Automated analysis approaches for coronary CT angiography

The main purpose of this thesis is to facilitate the automatic lesion reporting and risk stratification in a large cohort of patients and allow automatic follow-up comparison of quantitative parameters for coronary arteries on coronary computed tomography angiography images.We developed an automatic coronary artery tree (CAT) labeling algorithm to identify the anatomical segments for extracted coronary arteries from both right dominant and left dominant cases with an average precision of 91% in comparison with the manual annotations of two experts.A scoring system is developed to assess the CAT extraction quality which measured the quality of the manually refined CATs with higher scores than automatically extracted CATs with an average score as 82.0 (±15.8) and 88.9 (±5.4), respectively on a 100-point scale.Moreover, a model-guided method is developed to detect potential incorrect extractions and automatically improve the extracted CAT using the scoring system to monitor the improved extraction quality.We designed a method to automatically measure the plaque thickness changes between a CAT at baseline and at follow-up which allows the automatic comparison of plaque progression or regression. The average of the calculated plaque thickness difference is the same as the corresponding created value (standard deviation ±0.1mm).