Automatization of 3D reconstruction of coronary arteries from angiography projections using AI-enhanced segmentation techniques

Abstract

Fractional flow reserve (FFR) is one of the clinical diagnostics meas- urements that are performed to assess the physiological significance of stenosis potentially present in coronary arteries. Virtual fractional flow reserve (vFFR) is an alternative non-invasive approach that consists of performing the 3D recon- struction of the patient-specific coronary artery and afterwards applying tech- niques of computational fluid dynamics (CFD) to obtain the vFFR value. Within this paper, a new approach is presented, that combines the AI-enhanced segmen- tation of DICOM images obtained during X-ray angiography (XRA) examination and 3D reconstruction and finite element (FE) mesh generation. This enables an automated reconstruction and numerical simulations of blood flow through pa- tient-specific coronary arteries. The developed software is accurate, executes in a timely manner and is intuitive to use, which makes it a useful tool for the clini- cians to perform hemodynamic analyses of the state of coronary arteries. It can thus provide assistance in the treatment planning that is adapted to the particular patient.