Application of multiscale Poincaré short-time computation versus multiscale entropy in analyzing fingertip photoplethysmogram amplitudes to differentiate diabetic from non-diabetic subjects

• Published in: Computer methods and programs in biomedicine Vol. 166; pp. 115 - 121
• Main Authors: Haryadi, Bagus, Liou, Juin J., Wei, Hai-Cheng, Xiao, Ming-Xia, Wu, Hsien-Tsai, Sun, Cheuk-Kwan
• Format: Journal Article
• Language: English
• Published: Ireland Elsevier B.V 01.11.2018
• Subjects: Adult; Aged; Algorithms; Blood Pressure - physiology; Diabetes Mellitus - diagnosis; Diabetes type 2; Diagnosis, Computer-Assisted - methods; Entropy; Female; Heart Rate - physiology; Humans; Male; Middle Aged; Models, Statistical; Multiscale entropy (MSE); Multiscale Poincaré (MSP); Photoplethysmogram; Photoplethysmography; Pulse amplitudes; Reproducibility of Results; Signal Processing, Computer-Assisted; Software; Time Factors; Multiscale Poincaré (MSP); Multiscale entropy (MSE); Photoplethysmogram; Pulse amplitudes; Diabetes type 2
• ISSN: 0169-2607; 1872-7565; 1872-7565
• DOI: 10.1016/j.cmpb.2018.10.001

•Non-invasive detection of diabetic microvascular anomaly is vital for early control.•Multiscale Poincaré (MSP) sensitivity to diabetic microvascular change was assessed.•MSP was superior to multiscale entropy for detecting diabetic microvascular anomaly.•Compared with multiscale entropy, MSP has notably lower computation load. Multiscale Poincaré (MSP) plots have recently been introduced to facilitate the visualization of time series of physiological signals. This study aimed at investigating the feasibility of MSP application in distinguishing subjects with and without diabetes. Using photoplethysmogram (PPG) waveform amplitudes acquired from unilateral fingertip of non-diabetic (n = 34) and diabetic (n = 30) subjects, MSP indices (MSPI) of the two groups were compared using 1000, 500, 250, 100 data points. Data from Poincaré index (short-term variability/long-term variability [i.e. SD1/SD2] ratio, SSR) and multiscale entropy (MSE) were also obtained with the four corresponding data points for comparison. SSR and MSPI were both negatively related to glycated hemoglobin (HbA1c) and fasting blood sugar levels. Significant negative correlation was also noted between MSPI and pulse pressure. When only 500 and 250 data points were included, significant elevations in the non-diabetic group were only noted in MSPI (both p < 0.01). Furthermore, MSPI was significantly higher in non-diabetic than that in diabetic subjects on all scales (i.e., 1–10) but not using MSE when utilizing 1000 data points. The results demonstrated enhanced sensitivity of MSP in differentiating between non-diabetic and diabetic subjects compared to SSR and MSE, highlighting the feasibility of MSP application in biomedical data analysis to reduce computational time and enhance sensitivity.