A procedure is presented which allows the efficient production of a patient specific computer model of the femur, for surgical planning. Similar models require long processing times and/or high performance computing.

The method uses 24 key landmark points to customise a generic femur to patient data, using a desktop computer. By using non-linear elements a smooth, curved surface is obtained. A finite element mesh of a generic femur consisting of 384 elements was created using the analysis software CMISS (Bioengineering Institute, University of Auckland). A rectangular shaped host mesh was defined to enclose the generic femur. Datasets of 5 human femurs were obtained using a hand-held laser scanner on dry bones and the visible human dataset. Key landmark data points were selected on the generic femur along with corresponding target points on each data set. The host mesh was then deformed using a least squares algorithm, causing customisation of the generic femur to the patient specific model. Each customised model was compared with its entire dataset. The fitting process took less than 100 seconds on a 180 MHz 02 computer (SGI, CA, USA). The algorithm yielded an average root mean square (RMS) of 3.09mm with a standard deviation of 0.15mm. Operator time for positioning the projection points was less than 5 minutes.

This paper presents a novel means for customisation of human femoral geometry with generation of patient specific models on a PC from scan data in under 10 minutes. Current work is focusing on stress analysis, surgical simulation and planning.