Analysis of Total Hip Replacements Using Active Ellipses Stuart Kerrigan 1 , Stephen McKenna 1 , Ian W. Ricketts 1 and Carlos Wigderowitz 2 1 Division of Applied Computing, University of Dundee, Dundee, Scotland, DD1 4HN. 2 Orthopaedic and Trauma Surgery, Ninewells Hospital and Medical School, Dundee, Scotland, DD1 9SY Abstract. In this paper we propose a new method for the measurement of wear of a total hip replacement. Our method exhibits a greater degree of automation and is to be both accurate and repeatable. Measurement of wear can be quantified as the displacement of the centre of the femoral head relative to the centre of the acetabular cup or acetabular rim. Our method uses active ellipses - ellipses that, with prior knowledge of the intended contour, search for and alter shape to segment the boundary of the head and rim. A set of radiographs are manually annotated and the characteristics of the boundary of the femoral head and acetabular rim are learned. Two ellipses are sequentially placed on the radiograph, the first deforming around the boundary of the femoral head, the second placed using the previously learned average shape of the acetabular rim and converges around the wire marker. Once both ellipses have converged the distance between their two centres can be calculated and converted to mm as a measure of wear. Our method is validated by comparison with manual fitting of ellipses. A discussion of the results, the clinical relevance and further investigations concludes this paper. 1 Introduction With over 40,000 total hip replacement (THR) operations performed annually within the UK and over 5,000 of these are revisions [1]. There are clear benefits for providing early detection of implant failure and evaluation of surgical techniques and implant designs. The majority of failures are due to the displacement of the centre of the femoral head relative to the centre of the acetabular cup (see Fig 1a). Current findings are that the latter occurs most frequently [2]. The aim of this research is to create a new, automated method for the analysis of total hip replacements (THRs) that is more automated, precise, accurate and repeatable than existing methods. Analysis of THRs is conducted by a variety of methods including manual methods such as overlaying concentric circular transparencies on the radiograph and annotation via pencil and rulers on the film or a digitizing tablet. Another method of analysis is roentgen stereophotogrammetric analysis (RSA) [3] which requires the insertion of markers used as reference points in follow up X-rays. As the markers are inserted during the operation it cannot be used retrospectively. Semi-automated image processing techniques have been implemented [4] [5]. These techniques use edge detection but do not use any prior knowledge of the distinct contour of the femoral head or wire rim and require considerable user interaction. Also of note is Cootes and Taylor’s Active Shape Models (ASM), deformable models that are trained from previous examples to locate structures from computer images - including hip implants [6]. It is unclear however how they are to be used for the measurement of wear. Figure 1. a) A radiograph of the Zimmer CPT prosthesis. b) Annotation of the femoral head with 9 approximately equi-spaced points, numbers correspond to the order of annotation. c) Annotation of the acetabular rim with example of order of annotation on right wing of rim, points are concentrated in areas of high curvature. 2 Method Our method uses active ellipses and the normalized first order derivative of the grey levels of the femoral head and the acetabular rim from a set of manually digitized and annotated radiographs. Models of the grey level around the boundary of the femoral head and the acetabular rim are created. With these models searches can be undertaken 1) {skerrigan, stephen, ricketts}@computing.dundee.ac.uk, 2) cawigderowitz@lineone.net