Supervised Filter Learning for Coronary Artery Vesselness Enhancement Diffusion in Coronary CT Angiography Images

Abstract

In medical imaging, vesselness diffusion is usually performed to enhance the vessel structures of interest and reduce background noises, before vessel segmentation and analysis. Numerous learning-based techniques have recently become very popular for coronary artery filtering due to their impressive results. In this work, a supervised machine learning method for coronary artery vesselness diffusion with high accuracy and minimal user interaction is designed. The fully discriminative filter learning method jointly learning a classifier the weak learners rely on and the features of the classifier is developed. Experimental results demonstrate that this scheme achieves good isotropic filtering performances on both synthetic and real patient Coronary Computed Tomography Angiography (CCTA) datasets. Furthermore, region growing-based segmentation approach is performed over filtered images obtained by using different schemes. The proposed diffusion scheme is able to achieve higher average performance measures (87.8% ± 1.5% for Dice, 86.5% ± 1.3% for Precision and 88.5% ± 2.6% for Sensitivity). In conclusion, the developed diffusion method is capable of filtering coronary artery structures and suppressing nonvessel tissues, and can be further used in clinical practice as a real-time CCTA images preprocessing tool.

Keywords

• Computed tomography angiography
• Coronary artery filtering
• Vesselness enhancement
• Machine learning
• Vessel segmentation