Reliability of Semiautomated Computational Methods for Estimating Tibiofemoral Contact Stress in the Multicenter Osteoarthritis Study

Figure 2

Multiple evolutionary generations in the CMA-ES optimization approach show a steady march toward a single best alignment (top row), driven by a cost function that incorporates evaluation of agreement between a model silhouette projected onto the radiographic image plane and the bone edge detected from the actual radiograph (bottom row). Each CMA-ES generation includes 70 candidate spatial alignments, with those having the lowest cost function evaluation being used to drive subsequent generations.