Abstract

Geometric information about a patient’s anatomy is vital in pre-operative planning for orthopedic surgeries. At present, many physicians rely on the analysis of 2D scans, such as radiographs, magnetic resonance images (MRIs), or computed tomography (CT) scans, in order to visualize a patient’s geometry. Research is being done to develop patient-specific 3D bone geometry to aid in pre-operative planning and assist in creating personalized finite element models that can help predict complications or failure after surgery. While 3D models are incredibly useful and have been developed for several parts of the human anatomy, there is still a great need for cost-effective ways to create personalized total knee arthroplasty (TKA) models. Currently, methods for creating patient-specific knee geometries typically only consider patients without TKA implants, rely on expensive medical imaging, such as layered CT scans, or require a large database of 3D models. This work presents a novel semi-automated process to create patient-specific 3D tibia geometry for TKA patients using pairs of standard, low-cost bi-planar radiographs. The method presented involves two main stages. In the first stage, the geometric bone contours from each view of the tibia are extracted from two bi-planar radiographs using gradient thresholding and Canny edge detection on user-defined regions of interest. The second stage aligns the two extracted tibia contours with a generic 3D tibia model from the 6th SimTK Grand Challenge Dataset and then modifies the generic 3D model to match the radiographic contours using a full-ellipse scaling and shifting (FESS) method. The effectiveness of the FESS method is evaluated by comparing the contours of the generated patient-specific model in the anteroposterior (AP) and lateral views to the radiograph contours in the AP and lateral views using difference calculations for three TKA patients. The maximum difference between the patient-specific models and the radiographs in both AP and lateral views is found to be on the order of 1 mm.

This content is only available via PDF.
You do not currently have access to this content.