LAUSR

The quantitative assessment of the location and size of myocardial infarction has important implications for the diagnosis and treatment of ischemic cardiac diseases. In particular, the tasks of optical flow estimation are of increasing interest in the motion analysis in the field of computer vision. In this paper, we propose a deep learning constrained framework, integrating optical flow features for the classification and localization of myocardial infarction from medical image sequences. The framework is composed of two stages. In the first stage, a stacked denoising autoencoder allows for the extraction of the intensity and motion characteristics from images. Thereafter, a support vector machine model is employed to predict the anomaly scores of each input. Initial experiments are performed with two-dimensional cardiac MRI sequences.