Premium
Signal Processing and Machine‐learning Algorithm to classify anesthesia depth
Author(s) -
Reyes-Cruz Dario,
Mosquera Oscar Leonardo,
Botero-Rosas Daniel Alfonso,
Gallego-Correa John Jairo,
Leon-Ariza Henry H.,
Duarte-Tamara Santiago,
Botero-Machado Juan David,
Rozo-Cespedes Laura,
Garcia-Ramos Andres Felipe,
Padron-Ayala Antonietta
Publication year - 2020
Publication title -
the faseb journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.709
H-Index - 277
eISSN - 1530-6860
pISSN - 0892-6638
DOI - 10.1096/fasebj.2020.34.s1.05484
Subject(s) - anesthetic , statistic , medicine , anesthesia , artificial intelligence , artificial neural network , feedforward neural network , machine learning , algorithm , computer science , mathematics , statistics
Background An algorithm with complexity measures, nonlinear dynamics, and neural networks were developed to detect and classify the anesthetic deep (AD). It has been shown that extreme anesthetic depth has been correlated with an increased risk of mortality. Likewise, intraoperative awareness has been reported in the anesthetic drugs under dosage. Methods Our artificial intelligence classifier algorithm was developed using a Complexity Brainwave Index (CBI) and a heart rate variability parameter (HRVP). Data from 60 patients (25 men and 35 women), adults between 18 and 65 years of age undergoing surgical procedures (ASA I–II) were collected. The classifier was designed using multi‐layer feedforward neural networks with a hyperbolic activation function using the patient data set and their anesthesia records. Results The CBI by itself showed a better prediction probability with Pk of 0.935 when compared with Datex‐Ohmeda index in which State Entropy and Response Entropy had a Pk of 0.884 and 0.899 respectively; also, our CBI showed a performance of 99% of the awake state and 93.3% for deep anesthesia state. Additionally, when combining the CBI and the HRVP, the C‐Statistic was 99% for the awake state, 87.41% for light anesthesia, 82,46% for general anesthesia, and 93.33% for deep anesthesia. Conclusions Our results demonstrate that special biological signal filtering and a machine learning algorithm may be used to classify AD during a surgical procedure on ASA I–II patients, assisting anesthesiologists and clinicians in decision making. Support or Funding Information University of La Sabana