Classification of Type 2 Diabetes Mellitus Subtypes Defined by Traditional Chinese Medicine Based on High-order Brain Functional Network

Abstract Type 2 diabetes mellitus (T2DM) cases have various complications in multiple organs, including the brain. According to traditional Chinese medicine (TCM), the clinical appearance of T2DM can be divided into two syndromes, namely a Deficiency Syndrome and an Excess Syndrome, depending on the patient’s clinical symptoms, pulse and tongue appearance. However, the differences between TCM of T2DM-related subtypes remain unclear. Compared to the conventional static functional connectivity (FC), dynamic FC constitutes a promising strategy for objective and non-invasive identification of possible imaging biomarkers of T2DM Deficiency and Excess Syndromes and may serve as effective diagnostic features for differentiating these two subtypes. This work aimed to utilize high-order FC to identify abnormal connectomics patterns in the Deficiency and Excess Syndromes of Type 2 Diabetes Mellitus using a multivariable, machine learning-based method. We first constructed large-scale high-order brain networks by analyzing the temporal synchronization of the dynamic FC time series among various pairs of cerebral regions. Subsequently, this information was employed to classify the T2DM Deficiency and Excess Syndromes using a support vector machine model. The built model had an 80% accuracy in distinguishing T2DM Deficiency Syndrome from T2DM Excess Syndrome, and detected aberrant high-order FC patterns between them. Furthermore, widespread connectivity alterations were identified, which reflected T2DM subtypes. Therefore, these findings provide valuable insights into the TCM understanding of T2DM.