The radiomics features of the temporal lobe region related to menopause based on MR-T2WI can be used as potential biomarkers for AD.

Abstract

Menopause may be an important pathogenic factor for Alzheimer's disease (AD). The M1 polarization of microglia and neuroinflammatory responses occur in the early pathogenetic stages of AD. Currently, no effective monitoring markers are available for AD's early pathological manifestations. Radiomics is an automated feature generation method for the extraction of hundreds of quantitative phenotypes (radiomics features) from radiology images. In this study, we retrospectively analyzed the magnetic resonance T2-weighted imaging (MR-T2WI) on the temporal lobe region and clinical data of both premenopausal and postmenopausal women. There were three significant differences were identified for select radiomic features in the temporal lobe between premenopausal and postmenopausal women, i.e. the texture feature Original-glcm-Idn (OI) based on the Original image, the filter-based first-order feature Log-firstorder-Mean (LM), and the texture feature Wavelet-LHH-glrlm-Run Length Nonuniformity (WLR). In humans, these three features were significantly correlated with the timing of menopause. In mice, these features were also different between the sham and ovariectomy (OVX) groups and were significantly associated with neuronal damage, microglial M1 polarization, neuroinflammation, and cognitive decline in the OVX groups. In AD patients, OI was significantly associated with cognitive decline, while LM was associated with anxiety and depression. OI and WLR could distinguish AD from healthy controls. In conclusion, radiomics features based on brain MR-T2WI scans have the potential to serve as biomarkers for AD and noninvasive monitoring of pathological progression in the temporal lobe of the brain in women undergoing menopause.