An Analog Nearest Class with Multiple Centroids Classifier Implementation, for Depth of Anesthesia Monitoring

Vassilis Alimisis, Vassilis Mouzakis, Georgios Gennis, Errikos Tsouvalas, Paul P. Sotiriadis

2022 · 24 citations · 24 references

Concepts

Abstract

Monitoring the Depth of Anesthesia on a patient is crucial to maintain a safe sedation state during a surgical operation. A high dosage can directly affect the patient's health, while a low one may disrupt the operation and, in turn, lead to unavoidable damage. To this end, this work proposes a novel, low power <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(1.7\mu W)$</tex> , low voltage (0.6V) analog architecture of a Nearest Class with Multiple Centroids classifier for depth of anesthesia monitoring. The architecture consists of a bell-shaped function circuit and the Lazzaro argmax operator circuit. To verify the proper operation of the proposed classifier a real-world depth of Anesthesia dataset is utilized. Post-layout simulation results were compared with software-based ones to confirm the high accuracy of the proposed design. The implemented architecture was realized and simulated in a TSMC 90nm CMOS process, using the Cadence IC Suite.

References

24