Diagnosis of Early Mild Cognitive Impairment in Type 2 Diabetes Mellitus by Deep Learning of Multimodal Images and Metadata

Type 2 diabetes mellitus (T2DM) is a chronic metabolic disorder, which can result in abnormal brain alternations and mild cognitive impairment in elder adult lives. Therefore, identification of early cognitive impairment in patients with T2DM is of paramount importance for mitigating cognitive decline of patients and enhancing their quality of life. Moreover, clinical metadata is informative in T2DM diagnosis, which can provide prior knowledge in demonstrating different severities of T2DM. To this end, we develop a robust deep learning model for diagnosing early cognitive impairment in T2DM using multi-modal neuroimages, which incorporates the informative clinical metadata (i.e., MoCA, BMI and HbA1c) to design metadata-induced contrastive Laplacian regularization. This can effectively alleviate the problem of small medical dataset in deep learning. Extensive experiments have shown significant improvement in accuracy and AUC in the identification of T2DM with/without mild cognitive impairment in a dataset with 311 subjects, indicating the ability of the proposed method in understanding of associated brain alterations in T2DM and its potential applications on other brain disorders.