Visualizing Articular Cartilage Using Expectation Maximization To Compare Segmentation Ofmrimages By Independent Raters

When following the progression of osteoarthritis, it is important to precisely determine the presence or absence of cartilage in joints such as the knee. MRIis are useful for this, but the images are difYcult to segment. We modiYed an expectation maximization algorithm originally described by WarYeld et al. 5 to score the accuracy of segmentation of a series of images by multiple raters. In the process, we also gain information that is used to visualize the most likely location of cartilage.