Height Prediction For Growth Hormone Deficiency Treatment Planning Using Deep Learning

Prospective studies using longitudinal patient data can be used to help to predict responsiveness to Growth Hormone (GH) therapy and assess any suspected risks. In this paper, a novel Clinical Decision Support System (CDSS) is developed to predict growth (in terms of height) in children with Growth Hormone Deficiency (GHD) just before the start of GH therapy. A Deep Feed-Forward Neural Network (DFFNN) model is proposed, developed and evaluated for height prediction with seven input parameters. The essential input parameters to the DFFNN are gender, mother's height, father's height, current weight, chronological age, bone age, and GHD. The proposed model is trained using the Levenberg Marquardt (LM) learning algorithm. Experimental results are evaluated and compared for different learning rates. Measures of the quality of the fit of the model such as Root Mean Square (RMSE), Normalized Root Mean Square (N-RMSE), and Mean Absolute Percentage Error (MAPE) show that the proposed deep learning model is robust in terms of accuracy and can effectively predict growth (in terms of height) in children.