Early diagnosis of Alzheimer’s disease by ensemble deep learning using FDG-PET

Early diagnosis of Alzheimer’s disease (AD) is critical in preventing from irreversible damages to brain cognitive functions. Most computer-aided approaches consist of extraction of image features to describe the pathological changes and construction of a classifier for dementia identification. Deep learning technique provides a unified framework for simultaneous representation learning and feature classification, and thus avoids the troublesome hand-crafted feature extraction and feature engineering. In this paper, we propose an ensemble of AlexNets (EnAlexNets) algorithm for early diagnosis of AD using positron emission tomography (PET). We first use the automated anatomical labeling (AAL) cortical parcellation map to detect 62 brain anatomical volumes, then extract image patches in each kind of volumes to fine-tune a pre-trained AlexNet, and finally use the ensemble of those well-performed AlexNets as the classifier. We have evaluated this algorithm against seven existing algorithms on an ADNI dataset. Our results indicate that the proposed EnAlexNets algorithm outperforms those seven algorithms in differentiating AD cases from normal controls.