Artificial Neural Networks-An Alternative Method in the Selection of Human Origin Meant for the Varum Correcting Tibial Osteotomy

A Sfarghiu, Monica Serban, H Follet, R Bardonnet
2003 unpublished
One of the most frequent diseases is the femuro-tibial arthrosis manifested through the extremely painful wearing away of the cartilage of the knee joint. A widely used method of treatment is the open osteotomy whose technique consists in the bone's sectioning followed by the straightening of the axle through the insertion of a bony graft. Starting with 1993 the TBF Laboratory, specialised in tissue engineering, produces Phoenix grafts for osteotomies. The Phoenix grafts are selected on the
more » ... s of clinical and biological tests carried out on the raw material-the femoral apophyses that are the result of the inserting of the coxo-femoral joint's prosthesis. One of the major problems of the open osteotomy is obtaining an in time-resistant consolidation, and this implies a resistance to compression of the bony graft superior to 60 MPa. In this respect, the present paper describes a method based on neural networks as a tool in classifying the mechanical resistance of the grafts and, also, a comparison with the classical multivariable statistical method. The goal is to select only those samples with a mechanical resistance superior to the value of 60 MPa and, for that, we are looking after a simple criterion that can be directly applied to them without destroying the grafts by testing. This is governed both: by financial reasons and limited resources.