Cardiac NMR imaging experiments yield massive amounts of data, thus eliminating manual processing as a practical analysis approach. We describe a novel, fast, and reliable statistical segmentation algorithm for automatically determining left and right ventricular volume. The algorithm first locates that portion of the image containing the ventricles, analyzes the histogram via unsupervised parametric clustering to determine suitable intensity ranges for blood and tissue, and then applies a region growing to determine ventricular area. A similar procedure is followed for each of the numerous anatomic and temporal slices obtained in the experiment, with past results used to guide segmentation of the remaining slices.