Directional Drift in Biologically Meaningful Vector Planes: A Proposed Geometric Framework for Early Detection of Subthreshold Disease

BackgroundMost conventional diagnostic systems rely on fixed thresholds to differentiate disease states from normal. However, early pathological changes may begin before these thresholds are crossed. Therefore, a system that works in this pre-threshold state can potentially lead to earlier diagnosis. ObjectiveTo propose and evaluate a geometric framework that models early disease as a directional drift from a physiological plane to a pathological plane, allowing for pre-threshold detection using biologically interpretable variables. MethodsIn this modeling study on synthetic data derived from published clinical trends, two clinically meaningful variables were used to define a 2D feature space. The model was applied to a synthetic dataset of 4000 eyes divided into four phenotypes: normal stable (NS), early disease stable (ED_S), early disease progressive (ED_P), and pre-threshold progressive (PT_P). A physiological plane was constructed using range-normalized values from the NS group. A canonical disease vector was derived from the ED_P group. Each subjects follow-up data was transformed into a subject-specific drift vector, and the Composite Drift Score (CDS) was calculated as the product of directional alignment (Directional Emphasis Multiplier, DEM) and a Magnitude-to-Noise Ratio (MNR). ResultsCDS increased significantly over follow-ups in both ED_P and PT_P, distinguishing them from the two stable cohorts (p<0.001). DEM and MNR components showed consistent trends, with progressive cases exhibiting higher alignment with the disease vector and supra-noise magnitude of change. Visual and statistical analyses confirmed early drift detection even within numerically normal ranges. ConclusionIn this early modeling study based on simulation data, we could quantify the directional drift with a unitless, interpretable metric (CDS) and its derivatives. It showed similar trends in pre-threshold groups as early disease groups, showing potential for further evaluation.