Automatic noise reduction in coronary angiography video data by morphological operations

The aim in coronary angiography is the observation of coronary artery structure by the experts. Quantitative analysis and archiving is mostly skipped in clinics since an accurate, automatic and fast artery segmentation is necessary for that. However, high noise in the angio images prevents this. Thus, image processing techniques are first used for denoising from noise and background, and then a new artery segmentation algorithm is suggested. The contrast is enhanced by applying top-hat operator according to wide and narrow arteries and then non artery parts are deleted by applying an entropy-based threshold. Finally, artery structure is extracted at 98% accuracy compared to manually drawn ground truth. This algorithm is implemented as a Matlab-Simulink model with video input working at real time speed.