relative importance_based sensitivity analysis of inputs of neural networks

I have a vibration data trained with patternnet (neural network). I have extracted 9 features from the data as the inputs of neural network. now, I want to determine the relative importance sensitivity analysis of these inputs that has introduce by Garson (G.D. Garson, Interpreting neural-network connection weigths, AI Expert 6

(1991) 46–51.) and find the most effective input on the outputs of the neural network. how can I write the code for it?