A 65 nm Trustworthy Hypoglycemia Forecasting Engine Achieving 11.3 nJ per Inference

2026-06-05

AI Summary

This work addresses the unreliability of hypoglycemia prediction in continuous glucose monitoring caused by sensor noise and missing data. To this end, the authors propose a hybrid inference architecture based on probabilistic decision trees, which employs exact arithmetic evaluation in shallow layers and sampling-based reasoning in deeper layers, substantially reducing the complexity of soft decision trees. By integrating an on-chip RISC-V core with a 4×24×24 reconfigurable array of probabilistic nodes, this design is the first edge AI chip to jointly incorporate interpretability, noise robustness, and uncertainty awareness. Implemented in 65 nm CMOS technology, the system achieves an ultra-low energy consumption of only 11.3 nJ per inference while attaining an F1 score of 0.825 for 30-minute-ahead hypoglycemia prediction, demonstrating 4.1–16.1× improved robustness to sensor noise and data loss compared to conventional decision trees and random forests.

Abstract

Diabetes affects millions of people and requires reliable continuous glucose monitoring for early hypoglycemia warning. However, medical AI systems must be not only accurate and energy efficient, but also explainable, noise robust, and uncertainty aware. This work presents a 65 nm hypoglycemia forecasting engine based on probabilistic decision trees for trustworthy medical inference. The proposed hybrid architecture combines exact arithmetic evaluation for shallow tree layers with sampling based inference for deeper layers, reducing soft decision tree complexity from exponential to sample efficient traversal. A reconfigurable 4 by 24 by 24 probabilistic node array supports arbitrary tree structures with a maximum depth of 12, coordinated by an on chip low power RISC V core. Fabricated in 65 nm CMOS, the chip achieves 11.3 nJ per inference and a state of the art 30 min forecasting F1 score of 0.825 on continuous glucose monitoring data. Compared with conventional decision tree and random forest models, the proposed engine improves robustness to sensor noise and data point drop off by 4.1x to 16.1x. These results demonstrate an energy efficient, explainable, and uncertainty aware edge AI engine for trustworthy hypoglycemia forecasting.

Problem

• hypoglycemia forecasting
• trustworthy AI
• continuous glucose monitoring
• noise robustness
• uncertainty awareness

Innovation

• probabilistic decision trees
• energy-efficient AI
• trustworthy inference
• edge computing
• hypoglycemia forecasting