In this work, MacConaill's classification that the articular surface of the femoral head is better represented by ovoidal shapes rather than purely spherical shapes is computationally tested. To test MacConaill's classification, a surface fitting framework was developed to fit spheres, ellipsoids, superellipsoids, ovoids, and superovoids to computed tomography (CT) data of the femoral proximal epiphysis. The framework includes several image processing and computational geometry techniques, such as active contour segmentation and mesh smoothing, where implicit surface fitting is performed with genetic algorithms. By comparing the surface fitting error statistics, the results indicate that (super)ovoids fit femoral articular surfaces better than spherical or (super)ellipsoidal shapes.