Improved interventional X-ray appearance

Depth cues are an essential part of navigation and device positioning tasks during clinical interventions. Yet, many minimally-invasive procedures, such as catheterizations, are usually performed under X-ray guidance only depicting a 2D projection of the anatomy, which lacks depth information. Previous attempts to integrate pre-operative 3D data of the patient by registering these to intra-operative data have led to virtual 3D renderings independent of the original X-ray appearance and planar 2D color overlays (e.g. roadmaps). A major drawback associated to these solutions is the trade-off between X-ray attenuation values that is completely neglected during 3D renderings, and depth perception not being incorporated into the 2D roadmaps. This paper presents a novel technique for enhancing depth perception of interventional X-ray images preserving the original attenuation appearance. Starting from patient-specific pre-operative 3D data, our method relies on GPU ray casting to compute a colored depth map, which assigns a predefined color to the first incidence of gradient magnitude value above a predefined threshold along the ray. The colored depth map values are carefully integrated into the X-Ray image while maintaining its original grey-scale intensities. The presented method was tested and analysed for three relevant clinical scenarios covering different anatomical aspects and targeting different levels of interventional expertise. Results demonstrate that improving depth perception of X-ray images has the potential to lead to safer and more efficient clinical interventions.