Session P7F.2

Automatic Generation of Panoramic M-Mode from 2D Echocardiograms for Measurements Verification

A Amir*, S Ebadollahi

IBM Almaden Research Center
San Jose, CA, USA

In recent years echocardiography has become the most commonly used tool in diagnosis of heart disease. Its ability to provide continuous view of both structure and motion, to take measurements and to verify functionality using blood and tissue Doppler are ever expanding and becoming more accurate. Yet, the outcome of these tests greatly depend on accurate manual positioning of the sensor as well as of semi-automatic rulers and markers for estimation of volume, tissue thickness and wall motion. In particular, some of these measurements are taken in the one-dimensional M-mode scan.
We develop an automatic system for validation of M-mode measurements. Given a recorded echocardiogram, the standard 2D mode views are analyzed, and a family of panoramic M-modes is synthesized. Next the ECG signal is detected in the echo video and is used for temporal segmentation of heart cycles and timing of measurements. We then measure quantities from the panoramic M-mode and use those to validate the corresponding measurements taken by a cardiologist. We further extract other echo attributes for indexing and search. These allow us to search for patients with similar cases based on similarity of echocardiograms.
In the talk we will first present the algorithms and processing steps necessary to parse the video and to identify the required segments. Then we will show how we construct the M-mode panoramic and will discuss their advantages and disadvantages compared to manually captured M-mode. While machine M-mode are captured at higher temporal frequency, panoramic M-mode allow automatic correction for spatial location of central, maximal axis and provide an automatic way to alert for inconsistent measurements.

(Abstract Control Number: 235)