Multimodal Deep Learning for Early Alzheimer’s Disease Prediction Using MRI Texture Features and Gut Microbiome Data

Abstract

Early diagnosis of Alzheimer Disease (AD) has become one of the most significant problems in clinical neuroscience since structural and biological changes may occur decades before a symptom is visible to the naked eye. In this paper, a new multimodal Deep Learning (DL) model is suggested, which incorporates MRI texture features based on Gabor filters and gut microbiome data to better predict of Alzheimer disease in early. The structural brain changes are recorded by the use of Magnetic Resonance Imaging (MRI) and the Gabor filters may be applied to extract the discriminative texture features that have the capability of enhancing the faint profiles which are associated with the initial neurodegeneration. Simultaneously, microbial composition and diversity data, which are the gut microbiome, is added as a new biomarker associated with the gut-brain axis and cognitive impairment. The proposed model combines a hybrid DL architecture that involves the use of convolutional neural networks (CNNs) to handle Gabor-enhanced MRI data and fully connected layers to handle microbiome features data. Multimodal fusion approach is applied so that the heterogeneous sources of data can be properly combined so that the model could acquire complementary representations. The framework is experimented on the categorization at the initial level, between cognitively normal, mild cognitive impairment (MCI) and the Alzheimer disease. The evidence of the experiment proves that imaging and microbiome data integration are much more effective than unimodal in predicting the results. In addition, the sensitivity of the features to the primary structural changes of brain tissues is enhanced by Gabor filters. The mentioned study identifies the opportunities of the combination of neuroimaging and biological data with a DL to assist in the early diagnosis of the Alzheimer disease as a potential course in the precises medicine and preventive care.