Symbolic regression metamodel-based optimal design of patient-specific spinal implant (pedicle screw fixation)

Abstract

Pedicle screw-rod insertion is a common surgical procedure used for treating degenerative spinal diseases. Optimized design of such implants is necessary to avoid undue strains at the bone–implant interface. In this work, ideal optimized implant design is defined as one for which the strain difference between intact bone and bone after implantation at six interfacial positions is zero. To achieve this, genetic programming (GP) based symbolic regression (SR) metamodels are built from limited data obtained from expensive but highly accurate finite element (FE) models. The FE models are generated from CT scan data. A cumulative objective function is expressed in terms of GP-based SR metamodels which is then combined with a genetic algorithm (GA) to predict patient-specific optimum implant designs.