Abstract
Type 1 (T1D) and type 2 diabetes (T2D) are both associated with chronic inflammation and endothelial dysfunction, yet their discrimination based on conventional inflammatory biomarkers remains challenging. In this context, a proof-of-concept study exploring whether advanced hydrodynamic descriptors of leukocyte-endothelium interactions under flow conditions might disclose invaluable information to differentiate between diabetic phenotypes is presented. Neutrophil dynamics were investigated using a parallel-plate flow chamber and quantitative video tracking, from which an expanded set of novel hydrodynamic variables encompassing velocity-, acceleration-, rolling-, interaction-, and count-based hydrodynamic parameters was extracted. Supervised machine learning models based on linear discriminant analysis (LDA) were developed and validated using cross-validation strategies. While classical flow parameters and circulating inflammatory biomarkers showed limited discriminatory power, a parsimonious subset of novel hydrodynamic descriptors enabled robust separation between healthy, T1D, and T2D individuals, despite the moderate cohort size (n = 10 per group). Weak to moderate correlations with biomarkers indicate that leukocyte dynamics capture complementary functional information beyond static molecular markers. Overall, this approach highlights the potential of flow-based leukocyte dynamics and interpretable machine learning models as a promising functional platform for phenotyping inflammatory microvascular alterations in diabetes.