Dietary Behavior, Physical Activity, and 24-Hour Rhythm Coherence: A Digital Phenotyping Perspective on Vascular and Glycemic Health

Figures

• 

  Figure 1. Longitudinal multimodal phenotypic tracking of high and low rhythm coherence over 7 days. The side-by-side comparison shows daily behavioral synchronization across three rhythm layers: eating (top), activity (middle), and rest–sleep (bottom), from Day 1 to Day 7. Panel (A) shows a high-coherence profile (RCI = 0.82), characterized by restricted eating windows, synchronized physical activity peaks consistently occurring 30–60 min after food intake, and stable rest–sleep cycles with minimal day-to-day drift in mid-sleep timing. In contrast, Panel (B) shows a low-coherence profile (RCI = 0.28), marked by erratic and fragmented meal timing, frequent late-night eating, physical activity peaks uncoupled from metabolic inputs, and marked temporal misalignment of sleep windows. Digitized phenotypic markers are defined as follows: sun and cutlery icons denote meals; the sunrise-over-horizon icon denotes early-morning eating; the plain crescent-moon icon denotes evening eating before 22:00; dashed orange regions indicate eating windows; blue waveforms represent continuous activity levels; dashed blue circles highlight targeted post-meal activity; solid green bars indicate sleep intervals;…

• 

  Figure 2. Hypothesis-generating example of a possible multimodal digital phenotyping architecture for future evaluation of the Rhythm Coherence Framework. The four panels are conceptual design examples rather than a validated end-to-end system. Panel (A) illustrates possible anatomical placement of six candidate sensors: continuous glucose monitoring (CGM) on the upper arm; actigraphy and cuffless blood pressure (BP) on the wrist; skin-temperature sensing on the sternum; an electromyography (EMG) patch on the quadriceps; and a sweat patch on the forearm. Panel (B) illustrates a possible signal-processing workflow. Raw multisensor streams could undergo feature extraction to generate candidate metrics, including time in range (TIR), coefficient of variation (CV), postprandial area under the curve (PP AUC), and the Sleep Regularity Index (SRI). These examples show how a future candidate RCI might combine within-rhythm regularity and cross-rhythm coupling; the displayed 0–1 range, components, and workflow are hypothetical and unvalidated. Panel (C) provides a simplified synthetic 24 h data-stream example rather than participant data. Panel (D) is a speculative interface mock-up showing…

• 

  Figure 3. Cross-scale multimodal integration framework linking acute exposures to longitudinal cardiometabolic health trajectories through the 24 h Rhythm Coherence Index (RCI). The conceptual matrix organizes physiological, behavioral, and digital markers across five analytical dimensions: dietary behavior, physical activity, digital tools, vascular outcomes, and glycemic outcomes. These dimensions span three nested temporal scales: T1, T2, and T3. T1, the acute scale spanning minutes to hours, captures immediate postprandial nutrient flux, individual exercise bouts, and acute vascular or glycemic fluctuations. These fluctuations are detected using high-resolution digital tools, such as 1 min continuous glucose monitoring (CGM), wearable pulse amplitude tonometry (PAT), and continuous blood pressure (BP) monitoring. Thick horizontal arrows denote the temporal cascade through which acute T1 mechanisms aggregate into stabilized T2 behavioral patterns. T2 represents the 24 h scale and the primary scope of the framework. At the T2 core, integrated multisensor signals converge to operationalize the RCI through six multidimensional nodes: eating rhythm, activity rhythm, rest–sleep rhythm, 24 h…