3-D segmentation algorithm that was developed to be automatic, accurate, and adaptive. The segmented vascular tree is then processed to extract and quantify cerebral features that correlate with hypertension development. The cerebral vascular features were carefully selected based on clinical research and are reported to correlate with hypertension development, namely, change in cerebral blood vessels diameters and change in vascular tortuosity. These features are estimated and used to prepare the inputs for the neural networks' classification module. The classification module processes the data (cerebral vascular features) and makes a decision that would help physicians to decide if an individual is having the potential to be hypertensive or not. An MRA dataset has been acquired by the University of Pittsburgh and approved by the institutional review board (IRB) according to the relevant guidelines and regulations. It was used to test the CAD system and the experimental results reported a classification accuracy of ∼ 91% which supports the efficacy of the CAD system components to accurately model and discriminate between normal and hypertensive subjects. Clinicians would benefit from the proposed CAD system to detect and track cerebral vascular alterations over time for people with a high potential of developing hypertension and to prepare appropriate treatment plans to mitigate adverse events. Moreover, the functionality of the developed CAD system has been utilized to study the significance of using mean arterial vii pressure (MAP) in hypertension detection. In addition, it has been used study the impact of the hypertension-related cerebrovascular changes on different compartments of human brains (specifically, the anterior and the posterior regions). viii TABLE OF CONTENTS
doi:10.18297/etd/3823 fatcat:zpgaeryyofhvdguap37b723oay