DPA-Net: A Dual-Path Attention Neural Network for Inferring Glycemic Control Metrics from Self-Monitored Blood Glucose Data

TL;DR

This work tackles estimating AGP glycemic metrics from sparse SMBG data by introducing DPA-Net, a dual-path attention network that jointly reconstructs a CGM-like trajectory and directly predicts TIR, TAR, and TBR. The upper Spatial-Channel Attention path enforces physiologic trajectory reconstruction, while the lower multi-scale ResNet path yields direct metric predictions; their outputs are aligned through a dedicated loss to reduce bias and overfitting. An active point selector (AETCN) captures realistic SMBG sampling patterns, further improving robustness. Experimental results on Jaeb Center data show that DPA-Net achieves high accuracy with low bias, outperforming a No-Interp SMBG baseline and demonstrating potential for accessible, glucose-control monitoring in settings where CGM is not available.

Abstract

Continuous glucose monitoring (CGM) provides dense and dynamic glucose profiles that enable reliable estimation of Ambulatory Glucose Profile (AGP) metrics, such as Time in Range (TIR), Time Below Range (TBR), and Time Above Range (TAR). However, the high cost and limited accessibility of CGM restrict its widespread adoption, particularly in low- and middle-income regions. In contrast, self-monitoring of blood glucose (SMBG) is inexpensive and widely available but yields sparse and irregular data that are challenging to translate into clinically meaningful glycemic metrics. In this work, we propose a Dual-Path Attention Neural Network (DPA-Net) to estimate AGP metrics directly from SMBG data. DPA-Net integrates two complementary paths: (1) a spatial-channel attention path that reconstructs a CGM-like trajectory from sparse SMBG observations, and (2) a multi-scale ResNet path that directly predicts AGP metrics. An alignment mechanism between the two paths is introduced to reduce bias and mitigate overfitting. In addition, we develop an active point selector to identify realistic and informative SMBG sampling points that reflect patient behavioral patterns. Experimental results on a large, real-world dataset demonstrate that DPA-Net achieves robust accuracy with low errors while reducing systematic bias. To the best of our knowledge, this is the first supervised machine learning framework for estimating AGP metrics from SMBG data, offering a practical and clinically relevant decision-support tool in settings where CGM is not accessible.

Figures (10)

• Figure 1: AGP times-in-range “thermometer” (TAR/TIR/TBR).
• Figure 2: Overall architecture of the proposed Dual-Path Attention Network (DPA-Net). The input $\Theta_{\mathrm{in}}$ combines the data matrix ($M_s$), mask matrix ($M_m$), and positional embedding ($M_p$). The upper SCA Path reconstructs dense CGM from sparse SMBG, while the lower multi-scale ResNet path directly predicts long-term glycemic range metrics (TIR, TAR, TBR). Three complementary loss functions ($L_{\mathrm{rc}}, L_a, L_{\mathrm{tr}}$) jointly supervise training.
• Figure 3: Channel attention block.
• Figure 4: Spatial Transformer Block.
• Figure 5: Network structure for Spatial-Channel Attention (SCA).