Classification of gender and race in the distal femur using self organising maps.

In this study gender and race differences in distal femoral morphology were investigated. Reliable anatomic knee measurements were obtained for 60 knees via MRI and direct scanning of cadaver specimens. The MRI data comprised of 20 White males and 22 White females while the cadaver specimens comprised of 18 Black males. Possible differences were investigated using a type of artificial neural network to classify the data, namely the self-organising map (SOM). The SOM suggested that clear differences are present between genders when absolute measurements are used. Male knees tended to be larger over all the measurements considered. However, when data were normalised for size, the clear differences were diminished and definite clusters were difficult to define. Black male knees tended to have larger condyle radius to anterior-posterior length ratios compared to White males. White male knees tended to be wider than White female knees. It is however suggested than when corrected for size, there exists a large variation among individual knees regardless of gender or race. It is argued that with the large variation in populations it can become advantageous not to think about gender-specific or race-specific knee replacement designs, but rather patient-specific.

Contact stresses in a patient-specific unicompartmental knee replacement.

A custom, patient-specific unicompartmental knee replacement was developed using a unsupervised neural network trained on a database of healthy knee geometries. This custom implant was then compared to a conventional implant in terms of contact stress in a Finite Element Model. The custom implant experienced lower contact stresses at the tibiofemoral joint compared to the conventional implant. The custom implant stresses were further reduced with the use of a customized mobile bearing. The custom implant also displayed more uniform stress distribution at the bone-implant interface.