TY  - JOUR
PB  - Korean Institute of Electrical Engineers
SN  - 19750102
EP  - 2738
AV  - none
SP  - 2715
TI  - Mechanical Ventilator Pressure and Volume Control Using Classifier Machine Learning Algorithm for Medical Care
N1  - cited By 0
Y1  - 2024///
VL  - 19
A1  - Anitha, T.
A1  - Gopu, G.
A1  - Arun Mozhi Devan, P.
JF  - Journal of Electrical Engineering and Technology
UR  - https://www.scopus.com/inward/record.uri?eid=2-s2.0-85180178100&doi=10.1007%2fs42835-023-01744-0&partnerID=40&md5=e0cbba44a27c2ca6b4e82feaf28f9546
ID  - scholars19713
KW  - Controllers; Decision trees; Intensive care units; Iterative methods; Learning algorithms; Machine learning; Nearest neighbor search; Two term control systems; Ventilation
KW  -  'current; Asynchrony; Critically-ill patients; Iterative learning controller; Machine learning algorithms; Machine-learning; Mechanical; Mechanical ventilation; Neural-networks; Ventilator
KW  -  Three term control systems
N2  - The mechanical ventilation technique is crucial for saving the lives of critically ill patients in the Intensive Care Unit. However, there can be a mismatch between the patientâ??s needs and the ventilator settings, which can cause patient-ventilator asynchrony. Our research aims to tackle this issue by implementing a novel current cyclic feedback type iterative learning PID controller (ILCPID) to achieve the desired pressure and volume. The research also provides a concise and comprehensive study to identify the most effective machine learning methodology for developing adequate ventilation models. In addition, the ILCPID controllerâ??s accuracy and effectiveness are further validated using machine learning-based classifiers that can predict and identify the best model for the different ventilator modes, reducing the risk of mechanically ventilated patients. Among the different classifiers, the proposed narrow neural network achieved an accuracy of 92.4 and 89.29 for pressure and volume, respectively. Other techniques, such as wide neural networks, coarse trees, K-nearest neighbours, and decision trees, were also compared for accuracy, training duration, specificity, and sensitivity. © The Author(s) under exclusive licence to The Korean Institute of Electrical Engineers 2023.
IS  - 4
ER  -