AUTHOR=Duan Junwei , Liu Yang , Wu Huanhua , Wang Jing , Chen Long , Chen C. L. Philip 

TITLE=Broad learning for early diagnosis of Alzheimer's disease using FDG-PET of the brain

JOURNAL=Frontiers in Neuroscience

VOLUME=Volume 17 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2023.1137567

DOI=10.3389/fnins.2023.1137567

ISSN=1662-453X

ABSTRACT=Alzheimer’s Disease (AD) is a progressive neurodegenerative disease, the development of AD is irreversible. However, prevention measures in the presymptomatic stage of AD can effectively slow down the deterioration. Fluorodeoxyglucose Positron Emission Tomography (FDG-PET) can detect the metabolism of glucose in patients’ brain, which can help identify changes related to AD prior to brain damage. Machine Learning is useful for early diagnosis of AD patients with FDG-PET, while it needs a sufficiently large dataset, and it’s easy to overfit in small data set. Previous studies using machine learning for early diagnosis with FDG-PET either need extract tricky handcrafted feature or validated on a small dataset, and few studies have explored the refined classification of Early MCI (EMCI) and Late MCI (LMCI). This paper presents a broad network based model for early diagnosis of AD (BLADNet) by PET imaging of the brain, a novel broad-neural-network to enhance feature of FDG-PET extracted from 2D-CNN. BLADNet can search information in broad space by adding new BLS blocks without retraining the whole network, thus improve the accuracy of AD classification. The experiments conducted on a dataset that contains 2298 FDG-PET images of 1045 subjects from ADNI demonstrate that our methods are superior to previous studies of early diagnosis of AD with FDG-PET. Especially, our methods achieved state-of-the-art results in EMCI and LMCI classification with FDG-PET.