Medical Engineering & Physics 32 (2010) 1180–1188 Contents lists available at ScienceDirect Medical Engineering & Physics journal homepage: www.elsevier.com/locate/medengphy A method to reconstruct patient-specific proximal femur surface models from planar pre-operative radiographs P.E. Galibarov, P.J. Prendergast, A.B. Lennon ∗ Trinity Centre for Bioengineering, School of Engineering, Trinity College Dublin, Dublin 2, Ireland article info Article history: Received 28 May 2010 Received in revised form 20 July 2010 Accepted 17 August 2010 Keywords: Patient-specific 3D reconstruction Image processing Finite element analysis abstract Three-dimensional reconstruction from volumetric medical images (e.g. CT, MRI) is a well-established technology used in patient-specific modelling. However, there are many cases where only 2D (planar) images may be available, e.g. if radiation dose must be limited or if retrospective data is being used from periods when 3D data was not available. This study aims to address such cases by proposing an automated method to create 3D surface models from planar radiographs. The method consists of (i) contour extraction from the radiograph using an Active Contour (Snake) algorithm, (ii) selection of a closest matching 3D model from a library of generic models, and (iii) warping the selected generic model to improve correlation with the extracted contour. This method proved to be fully automated, rapid and robust on a given set of radiographs. Measured mean surface distance error values were low when comparing models reconstructed from matching pairs of CT scans and planar X-rays (2.57–3.74 mm) and within ranges of similar studies. Benefits of the method are that it requires a single radiographic image to perform the surface reconstruction task and it is fully automated. Mechanical simulations of loaded bone with different levels of reconstruction accu- racy showed that an error in predicted strain fields grows proportionally to the error level in geometric precision. In conclusion, models generated by the proposed technique are deemed acceptable to perform realistic patient-specific simulations when 3D data sources are unavailable. © 2010 IPEM. Published by Elsevier Ltd. All rights reserved. 1. Introduction In clinical practice medical images are widely used for diagnos- tic, pre-operative planning and other purposes. Common imaging techniques such as CT, MRI, ultrasound, and conventional radio- graphy are capable of capturing important data for applications, where patient-specific morphology is required. Volumetric images (CT, CT, MRI, 3D ultrasound) are highly desired by researchers who are interested in precise anatomical morphology. However, these images are not always available for research and clinical use. Often just planar radiographs are used for pre-operative planning due to their cost efficiency and lower radiation dose [1]. Due to the lack of 3D data it is more difficult to obtain a precise geometry from such a planar image. Conventional 3D reconstruction from a volumetric image begins with a stack of 2D images (slices) with known spa- tial dimensions. This process consists of several steps: (i) image ∗ Corresponding author at: Trinity Centre for Bioengineering, School of Engineer- ing, Parsons Building, Trinity College Dublin, Dublin 2, Ireland. Tel.: +353 01 896 2396; fax: +353 01 679 5554. E-mail addresses: alex.lennon@tcd.ie, lennonab@tcd.ie (A.B. Lennon). processing—enhancing initial images to remove noise and scanning artefacts, (ii) image segmentation—segmenting 2D/3D images into several regions representing objects or, alternatively, a 3D image can be converted directly into a 3D model using voxels (e.g. vol- ume ray casting [2] or finite element models for stress analysis [3]); and, finally, (iii) surface extraction—defining the objects surface representation. An example of surface extraction is the generation of isosurfaces based on voxel intensities, e.g. the Marching Cubes algorithm [4] (closest voxels with a specified Hounsfield value [5] are connected into volumetric polygons to organize a 3D surface.) Alternatively, 2D contours extracted from image slices can be swept into a single surface as has been done by many researchers [6–9]. Many open-source and commercial software packages are avail- able for solving these tasks in an automated and semi-automated manner, e.g. Slicer (Harvard Medical School, Boston, MA, USA), 3D Doctor (Able Software, Lexington, MA, USA), Amira (Visage Imaging Inc., Richmond, Australia), etc. In cases when high quality volumetric data is not available sta- tistical shape modelling (SSM) can be employed to reconstruct a body part. Barratt et al. [10] proposed a technique to register a mean statistical femur model to 3D ultrasound image data. Comparison of the results to corresponding CT-based models showed that they were favourable compared to other existing techniques (average 1350-4533/$ – see front matter © 2010 IPEM. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.medengphy.2010.08.009