Uncertainty-Aware Real-Time Visualization of Neurovascular Coupling Degradation Along the Visual Ventral Stream in Type 2 Diabetes]

Abstract

We present a real-time, zero-install WebGL pipeline that ingests per-region brain-parameter data and renders inter-subject and group differences directly in 3D. Applied to hemodynamic response function (HRF) parameters, our anatomically grounded Glass Brain maps three quantities to three simultaneous visual channels: a wavefront whose width represents coupling coherence, a glow radius that scales with inter-subject standard deviation as a spatial uncertainty footprint, and glow luminance encoding peak blood-oxygen-level-dependent (BOLD) amplitude. Unlike conventional uncertainty visualization, which relegates inter-subject variability to error bars or separate 2D maps [2, 3], we make it a primary spatial dimension embedded in the anatomy. Any parameter set in the same format renders identically; we demonstrate the pipeline on retinotopic-mapping fMRI from young controls (n = 24), an age-matched older control group (n = 25), and Type 2 Diabetes patients (n = 17) along the ventral stream (V1 → V4). The system renders each group’s parameter differences as directly legible halos without numerical decoding; in this dataset most differences are not statistically significant (the exception is reduced V1 response amplitude in T2DM versus young controls), which the visualization conveys faithfully rather than exaggerating. The renderer runs at 60 fps in an unmodified browser. We frame this as a visualization-pipeline and encoding-design contribution; the group differences shown are preliminary (see Sec. 6).