Use of Neural Networks in Children’s Cerebral Palsy Recognition by Gait Analysis Data

The classification systems for cerebral palsy (CP) need to be continuouslyupdated, according to specific objectives and to significant changes observed over theyears in the panorama of CP. Ferrari et al. [1], recently proposed a classification systemthat aimed at subdividing the diplegic children into four main clinical sub-forms, on thebase of their walking pattern. This preliminary study deals with the classification of thediplegic children affected by CP and it is based on the walking pattern classification systemproposed by Ferrari and utilized in LAMBDA motion analysis laboratory at S. MariaNuova Hospital of Reggio Emilia. Using kinematics data recorded by means of anoptoelectronic system on children affected by CP, an Artificial Neural Network (ANN) wasimplemented to allow an automatic recognition of the form of the palsy. The ANN proposedcorrelates a set of suitable statistical parameters of the kinematics of walking with the typeof diplegic clinical form. The effectiveness of the resulting neural network has been provedon a control set of data