Edge-Based Temporal Fusion Transformer for Multi-Horizon Blood Glucose Prediction
Taiyu Zhu, Tianrui Chen, Lei Kuangt, Junming Zeng, Kezhi Li, Pantelis Georgiou
Abstract
Deep learning models have achieved the state of the art in blood glucose (BG) prediction, which has been shown to improve type 1 diabetes (T1D) management. However, most existing models can only provide single-horizon prediction and face a variety of real-world challenges, such as lacking hardware implementation and interpretability. In this work, we introduce a new deep learning framework, the edge-based temporal fusion Transformer (E-TFT), for multi-horizon BG prediction, and implement the trained model on a customized wristband with a system on a chip (Nordic nRF52832) for edge computing. E-TFT employs a self-attention mechanism to extract long-term temporal dependencies and enables post-hoc explanation for feature selection. On a clinical dataset with 12 T1D subjects, it achieved a mean root mean square error of 19.09 ± 2.47 mg/dL and 32.31 ± 3.79 mg/dL for 30 and 60-minute prediction horizons, respectively, and outperformed all the considered baseline methods, such as N-BEATS and N-HiTS. The proposed model is effective for multi-horizon BG prediction and can be deployed on wearable devices to enhance T1D management in clinical settings.