Non-alcoholic fatty liver disease (NAFLD) is characterized by dysregulated carbohydrate and lipid metabolism, which are its primary features. However, traditional biochemical markers pose challenges for accurate quantification and visualization of metabolic states. This study introduces a novel states-based approach for accurate NAFLD assessment.

Joint probabilistic distributions of triglycerides and glycemia were constructed using dual-indicator probabilistic scatter plots based on clinical data (healthy controls: n=1978; NAFLD patients: n=471). Patterns of metabolic dysregulation were revealed through comparison against healthy profiles. Self-organizing feature mapping (SOFM) clustered the distributions into four dominant states.

Healthy scatter plots demonstrated a distinct progression of sub-states ranging from very healthy to sub-healthy. In contrast, NAFLD plots exhibited shifted probability centers and outward divergence. SOFM clustering classified the states into: mild; moderate and severe lipid metabolism disorders; and carbohydrate metabolism disorders.

Probabilistic scatter plots, when combined with SOFM clustering, facilitate a states-based quantification of NAFLD metabolic dysregulation. This method integrates multi-dimensional biochemical indicators and their distributions into a cohesive framework, enabling precise and intuitive visualization for personalized diagnosis and monitoring of prognostic developments.