Automatic Landmark Detection on Epicondyles of Distal Femur in X-Ray Images

Orthopaedic condition such as degenerative arthritis (e.g., osteoarthritis) can deteriorate the osteoarticular systems, causing pain and affecting activities of daily living. Treatments such as Total Knee Arthroplasty (TKA) or Spinal Arthroplasty are commonly used to relieve the pain/disability and provide mobility. This requires an accurate definition of key anatomical landmarks both for surgical operation and implant design. Utilization of the X-ray images is essential for diagnostic purposes (e.g., to identify the extent of pathological conditions), and demand for their use is ever-increasing. Moreover, radiological images are used for shape extraction/bone geometry for orthopaedic application (such as implant design or surgical/non-surgical planning) [1,2,3]. However, visual inspection/manual extraction is a subjective task and error prone, and geometrical analysis (through shape extraction), in particular, is a time consuming process [4]. In this paper we present a method to find anatomical landmarks from X-ray images, using MATLAB software (The MathWorks Inc., Natick, MA, USA) [11]. Segmentation and shortest path algorithm together with implementation of proper constraints were employed to find medial and lateral edges of the femur, enabling automatically extract parameters such as the epicondylar width of the femur. Some of the constraints applied to the program include adjusting image intensity values [10] particularly on the medial side of the knee, and separating image in to two parts, due to differences in the intensity of the image caused by patella projection. Initially, twenty series of digital X-ray images of the distal femur were used as input data to extract their anatomical landmarks, providing a means to automatically measure morphometric dimensions in different populations (by exporting the detected edges in 2-D coordinate system). The results show that the developed algorithm can detect the landmark in a robust way. Further integration of the result with Principal Component Analysis (PCA) technique is envisaged to identify the statistically important parameters required for orthopaedic planning or optimization of implants in the knee & spine area.