Method for Automatic Selection of Parameters in Normal Tissue Complication Probability Modeling

Purpose: To present a fully automatic method to generate multiparameter normal tissue complication probability (NTCP) models and compare its results with a published model of the same patient cohort. Methods and Materials: Data were analysed from 345 rectal cancer patients treated with external radiotherapy to predict the risk of patients developing grade 1 or ≥2 cystitis. In total 23 clinical factors were included in the analysis as candidate predictors of cystitis. Principal component analysis (PCA) was used to decompose the bladder dose volume histogram (DVHs) into 8 principal components (PCs), explaining more than 95% of the variance. The dataset of clinical factors and PCs was divided into training (70%) and test (30%) datasets, with the training dataset used by the algorithm to compute an NTCP model. The first step of the algorithm was to obtain a bootstrap sample, followed by multicollinearity reduction using the variance inflation factor (VIF) and genetic algorithm optimisation to determine an ordinal logistic regression model that minimises the Bayesian information criterion (BIC). The process was repeated 100 times and the model with the minimum BIC was recorded on each iteration. The most frequent model was selected as the final ‘automatically generated model’ (AGM). The published model and AGM were fitted on the training datasets and the risk of cystitis was calculated. Results: The two models had no significant differences in predictive performance both for the training and test datasets (p-value>0.05), and found similar clinical and dosimetric factors as predictors. Both models exhibited good explanatory performance on the training dataset (p-values>0.44) which was reduced on the test datasets (p-values