The most widely used system of grading Legg-Calvé-Perthes disease (LCPD) is still the radiographic grouping by Catterall, although it lacks interobserver reliability, especially in the early stages. The predictive value of Salter-Thompson classification and Herring's "lateral pillar" classification are still being studied. In Part 1, we demonstrated the good reliability and the predictive value of four magnetic resonance imaging (MRI) indices (extension of necrosis, lateral extrusion, physeal involvement, metaphyseal changes) through their correlation to clinical and radiographic conditions of the hips at follow-up. The same good results were obtained by submitting to statistical analysis a second group of 31 patients (French series). On the basis of these statistical studies, a new classification has been proposed. It takes into account the extent of necrosis and two MRI risk signs: lateral extrusion and physeal involvement. The extent of necrosis up to or more than 50% separates two main groups, A and B. The associated MRI risk factors distinguished six classes with different prognoses. Appropriated treatment also is suggested for each class. Our experience on MRI in LCPD led us to draw a pathomorphogenetic model called "packed capsule." According to this biomechanical model, the femoral head is considered a segment of a sphere made of viscoelastic material and hermetically sealed. The deformation of the head depends on the behavior of the necrotic fluid collected inside the capsule under the weight-bearing forces. Finally, our suggestion in the treatment of Perthes disease is to relieve weight bearing up to the fragmentation stage, whether the diagnosis has been made by the use of MRI or without it. During the fragmentation stage, MRI is extremely useful in performing prognosis; at this time our classification can be applied, and the corresponding treatment can be followed.