Dual-domain analysis of CGM data for personalized metabolic health management

Abstract

Time-domain metrics on a continuous glucose monitoring (CGM) miss hidden dynamics, but frequency-domain features detect up on oscillating patterns, which makes it easier to measure glucose variability and gain clinical understanding. This study aimed to analyze the time and frequency domain features of CGM data for three patients (PtID 65, PtID 109, and PtID 134) to assess glucose variability and control. CGM data were collected over a specified period, with key metrics such as mean glucose, standard deviation (SD), time in range (TIR), and spectral features being computed. The analysis included both time domain and frequency domain features. PtID 109 exhibited the highest mean glucose (207.86 mg/dL) and greatest variability, reflected in the highest SD (85.57 mg/dL) and the lowest TIR (38.24%). PtID 134 showed the most stable glucose levels with the highest TIR (73.60%) and the lowest SD (59.81 mg/dL). Frequency domain analysis revealed that PtID:134 had the highest spectral flatness, indicating more variability in the glucose signal’s frequency content. PtID 109 had the lowest TIR at 38.24%, whereas PtID 134 achieved 73.60%, indicating a 24% enhancement and demonstrating superior glucose stability. Dual-domain analysis offers thorough framework for comprehending glucose variability, facilitating individualized treatment and enhanced metabolic health care.