Computerized characterization of cardiac wall motion is provided. Quantities for cardiac wall motion are determined (34) from a four-dimensional (i.e., 3D + time) sequence of ultrasound data. A processor (12) automatically processes the volume data to locate (28) the cardiac wall through the sequence and calculate (34) the quantity from the cardiac wall position or motion. Various machine learning is used for locating (28) and tracking (30) the cardiac wall, such as using a motion prior learned from training data for initially locating the cardiac wall and the motion prior, speckle tracking, boundary detection, and mass conservation cues for tracking with another machine learned classifier. Where the sequence extends over multiple cycles, the cycles are automatically divided (26) for independent tracking of the cardiac wall. The cardiac wall from one cycle may be used to propagate (32) to another cycle for initializing the tracking. Independent tracking (32) in each cycle may reduce or avoid inaccuracies due to drift.
