Analysis of spinal lumbar interbody fusion cage subsidence using Taguchi method, finite element analysis, and artificial neural network

Abstract

Subsidence, when implant penetration induces failure of the vertebral body, occurs commonly after spinal reconstruction. Anterior lumbar interbody fusion (ALIF) cages may subside into the vertebral body and lead to kyphotic deformity. No previous studies have utilized an artificial neural network (ANN) for the design of a spinal interbody fusion cage. In this study, the neural network was applied after initiation from a Taguchi L18 orthogonal design array. Three-dimensional finite element analysis (FEA) was performed to address the resistance to subsidence based on the design changes of the material and cage contact region, including design of the ridges and size of the graft area. The calculated subsidence is derived from the ANN objective function which is defined as the resulting maximum von Mises stress (VMS) on the surface of a simulated bone body after axial compressive loading. The ANN was found to have minimized the bone surface VMS, thereby optimizing the ALIF cage given the design space. Therefore, the Taguchi-FEA-ANN approach can serve as an effective procedure for designing a spinal fusion cage and improving the biomechanical properties.