Texture and signal features from hippocampal T2 maps as biomarkers for MCI to AD progression

Abstract

Alzheimer's disease (AD) is the most common type of dementia and predicting who will convert from Mild Cognitive Impairment (MCI) to AD is crucial to patient benefits as well as medical research. To fulfill this purpose, in recent years it has been reported that the texture of magnetic resonance images can be an effective biomarker. In this study we used images from the Alzheimer's Disease Neuroimaging Initiative database to create T2 maps and identify features related to the texture and signal distribution for the prediction of AD. We extracted 3S features from the left and right hippocampus for 40 patients with MCI who either progressed to AD (18) or remained stable (22) and measured the mean and absolute difference of these contralateral features. We also kept the original volume of each region, yielding a total of 7S features. We used 7 machine learning methods to analyze whether by adding these imaging features to the neuropsychological studies currently used for diagnosis, we could more accurately identify who would develop the disease. We found 11 features significantly different between groups. Furthermore, all but one of the machine learning methods improved their accuracy by adding the signal- and texture-related features, and the volumetric information was non-significant. Our results suggest that these imaging features from hippocampal T2 maps should be further investigated as potential MRI biomarkers for the prediction of AD.