Article Text

SAT0327 Predicting hip osteoarthritis progression using active appearance modelling (AAM) applied to hip DXA images from the auckland calcium study
  1. K. Yoshida1,
  2. J.S. Gregory1,
  3. B. Mason2,
  4. I.R. Reid2,
  5. D.M. Reid1
  1. 1University of Aberdeen, Aberdeen, United Kingdom
  2. 2University of Auckland, Auckland, New Zealand


Background Dual energy Xray absorptiometry (DXA) scanners are used to assess osteoporosis but recently, DXA images have also been shown to adequately assess osteoarthritis (OA) and Active Shape modelling (ASM) to predict hip OA progression1. AAM is an extension of ASM incorporating information on both bone shape and density.

Objectives To identify hip OA cases and determine if AAM applied to hip DXA images can predict the progression of hip OA.

Methods Hip DXA scans taken at 30 month intervals over 5 years from postmenopausal women participating in the Auckland Calcium Study2were scored for the presence of hip OA using Kellgren Lawrence grades (KL). Hip OA progression was defined as a) Hip arthroplasty (excluding hip fracture and avascular necrosis) or b) KL change of ≥1/2 grade based on progressive OA features and KL≥2 on follow up scan. Frequency matching for age was used to obtain 1:3 ratio of OA progression cases (a and b) to never OA controls (n=378). 48-point proximal femur AAM was applied to the DXA images.

Results Significant correlations (P<0.001) were found between baseline KL grade and Appearance modes (App)1,6,8,13,15 and 16 (rho=0.23, -0.17, 0.20, -0.24, -0.21, 0.16). When OA progression was defined as hip arthroplasty of the scanned hip (n=21; controls including 41 non-progressive hip OA cases, n=419) App7,8,13,16,18 and 19 were respectively found to be significant predictors, but only App16,18 and 19 remained statistically significant after adjustment for KL. When controls were widened to include all nonarthroplasty cases (n=495), App7,8,13,16,18 and 19 were found to be significant predictors of hip arthroplasty but only App13 and 16 remained significant after KL adjustment. When new incident hip OA was defined as KL≥2 at follow up in absence of baseline OA (n=47), both App7 and 13 remained significant after adjustment. When OA cases were most broadly defined as KL change≥1/2 grade or arthroplasty (n=126) and controls included non-progressive hip OA cases (n=419), 6 modes (App7,13,16 inclusive) remained significant after KL adjustment (Table 1).

Table 1. Selected appearance modes. Odds ratio (95% confidence interval), P values

Conclusions AAM applied to hip DXA images can predict the incidence and progression of hip OA.

  1. Reid IR et al. American Journal of Medicine. 2006;119(9):777-85

  2. Barr RJ et al. doi:10.1093/rheumatology/ker382

Disclosure of Interest None Declared

Statistics from

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.