Deep Learning-based Feature Extraction with MRI Data in Neuroimaging Genetics for Alzheimer’s Disease

Abstract

The prognosis and treatment of the patients suffering from Alzheimer's disease (AD) have been one of the most important and challenging problems over the last few decades. To better understand the mechanism of AD, it is of great interest to identify genetic variants associated with brain atrophy. Commonly in these analyses, neuroimaging features are extracted based on one of many possible brain atlases with FreeSurf and other popular softwares, which however may lose important information due to our incomplete knowledge about brain function embedded in these suboptimal atlases. To address the issue, we propose convolutional neural network (CNN) models applied to three-dimensional MRI data for the whole brain or multiple divided brain regions to perform completely data-driven and automatic feature extraction. These image-derived features are then used as endophenotypes in Genome-Wide Association Studies (GWAS) to identify associated genetic variants. When applied to the ADNI data, we identified several associated SNPs which have been previously shown to be related to several neurodegenerative/mental disorders such as AD, depression and schizophrenia. Code and supplementary materials are available at https://github.com/Dipnil07. The codes have been implemented using Python, R and Plink softwares.

Deep causal feature extraction and inference with neuroimaging genetic data

“…For example, there are at least 66 existing atlases for the whole brain structural MRI (sMRI) data; 12 which one is best to use? Given the tremendous successes of CNNs, especially for automatic feature extraction in image analysis, it is natural to apply CNNs to extract features from neuroimaging data, and use these features as the traits (ie, endophenotypes) to be associated with SNPs 17 . These extracted features can go beyond any given ROIs.…”

“…The sample size of the ADNI, around 800, is still small. Other techniques, such as transfer learning, 44 data augmentation 17 and self‐supervised learning can be explored and incorporated in the future for better performance. Third, perhaps most importantly, it is still quite challenging to interpret CNN‐extracted features.…”

“…A convolutional auto‐encoder network is proposed to reduce dimension from MRI scans 15,16 . A recent study proposes to extract features by training 3D convolutional neural networks in an AD classification task, reporting an AUC score of 0.75 on test data 17 . They also perform GWAS scans to identify genetic variants associated with extracted features.…”

A Novel Artificial Intelligence (AI) Method to Classify and Predict the Progression of Alzheimer’s Disease

“…MRI is the most commonly employed imaging technique for predicting stages of AD [13]. In order to reveal hidden information and eliminate noise inherent in complex MRI sequences, the extraction of important features becomes imperative [14]. Numerous studies have leveraged these critical input features extracted from MRI data to optimize performance [15][16][17][18].…”