Features extraction and analysis methods for sequences of ultrasound images

Abstract

Our principal motivation is to study time sequences of echocardiographic raw data to track specific anatomical structures. First, we show that the image processing can make direct use of the audio signal data, avoiding loss of information and yielding optimal results. Secondly, we develop a strategy which takes a time sequence of raw data as input, computes edges, initiates a segmentation of a pre-selected anatomical structure and uses a deformable model for its temporal tracking. This approach is validated in a real time sequence of ultrasound images of the heart to track the left auricle and the mitral valve.