CLCA: Contrastive Learning Using Combined Additional Information for Alzheimer's Diagnosis

Alzheimer's disease (AD) is a cognitive disorder that often occurs in the elderly. In the early stage of the disease, memory loss is the primary manifestation, which is clinically called mild cognitive impairment (MCI). At present, there is no way to prevent or reverse AD. Therefore, it is essential to diagnose the disease in time. There are many studies using computers to diagnose AD automatically. Because the cause of the disease is unclear, using deep learning to diagnose AD from MRI data is a hot research topic today. In addition, for the 3-D medical imaging field, it is challenging to build a large dataset for a particular disease. Contrastive learning demonstrates strong capabilities and provides a way to learn knowledge from the images themselves. However, in this process, the additional information of participants is not used, and many studies have found that factors such as sex and age significantly influence the disease. So we improved the existing training method to take the age and sex of the participants into account and improve the encoding ability of the neural network. We validate performance on the ADNI dataset; our method outperforms training from scratch and state-of-the-art methods on AD diagnosis tasks after fine-tuning.