An Improved Pre-processing Machine Learning Approach for Cross-Sectional MR Imaging of Demented Older Adults

Data pre-processing is the foremost step employed in building any machine learning (ML) model. It has a significant effect on the generalization performance of the model. In the present study, we have attempted to present the data pre-processing techniques for analysis of cross-sectional Magnetic Resonance Imaging (MRI) data of demented and non-demented older adults. The MRI dataset consists of 434 MR sessions of 416 subjects, aged between 18 to 96 years. Before performing classification of MRI data and its pattern analysis, characteristics of the dataset, such as sampling, imbalanced dataset, missing values, outliers, incompleteness, and the presence of irrelevant features, had been addressed. We involved ten major steps to process dataset for the ML model building. Experimental results on the cross-sectional data indicated a significant relative improvement in the pattern analysis achieved by doing PCA (Principal Component Analysis). Pattern analysis employing PCA resulted in noteworthy advancement in pattern recognition with an explained variance of 0.97377.