R. de Luis-García, J. Ignacio Arribas, S. Aja-Fernández, and C. Alberola-López (Spain)
Medical Image Analysis, Neural Network, Bone Age Assessment, Model Selection
In this paper, we present a particular Neural Architec ture named Generalized Softmax Perceptron (GSP) suit able to estimate probabilities at the output after having been trained in supervised mode, together with a Posterior Prob ability Model Selection (PPMS) algorithm which automat ically determines the optimal architecture size based on statistical considerations to estimate bone age from a tiny sized hand-wrist radiograph data set. Traditional feature definition, selection and extraction strategies are carried out in order to improve final neural classification step, yielding promising results shown.
Important Links:
Go Back
