2026conference

SafePulse: A Memory-Safe Rust Architecture for Deterministic Real-Time ECG Alerting with Optional LLM Enrichment

Suyu Jiang, Haozhe Ruan, Shihao Wang, Jiuyuan Zhao, Luo Chen, Aquil Mirza Mohammed

2026 International Conference on Artificial Intelligence for Health and Education (ICAIHE) · Track 1: Emerging AI Technologies for Health and Education

Abstract

Processing an LLM call is simple and slow, but fast recovery alerts are essential for real-time ECG decision support. We introduce SafePulse, a system in which an uncut alarm priority contract is implemented: time constraints allow the LLM to interpret and accumulate asynchronously, but ECG alerts are transmitted synchronously via a dependent optical path. By separating each device into an individual SessionActor, the risk of shared security state at the application layer is reduced. On MIT-BIH records 100–109 the detector achieved Se = 98.6% and PPV = 98.7%. In an injected-fault matrix of 16 cells (4 fault profiles × 4 concurrency tiers, 1000 trials/cell), event-level deterministic streams showed tuple mismatch count 0, with Δalerts = 0 and Δp95/p99/p999 = 0.0 μs. This study integrates LLM enrichment in a safety-constrained monitoring pipeline and demonstrates its architecture and validation mode. Extension code, figures, and complete results are provided as Electronic Supplementary Material.

Affiliations

Hubei University of Arts and Science, China · The Hong Kong Polytechnic University (PolyU), Hong Kong · City University of Hong Kong, Hong Kong

Status: Accepted as Full Paper – ICAIHE 2026, Waseda University, Tokyo (Paper ID 68; reviewer scores 3 / 4)

Presenter: Suyu Jiang

Keywords

ECG monitoringrustactor modelLLM enrichmentdigital health decision support

BibTeX

@inproceedings{jiang2026safepulse,
  author = {Jiang, Suyu and Ruan, Haozhe and Wang, Shihao and Zhao, Jiuyuan and Chen, Luo and Mohammed, Aquil Mirza},
  title = {SafePulse: A Memory-Safe Rust Architecture for Deterministic Real-Time ECG Alerting with Optional LLM Enrichment},
  booktitle = {2026 International Conference on Artificial Intelligence for Health and Education (ICAIHE)},
  year = {2026},
  note = {Track 1: Emerging AI Technologies for Health and Education},
  keywords = {ECG monitoring, rust, actor model, LLM enrichment, digital health decision support}
}