Statistical and entropy-based features can efficiently detect the short-term effect of caffeinated coffee on the cardiac physiology.

An electrocardiograph (ECG) is the most effective way to find the changes in cardiac physiology. It is the representation of the electrical activities of the heart and can be understood using different waves, peaks, and intervals. Several factors affect the functionality of the heart that includes lifestyle, stress, daily diet, etc. Coffee, the most widely consumed beverage in the world, is an integral part of everyday life. Caffeine, the prime constituent of coffee, is believed to affect the heart physiology. However, the effect of consumption of caffeinated coffee on the cardiac electrophysiological changes, estimated from the morphological features (e.g., peaks, waves, intervals), is controversial. This has led to the exploration of other feature extraction methods to detect the changes accurately. In recent years, the statistical and entropy-based features have emerged as an efficient method to extract hidden patterns from the ECG signal. These features have been successfully explored in arrhythmia detection, noise removal, biometric identification, etc. Hence, we hypothesized that the statistical and entropy-based features could be efficiently used in detecting the changes in the ECG signal after coffee consumption. For the evaluation of our hypothesis, 5-sec ECG segments were extracted from the recorded ECG signals from 14 volunteers in pre- and post-coffee consumption conditions. From each segment, the statistical and entropy-based features were computed. Then, the statistically significant features were extracted using Wilcoxon's signed-rank test. The results showed a significant difference in the statistical parameters post-consumption of coffee. Further, to validate our findings, several machine learning models were used for the automatic detection of these changes, and the results show the highest classification accuracy of 75%. The results support our hypothesis that the statistical and entropy-based features can efficiently detect the changes in the ECG signals, which is induced by coffee consumption. The findings of the proposed hypothesis may open up a new research arena of detecting the presence of different drugs and alcohol in the human body by analyzing the ECG signals.