Automatic left ventricular myocardium segmentation and volumetric classification of scar

Cardiac function is of predominant importance for both prognosis and treatment
of different pathologies such as mitral regurgitation,ischmia,dyssynchrony and
myocarditis. Cardiac behaviour is determined by structural and functional
features. In both cases, the analysis of medical imaging studies requires to detect
and segment the myocardium. Nowadays, magnetic resonance imaging is one of
the most relevant and accurate non-invasive diagnostic tools for cardiac structure
and function. In this, we present deep learning technique stack auto encoder
architecture to segment images from a series of short-axis cardiac magnetic
resonance slices, we propose to tackle the problem of automated left ventricle
segmentation through the application of deep learning technique.It takes as input
raw magnetic resonance imaging images,requires no manual pre processing or
image cropping and is trained to segment the endocardium of the left ventricle,
as well as the center of the left ventricle. The quantitative evaluation is based on
three commaon metrics which are accuracy, dice and sensitivity respectively and
classifciation using random forest classifier. The classifciation is to be done on
the image set to determine whether the myocardium is viable or normal.

MPhil CS 2018-2019 INT Dr Elizabeth Sherly