Predictive estimation of ovine hip joint centers: Neural networks vs. linear regression.

Journal: Journal of biomechanics
PMID:

Abstract

The purpose of this study was to investigate the utility of neural networks to estimate the hip joint center location in sheep and compare the accuracy of neural networks to previously developed linear regression models. CT scans from 16 sheep of varying ages, weight, sex, and phenotypes were acquired and the data was used to calculate the known hip joint center by sphere fitting the femoral head. A variety of neural networks were created to estimate the location of the hip joint center in the absence of CT data using different input criteria including anatomical measurements or landmark coordinates from the CT data and additional subject information. Neural networks significantly outperformed the current practice of estimating the hip joint center as the greater trochanter, however, did not significantly differ in performance to the linear regression-based approach. This study demonstrated the ability of neural networks to accurately predict the hip joint center in sheep but also highlights the limitations of developing these neural networks in comparison to traditional linear regression models.

Authors

  • Aaron Henry
    Department of Multidisciplinary Engineering, College of Engineering, Texas A&M University, United States of America. Electronic address: ahenry@tamu.edu.
  • Carson Benner
    J. Mike Walker '66 Department of Mechanical Engineering, Texas A&M University, United States of America. Electronic address: carsonmbenner@tamu.edu.
  • Anish Easwaran
    Department of Biomedical Engineering, College of Engineering, Texas A&M University, United States of America. Electronic address: anish.easwaran@tamu.edu.
  • Likhitha Veerapalli
    Department of Biomedical Engineering, College of Engineering, Texas A&M University, United States of America. Electronic address: likhithav@tamu.edu.
  • Dana Gaddy
    Department of Veterinary Integrative Biosciences, School of Veterinary Medicine & Biomedical Sciences, Texas A&M University, United States of America. Electronic address: dgaddy@tamu.edu.
  • Larry J Suva
    Department of Veterinary Physiology & Pharmacology, School of Veterinary Medicine & Biomedical Sciences, Texas A&M University, United States of America. Electronic address: lsuva@cvm.tamu.edu.
  • Andrew B Robbins
    Department of Mechanical Engineering, University of Texas at Tyler, United States of America. Electronic address: arobbins@UTTyler.edu.