Right Ventricle Segmentation Using Cardiac Magnetic Resonance Images

• Main Author: Rosado-Toro, Jose A.
• Other Authors: Rodriguez, Jeffrey J.
• Language: en_US
• Published: The University of Arizona. 2016
• Subjects: Polar Dynamic Programming; Polar Variance; Right Ventricle; Segmentation; Short Axis; Electrical & Computer Engineering; Four Chamber
• Online Access: http://hdl.handle.net/10150/612450; http://arizona.openrepository.com/arizona/handle/10150/612450

The world health organization has identified cardiovascular disease as the leading cause of non-accidental deaths in the world. The heart is identified as diseased when it is not operating at peak efficiency. Early diagnosis of heart disease can impact treatment and improve a patient's outcome. An early sign of a diseased heart is a reduction in its pumping ability, which can be measured by performing functional evaluations. These are typically focused on the ability of the ventricles to pump blood to the lungs (right ventricle) or to the rest of the body (left ventricle). Non-invasive imaging modalities such as cardiac magnetic resonance have allowed the use of quantitative methods for ventricular functional evaluation. The evaluation still requires the tracing of the ventricles in the end-diastolic and end-systolic phases. Even though manual tracing is still considered the gold standard, it is prone to intra- and inter-observer variability and is time consuming. Therefore, substantial research work has been focused on the development of semi- and fully automated ventricle segmentation algorithms. In 2009 a medical imaging conference issued a challenge for short-axis left ventricle segmentation. A semi-automated technique using polar dynamic programming generated results that were within human variability. This is because a path in a polar coordinate system yields a circular object in the Cartesian grid and the left ventricle can be approximated as a circular object. In 2012 there was a right ventricle segmentation challenge, but no polar dynamic programming algorithms were proposed. One reason may be that polar dynamic programming can only segment circular shapes. To use polar dynamic programming for the segmentation of the right ventricle we first expanded the capability of the technique to segment non-circular shapes. We apply this new polar dynamic programming in a framework that uses user-selected landmarks to segment the right ventricle in the four chamber view. We also explore the use of four chamber right ventricular segmentation to segment short-axis views of the right ventricle.