Automatic detection of anomalies in blood glucose using a machine learning approach

Some papers find irregular patterns in vital signs time series such as abnormal episodes in ECG pulses, SpO2 signal, and blood glucose level, mostly discovering unusual temporal patterns in continuous data.

“Some of the papers considered finding irregular patterns in vital signs time series such as abnormal episodes in ECG pulses [31,32], SpO2 signal [33] and blood glucose level [28], which mostly discover unusual temporal patterns in continuous data.”

SVM, Markov models, and Wavelet analysis are used in health monitoring systems for anomaly detection.

“For example, support vector machines [27], Markov models [28] and Wavelet analysis [29] are used in health monitoring systems for anomaly detection.”

Anomaly detection can use offline techniques to detect abnormal readings for each individual based on historical measurements, since abnormal patterns can differ per subject.

“However, anomaly detection task could use offline techniques in order to detect abnormal reading for each individual based on the historical measurements for the person [28], since the definition of the abnormal patterns can be different for each subject.”

Data simulation is useful when the focus of the data processing method is on efficiency and robustness of information extraction rather than handling real-world data including artifacts, errors, and conditions of the data gathering environment.

“Data simulation would be useful when the more focus of data processing method is on the efficiency and robustness of information extraction [28,57] rather than handling real-world data including the artifact, errors, conditions of data gathering environment, etc.”

Another reason to create and use simulated data is the lack of long term and large scale data sets.

“Another reason to create and use simulated data is the lack of long term and large scale data sets [28] which helps the proposed data mining systems to deal with huge amount of data.”

Blood glucose monitoring is an example of long term data analysis for the sake of right decision making.

“Blood glucose monitoring is an example of long term data analysis for the sake of right decision making [23,28].”

Due to the nature of vital signs, some are essentially counted as continuous such as ECG, heart rate and respiration rate, or discrete like blood glucose level.

“Due to the nature of vital signs, some of them are essentially counted as continuous such as ECG, heart rate and respiration rate [31,42] or discrete like blood glucose level [28].”

Usually single sensor data have been used for specific analysis on individual physiological data such as ECG signal analysis or blood glucose monitoring.

“Usually single sensor data have been used for specific analysis on individual physiological data such as ECG signal analysis [32,54,92] or blood glucose monitoring [23,28].”