Poster 04

The most common method for detecting coronary artery stenosis is interventional coronary angiography (ICA). However, 2-D angiogra- phy has limitations because it displays complex 3-D structures of arteries as 2-D X-ray projections. To overcome these limitations, 3-D models or tomographic images of the arterial tree can be reconstructed from 2-D projections. The 3-D modeling process of the arterial tree requires accurate acquisition geometry since in many ICA acquisitions the patient table is translated to cover the entire area of interest, the original cal- ibrated geometry is no longer valid for the 3-D reconstruction process. This study presents methods for identifying the frames acquired during table translation in an angiographic scene. Spatio-temporal methods based on deep learning were used to identify translated frames. Three different architectures − 3-D convolutional neural network (CNN), bi- directional convolutional long short term memory (CONVLSTM), and fusion of bi-directional CONVLSTM and 3-D CNN − were trained and tested. The combination of CONVLSTM and 3-D CNN surpasses the other two methods and achieves a macro f1-score (mean f1-scores of twoclasses) of 93%.

Postersession 1, Imaging and Image Reconstruction