Multivariate image analysis (MIA) descriptors have been applied to the quantitative structure–property relationship (QSPR) study of 13C-NMR chemical shifts of 2-mono substituted pyridines. In this method, descriptors are generated from pixels of images and are analyzed with different multivariate methods. Correlation ranking–principal component regression and correlation ranking–principal component–artificial neural networks were applied in constructing predictor models. In this article, the role of weight update function in artificial neural networks was investigated too. Obtained results using the correlation ranking–principal component–artificial neural network method showed high performance for predicting of 13C-NMR chemical shifts of pyridine derivatives. Also, these results indicated that MIA descriptors may be useful to predict 13C-NMR chemical shifts. Finally, The MIA-QSPR approach coupled to artificial neural networks revealed that the predictive ability of MIA descriptors is comparable or even superior for the pyridine derivatives when compared with the ChemDraw program or gauge included atomic orbital procedure for 13C chemical shifts calculations. Copyright © 2012 John Wiley & Sons, Ltd.