Utilization of second derivative photoplethysmographic features for myocardial infarction classification

Myocardial infarction (MI) is a common disease that causes morbidity and mortality. The current tools for diagnosing this disease are improving, but still have some limitations. This study utilised the second derivative of photoplethysmography (SDPPG) features to distinguish MI patients from healthy control subjects. The features include amplitude-derived SDPPG features (pulse height, ratio, jerk) and interval-derived SDPPG features (intervals and relative crest time (RCT)). We evaluated 32?MI patients at Pusat Perubatan Universiti Kebangsaan Malaysia and 32 control subjects (all ages 37–87?years). Statistical analysis revealed that the mean amplitude-derived SDPPG features were higher in MI patients than in control subjects. In contrast, the mean interval-derived SDPPG features were lower in MI patients than in the controls. The classifier model of binary logistic regression (Model 7), showed that the combination of SDPPG features that include the pulse height (d-wave), the intervals of “ab”, “ad”, “bc”, “bd”, and “be”, and the RCT of “ad/aa” could be used to classify MI patients with 90.6% accuracy, 93.9% sensitivity and 87.5% specificity at a cut-off value of 0.5 compared with the single features model.     

Analysis of myocardial infarction signals using optical technique

The risk of heart attack or myocardial infarction (MI) may lead to serious consequences in mortality and morbidity. Current MI management in the triage includes non-invasive heart monitoring using an electrocardiogram (ECG) and the cardic biomarker test. This study is designed to explore the potential of photoplethysmography (PPG) as a simple non-invasive device as an alternative method to screen the MI subjects. This study emphasises the usage of second derivative photoplethysmography (SDPPG) intervals as the extracted features to classify the MI subjects. The statistical analysis shows the potential of “a-c” interval and the corrected “a-cC” interval to classify the subject. The sensitivity of the predicted model using “a-c” and “a-cC” is 90.6% and 81.2% and the specificity is 87.5% and 84.4%, respectively.