FEATURE SELECTION AND CLASSIFICATION TO AUTOMATICALLY DETECT KNEE OSTEOARTHRITIS USING MRI

Open Access
Author:
Keffalas, Matthew George
Area of Honors:
Electrical Engineering
Degree:
Bachelor of Science
Document Type:
Thesis
Thesis Supervisors:
  • David Miller, Thesis Supervisor
  • Kenneth Urish, Thesis Supervisor
  • Jeffrey Louis Schiano, Honors Advisor
Keywords:
  • support vector machine
  • feature selection
  • osteoarthritis
  • machine learning
  • MRI
Abstract:
Osteoarthritis (OA) is a common joint disease, affecting roughly half of people age 55 or older. Currently, there is no reliable, noninvasive method for OA diagnosis in advance of the onset of symptoms. Recently, MRI, as an alternative to radiography, has shown promise for identifying pre-radiographic disease signatures. In this work, we predict changes due to OA prior to both their symptomatic presentation and radiographic detection. Analysis of data shows an acceptable accuracy. This presents a viable method for detecting OA in the early stages of the disease, and with further development the analysis could become a significant tool for early clinical OA diagnosis and for identifying study populations for both epidemiological and drug studies.