A comparison of various MRI feature types for characterizing whole brain anatomical differences using linear pattern recognition methods

• Published in: NeuroImage (Orlando, Fla.) Fla.), 2018-09, Vol.178, p.753-768
• Main Authors: Monté-Rubio, Gemma C., Falcón, Carles, Pomarol-Clotet, Edith, Ashburner, John
• Format: Article
• Language: English
• Subjects: Accuracy; Algorithms; Alzheimer's disease; Body mass index; Brain - anatomy & histology; Brain research; Datasets as Topic; Diffeomorphism; Gaussian process; Humans; Image Processing, Computer-Assisted - methods; Learning algorithms; Machine Learning; Magnetic resonance imaging; Magnetic Resonance Imaging - methods; Medical imaging; Model selection; Neuroimaging; Neuroimaging - methods; Noise; Pattern recognition; Pattern Recognition, Automated - methods; Registration; Regression analysis; Scalar momentum; Structural MRI; Variables; VBM
• ISSN: 1053-8119; 1095-9572
• DOI: 10.1016/j.neuroimage.2018.05.065

There is a widespread interest in applying pattern recognition methods to anatomical neuroimaging data, but so far, there has been relatively little investigation into how best to derive image features in order to make the most accurate predictions. In this work, a Gaussian Process machine learning approach was used for predicting age, gender and body mass index (BMI) of subjects in the IXI dataset, as well as age, gender and diagnostic status using the ABIDE and COBRE datasets. MRI data were segmented and aligned using SPM12, and a variety of feature representations were derived from this preprocessing. We compared classification and regression accuracy using the different sorts of features, and with various degrees of spatial smoothing. Results suggested that feature sets that did not ignore the implicit background tissue class, tended to result in better overall performance, whereas some of the most commonly used feature sets performed relatively poorly. •We tested various types of features, derived from segmented T1-weighted MRI, according to their ability to be used by a linear pattern recognition method to predict age, sex, BMI, and diagnostic status.•Scalar momenta, and other methods that do not ignore the implicit background class provided more effective strategies.•The most widely used feature type, namely “modulated” GM, was found to be relatively ineffective.