Angiographic image analysis for the diagnosis of coronary disease in young patients

Abstract

Congenital heart disease (CHD) is the leading cause of birth defects worldwide. In Canada, about 1 in 100 of newborns are with CHD. New methods and protocols to both diagnose and treat CHD are needed. To diagnose or treat newborns or young patients, cardiologists often use percutaneous catheter-based interventions (PCI). PCI is based on using a real-time twodimensional X-ray moving sequence to navigate inside the patient’s heart. This moving sequence will display the artery of the patient during one or more cardiac cycles. During one cardiac cycle, the diameter of a healthy coronary artery (CA) changes: it expands in diastole, when the blood flow in the artery increases, and it shrinks in systole. Capturing these diameter changes is one of the measures used to evaluate the state of the CA as an indicator for a vascular disease. These measurements are usually done manually on X-ray frames or using other image modalities such as ultrasounds. This research sheds new light on an automatic method that can guide the diagnosis of coronary arteries from 2D X-ray sequences. Measuring CA diameter automatically in the X-ray sequence is not a straightforward task. First, a segmentation step is needed to extract CA in 2D X-ray angiograms. Second, the CA is tracked during the moving sequence. Finally, with the segmented and tracked CA, an automatic method measures the diameters of CA at each frame of the sequence and an assessment of diameter variations during the sequence is presented. Each step presents different challenges egarding : the complexity of CA structure, the lack of contrast in the images, the presence of motion artifacts, and the small variations of diameters from one frame to the next. This work introduces new methods for segmentation, tracking and diameter measurements. Our project is held in collaboration with the cardiology department at Sainte Justine’s children hospital. The methods developed in this research have been tested and evaluated on a dataset of X-ray sequences of young patients. Qualitative and quantitative results show the efficiency of the methods from segmenting and tracking to measuring the diameter’s variation from X-ray moving sequence.

Date	29 Jun 2016
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Luc Duong (Supervisor) & Mohamed Cheriet (Co-supervisor)