LightMoDAD: A Lightweight Diagnosis Network for Alzheimer's Disease with Small-Scale Multi-Modal Data

Alzheimer's Disease (AD) is a chronic neurodegenerative disease without effective medications or supplemental treatment. Early and accurate diagnosis of AD is crucial for effective treatment and patient management. However, AD diagnosis model training is often suffered from small datasets. This paper proposes a lightweight AD diagnosis network trained by using small multi-modal datasets. First, multimodal medical images are produced by fusing structural information from magnetic resonance imaging (MRI) of AD patients with brain activity information from positron emission tomography (PET) images. Then, a lightweight neural network is constructed by integrating convolutional neural networks (CNNs) with transformer networks to extract essential features and classify Alzheimer's disease images. Meanwhile, transfer learning makes the AD diagnosis model less dependent on data. Our model achieves promising results in terms of accuracy, sensitivity, and specificity using a small subset from the Alzheimer's Disease Neuroimaging Initiative (ADNI) public dataset.