The publication is devoted to the development of a new stereofluoroscopy-based method of cardiac navigation, as well as the assessment of results. In recent years, the improved efficiency of cardiac surgery is associated, first of all, with emerged high computer technologies, for example, with the processing of various kinds of video information received in real-time mode. Recently, there has been a tendency to use convolutional neural networks as an aid in diagnosing or monitoring the progress of an operation. The implementation of the cardiac navigation method using radiographic x-ray images is proposed based on these developing tools. The suggested method uses a software object detector based on a neural network and stereofluoroscopic images obtained on an X-ray complex consisting of two X-ray generators. The object detector model is responsible for detecting the catheter's coordinates in the left and right images. Then the third coordinate is calculated to construct the voxel.