Diagnosis of osteoporosis by quantification of trabecular microarchitectures from hip radiographs using artificial neural networks

Ju Hwan Lee, Yoo Na Hwang, Sung Yun Park, Jae Hoon Jeong, Sung Min Kim

Research output: Contribution to journalArticlepeer-review

7 Scopus citations

Abstract

The purpose of this study was to assess the diagnostic efficacy of an artificial neural network (ANN) in identifying postmenopausal women with low bone mineral density (BMD) by quantifying trabecular bone microarchitectures. The study included 53 post-menopausal women, who were classified as normal (n = 17) and osteoporotic (n = 36) according to T -scores. BMD was measured on the femoral neck by dual-energy X-ray absorptiometry. Morphological features were extracted to find optimum input variables by quantifying microarchitectures of trabecular bone. Principal component analysis was used to reduce the dimensionalities and improve classification accuracy. For the classification, a two-layered feed forward ANNs was designed using the Levenberg-Marquardt training algorithm, and was used to evaluate classification performance in terms of sensitivity, specificity and accuracy. The experimental results indicated the superior performance of the proposed approach for discriminating osteoporotic cases from normal. Moreover, our method considerably reduced the level of misclassification rates, and revealed the best classification results. Based on these results, we found the clinical usefulness of our method for diagnosing osteoporosis.

Original languageEnglish
Pages (from-to)1115-1120
Number of pages6
JournalJournal of Computational and Theoretical Nanoscience
Volume12
Issue number7
DOIs
StatePublished - 1 Jul 2015

Keywords

  • Artificial neural network
  • Bone mineral density
  • Dual-energy x-ray absorptiometry
  • Microarchitecture
  • Osteoporosis
  • Trabecular bone

Fingerprint

Dive into the research topics of 'Diagnosis of osteoporosis by quantification of trabecular microarchitectures from hip radiographs using artificial neural networks'. Together they form a unique fingerprint.

Cite this