Arthritis & Rheumatism, Volume 60,
October 2009 Abstract Supplement

The 2009 ACR/ARHP Annual Scientific Meeting
Philadelphia October 16-21, 2009.

Active Appearance Modelling of DXA Images to Assess Severity of Osteoarthritis of the Hip

Barr1,  Rebecca J., Gregory1,  Jennifer S., Yoshida1,  Kanako, Alesci2,  Salvatore, Aspden1,  Richard M., Reid1,  David M.

University of Aberdeen, Aberdeen, United Kingdom
Wyeth Research, Collegeville, PA


The use of plain radiographs for the assessment of Osteoarthritis (OA) is well established. Our recent studies have found that new generation Dual Energy X-ray Absorptiometry (DXA) scanners, normally used for measuring Bone Mineral Density (BMD), produce images of sufficient quality to allow grading of OA severity using the Kellgren Lawrence (KLG) scoring system. We have also shown that Active Shape Modelling (ASM), a method for quantifying variations in the shape of the femur, can identify early onset and progression of hip OA from radiographs. The aim of this study was to develop this further by using Active Appearance Modelling (AAM) and DXA images to model not only the shape, but also the variation in image intensity (BMD distribution) within the femur.


Sixty two subjects with hip OA were identified as having had a pelvic radiograph in the last 12 months using the local Radiology Information System, and invited to undergo a dual femur DXA scan using an iDXA scanner (GE Medical Systems). DXA images of each hip were graded using the KL system. An 85 point AAM template, that included the proximal femur, osteophytes, and parts of the pelvis, was applied to each image to provide a comprehensive model of the hip joint. Outputs from the AAM are independent "modes of variation" that describe changes in the shape of the hip and BMD distribution occurring together in the dataset. Correlations between the first 10 modes and KLG were assessed to identify modes likely to be of interest for assessing OA progression. These were then analysed using one-way ANOVA.


Significant correlations were found between KLG and modes 2, 4 and 7 (P < 0.001, r = 0.22 – 0.49, Fig 1A). These modes reflected changes in BMD distribution, identifying common areas of sclerosis as well as changes in the shape of the femur, the pelvis and joint space narrowing (Fig 1B). Results were confirmed by one-way ANOVAs (P < 0.001).

Fig. 1.


Active Appearance Modelling can be used with DXA images to extend the ASM of the hip to collect information on BMD distribution. This additional information may further improve our ability to assess the early onset, severity and progression of hip OA.

To cite this abstract, please use the following information:
Barr, Rebecca J., Gregory, Jennifer S., Yoshida, Kanako, Alesci, Salvatore, Aspden, Richard M., Reid, David M.; Active Appearance Modelling of DXA Images to Assess Severity of Osteoarthritis of the Hip [abstract]. Arthritis Rheum 2009;60 Suppl 10 :215
DOI: 10.1002/art.25298

Abstract Supplement

Meeting Menu