The thesis is divided into four additional chapters: (i) background (ii) implementation (iii) experiments and results (iv) and discussion. The background chapter will elaborate further regarding the novel monitoring technique. Additionally, it will introduce the various relevant machine learning concepts. The implementation chapter will explain and reason behind the various choices done regarding the machine learning model.
The experiments and results chapter outlines the experiments conducted and presents the results along with a short analysis where appropriate.
Lastly, the discussion chapter contains a general discussion along with a conclusion of the thesis and suggestions for future work.
Chapter 2
Background
This chapter will attempt to give an introduction to the theory relevant to the thesis. A medical background is presented first, which is intended to give an understanding of the heart and its motion. Next is a part which describes a novel technique and the past research leading which lead to this thesis. The following section, section 2.3 introduces machine learning in general and the three machine learning algorithms used.
2.1 Medical background
2.1.1 The heart
The heart (cardiac), seen in figure 2.1, is a muscular organ roughly shaped as a cone, with a wide basis and a pointed apex (tip). It is located behind and slightly left of the breastbone, with the apex pointing downwards and to the left. The heart contains four chambers: upper left and right atria and lower left and right ventricles. The atria work as blood collection chambers for the ventricles which expel blood out of the heart. Blood enters the right atrium which pumps it downwards to the right ventricle. The right ventricle pumps blood to the pulmonary system, i.e., the lungs, which fills the blood with oxygen and the left atria then collect it. The left atrium pumps it downwards to the left ventricle which pumps it back out to the circulatory system, i.e., the body [58].
2.1.2 The cardiac cycle
The cardiac cycle refers to a complete heartbeat from its generation to the beginning of the next beat. A normal resting heart rate for adults ranges from 60 to 100 beats a minute, which implies that the heart rate lasts from 0.6 to 1 seconds. The cardiac cycle is split into two main phases: the diastole and the systole. The diastole refers to the part of the cycle where the ventricles are relaxed in preparation for refilling with circulating blood, whereas the systole refers to the ejection of blood into an adjacent chamber or vessel. This is seen in figure 2.2. During most of the diastole, blood is passively flowing from the atria to the ventricles. The atria contract at
Figure 2.1: Diagram of the heart [4] showing the four chambers and blood vessel connected to the heart.
the end of diastole, which propels an additional amount of blood into the ventricles. Systole represents the time in which the myocardium (cardiac muscle) contracts. In the first phase of systole, blood is flowing from the atria to the ventricles while the myocardium contracts. In the second phase of systole, blood is expelled out of the ventricles to the body and lungs while the myocardium still contracts. Several other phases within the
Figure 2.2: Figure showing the difference between diastole and systole [63].
cardiac cycle are defined. The two most interesting ones in this project are the mid-systole and the isovolumetric relaxation (IVR) phase and is reasoned in section 3.3. Formally, mid-systole begins after the first of two heart sounds and ends before the second heart sound [52]. Mid-systole is the phase where the valves which keep the blood from flowing towards the lungs and body is are open, and blood is ejected through them. The IVR phase occurs at the end of systole, where the ventricular muscle relaxes and decreases its tension without lengthening so that ventricular volume
2.1. MEDICAL BACKGROUND
remains the same [21]
2.1.3 Cardiac motion
This section aims to provide an overview regarding how the heart moves during the cardiac cycle. The motion will be decomposed into three axes, shown in figure 2.3, because the motion data at hand is decomposed in the same way. Hence, it can give a better understanding of the data.
Figure 2.3: Figure showing the decomposition of cardiac movement into a radial, circumferential and longitudinal axes [3].
Cirumferential motion Circumferential motion is the rotation of the heart, that is, the movement upon itself. Usually, the rotation is measured from the left ventricle and in the apical region (lower part of the heart) and is given in degrees. The rotation of the left ventricle apical region is about 11 ± 5◦ in situ healthy human heart [54], reaching peak velocities at the beginning of ventricular systole [13]. The rotation is in a counterclockwise rotation as seen from the apex. By the middle of ventricular systole, the basal and mid-ventricular segments reverse their circumferential motion, reaching peak clockwise velocities by the end of this phase. The apical segments, however, continues their counter-clockwise circumferential motion until ventricular repolarization occurs.
Longitudinal motion Regarding movement in the longitudinal direction, there is little apical displacement during systole. All segments move downwards towards the left ventricular apex. Before the end of systole, a fast recoil motion occurs, moving all segments upwards. The movement continues until the blood stops being expelled from the left ventricle.
Radial motion The radial motion on the epicardial surface reduces due to the wall thickening and thinning that occur during the phases when the endocardial surface moves inward and outward, respectively.
Consequently, the dominant motion of the left ventricle epicardial apical region, where the motion sensor is placed is circumferential displacement or rotation [54].
2.1.4 Monitoring the heart
The following sections will give an introduction in various techniques used to do cardiac monitoring and their weaknesses.
Figure 2.4: Figure showing the characteristic ECG trace for a cardiac cycle [2, 8].
Electrocardigraphy Electrocardiography (ECG) is the process of monitor-ing the electrical activity of the heart over time. The monitormonitor-ing is achieved by placing electrodes on the skin that detects the tiny electrical changes that arise from the depolarization of the heart muscle during each heart beat. The electrical changes are displayed as a graph that shows voltage over time. A healthy heart will show the characteristic ECG tracing where each wave is named, seen in figure 2.4.
A typical ECG tracing is a repeating cycle of three electrical entities: a P wave, which shows the atrial depolarization (contraction), a QRS complex which shows ventricular depolarization, and finally a T wave which shows ventricular repolarization (expansion/relaxation). To the trained clinician, an ECG conveys a large amount of information about the structure of the heart and the function of its electrical conduction system. Ischemia causes a decreased blood flow to muscle tissues to the heart. This causes a de-polarization of the resting membrane potential of the ischemic region to the resting membrane potential of the normal region, which is manifested as either an elevated or depressed ST segment in the ECG, though not al-ways. There are several drawbacks of using ECG as detection method for ischemia. These includes: (i) The ST-depression or elevation is not present in the ECG trace (ii) Baseline drift (iii) Varying ST-T patterns in ECG of the same patient (iv) Noise (v) Power line interference (vi) Patient dependent