Algorithmic assessment of cardiac viability using magnetic resonance imaging

Gao, Gang (2006) Algorithmic assessment of cardiac viability using magnetic resonance imaging. PhD thesis, University of Glasgow.

Abstract

MRI is a non-invasive imaging method which produces high resolution images of human tissues from inside the human body. Due to its outstanding ability, it is quickly becoming a major tool for medical and clinical studies, including high profile areas such as neurology, oncology, cardiology and etc. MRI technology developed relatively slowly compared to other methods such as x-ray. A decade ago, it took more than 5 minutes to construct an MR image. However more recently, with several significant inventions such as echo planar imaging and steady state free procession techniques, the acquisition time of MRI has significantly reduced. At present, it is possible to capture dozens of MR images in a second. Those techniques are generally called ultra-fast MRI. The fast MR acquisition techniques enable us to extend our studies to the moving tissues such as the myocardium. Using the ultra-fast MRI, multiple images can be acquired during a cardiac cycle allowing the construction of cardiac cinematographic MR images. Cardiac motion can therefore be revealed. Abnormal cardiac motion is often related to cardiac diseases such as ischaemic myocardium and myocardial infarction. With advanced MRI techniques, cardiac diseases can be more specifically defined. For example, the late contrast enhanced MRI highlights acute myocardial infarction. The first-pass perfusion MRI suggests the existence of ischaemic myocardium. At the present time the majority of the analysis of MR images can be performed either qualitatively or quantitatively. The qualitative assessment is an eye-ball assessment of the images on a MRI workstation, which is subjective and inaccurate. The quantitative assessment of MR image relies on the computer technologies of both hardware and software. In recent years, the demands for the quantitative assessment of MR images have increased sharply. Many so-called computer aided diagnosis systems were developed to process data either more accurately or more efficiently. In this study, we developed an algorithmic method to analyse the late contrast enhanced MR images, revealing the so-called hibernating myocardium. The algorithm is based on an efficient and robust image registration algorithm. Using the image registration algorithm, we are able to integrate the static late contrast enhanced MR image with its corresponding cardiac cinematography MR images, and so constructing cardiac CINE late enhanced MR images. Our algorithm was tested on 20 subjects. In each of the subject, the mean left ventricle diastolic volume and systolic volume was measured by planimetry from both the original CINE images and the constructed late enhanced CINE images. The results are: left ventricle diastolic volume (original / constructed) = 206 / 215 ml, p = 0.35. Left ventricle systolic volume (original / constructed) = 129 / 123 ml, p = 0.33. With our algorithm, the cardiac motion and the myocardial infarction can therefore be studied simultaneously to locate the hibernating myocardium which moves abnormally. The accurate location of the hibernating myocardium is important because it could turn into the irreversible myocardial infarction. On the other hand, with proper medical treatment or cardiac surgery, the hibernating myocardium could be revitalised. The experimental results show there are no significant differences between the artificial cine late contrast enhanced MR images and the original cinematography MR images in left ventricle diastolic volume, left ventricle systolic volume. The method therefore appears promising as an improved cardiac viability assessment tool. In addition, we extended the method to a semi-automatic cardiac contour definition algorithm, which has produced a satisfactory result in contour definition for cardiac cinematography MR images from 34 subjects including 20 healthy volunteers and 14 patients. Although it is a semi-automatic method, the diagnosis time could be significantly reduced compared to the manual method. The algorithm was preliminarily tested on 10 first-pass perfusion MR sequences and 10 aortic MR sequences. The experimental results were satisfactory. Although, minor manual correction is required on some occasions, we believe our method could be clinically useful for the study of cardiac cinematography MR images, first-pass perfusion MR images and aortic MR images.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Additional Information: Adviser: Paul Cockshort
• Keywords: Biomedical engineering, Medical imaging
• Date of Award: 2006
• Depositing User: Enlighten Team
• Unique ID: glathesis:2006-74079
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 23 Sep 2019 15:33
• Last Modified: 23 Sep 2019 15:33
• URI: https://theses.gla.ac.uk/id/eprint/74079