Swarm intelligence based feature selection and ensemble classifiers for intensive care unit icu false alarms in arterial blood pressure signal

Abstract

Patient monitoring in Intensive Care Unit (ICU) requires collecting and processing high volumes of data. The high sensitivity of sensors leads to many false alarms, which cause alarm fatigue. Reduction of false alarms can lead to a better reaction time for medical personnel. A subset from Multiparameter Intelligent Monitoring in Intensive Care (MIMIC II) dataset was processed and annotated to aid the research related to the suppression of false alarms from ICU monitors. Applying signal processing and data mining techniques to the raw ICU measurements reduces the false alarms. Recently, automated feature engineering was performed using the signal for Arterial Blood Pressure (ABP) and a processed signal that contained the times of each heartbeat from the ABP signal. Next, Support Vector Machine (SVM), Random Forest (RF), and Extreme Random Trees (ERT) classifiers were trained to create classification models. However, the existing methods still have some issues, like noises presented in the ABP signal, QRS detection is the combination of Q wave, R wave and S wave becomes very challenging, and volumes of data need to be processed, which requires considerable computational time. newlineFast Independent Component Analysis (FICA), Intersection Kernel Principal Component Analysis (IKPCA) and Enhanced Independent Component Analysis (EICA) algorithms are proposed to reduce the noise in the ABP signal. It increases the false alarm detection rate. Haar Wavelet Transform newline