Zhao Lulu, Yang Licai, Su Zhonghua, Liu Chengyu
School of Control Science and Engineering, Shandong University, Jinan, China.
Second Affiliated Hospital of Jining Medical College, Jining, China.
Front Physiol. 2019 Mar 29;10:359. doi: 10.3389/fphys.2019.00359. eCollection 2019.
This study used entropy- and cross entropy-based methods to explore the cardiorespiratory coupling of depressive patients, and thus to assess the values of those entropy methods for identifying depression patients with different disease severities.
Electrocardiogram (ECG) and respiration signals from 69 depression patients were recorded simultaneously for 5 min. Patients were classified into three groups according to the Hamilton Depression Rating Scale (HDRS) scores: group Non-De (HDRS 0-7), Mid-De (HDRS 8-17), and Con-De (HDRS >17). Sample entropy (SEn), fuzzy measure entropy (FMEn) and high-frequency power (HF) were computed on the original RR interval time series and breath-to-breath interval time series. Cross sample entropy (CSEn) and cross fuzzy measure entropy (CFMEn) were computed on interval time series resampled at 2 Hz and 4 Hz, respectively. The difference among three patient groups and correlation between entropy values and HDRS scores were analyzed by statistical analysis. Surrogate data were also employed to confirm the validation of entropy measures in this study.
A consistent increasing trend has been found among most entropy measures from Non-De, to Mid-De, and to Con-De groups, and a significant ( < 0.05) difference in FMEn of RR intervals exists between Non-De and Mid-De or Con-De groups. Significant differences have been also found in all cross entropies, between Non-De and Con-De groups and between Mid-De and Con-De groups. Furthermore, significant correlations also exist between HDRS scores and FMEn of RR intervals ( = 0.24, < 0.05), CSEn at 4 Hz ( = 0.26, < 0.05) or 2 Hz ( = 0.28, < 0.05) resampling, and CFMEn at 4 Hz ( = 0.31, < 0.01) or 2 Hz ( = 0.30, < 0.05) resampling. A significant difference of cardiorespiratory coupling parameters between different depression stages and significant correlations between entropy measures and depression severity both indicate central autonomic dysregulation in depression patients and reflect varying degrees of vagal modulation reduction among different depression levels. Analysis based on surrogate data confirms that the non-linear properties of the physiological signals played a major role in depression recognition.
The current study demonstrates the potential of cardiorespiratory coupling in the auxiliary diagnosis of depression based on the entropy method.
本研究采用基于熵和交叉熵的方法来探究抑郁症患者的心肺耦合情况，从而评估这些熵方法在识别不同疾病严重程度的抑郁症患者方面的价值。
同时记录69名抑郁症患者的心电图（ECG）和呼吸信号，持续5分钟。根据汉密尔顿抑郁量表（HDRS）评分将患者分为三组：非抑郁组（HDRS 0 - 7）、中度抑郁组（HDRS 8 - 17）和重度抑郁组（HDRS > 17）。在原始RR间期时间序列和逐次呼吸间期时间序列上计算样本熵（SEn）、模糊测度熵（FMEn）和高频功率（HF）。分别在2Hz和4Hz重采样的间期时间序列上计算交叉样本熵（CSEn）和交叉模糊测度熵（CFMEn）。通过统计分析三组患者之间的差异以及熵值与HDRS评分之间的相关性。还采用替代数据来确认本研究中熵测量的有效性。
在大多数熵测量中，从非抑郁组到中度抑郁组再到重度抑郁组呈现出一致的上升趋势，非抑郁组与中度抑郁组或重度抑郁组之间RR间期的FMEn存在显著差异（< 0.05）。在所有交叉熵中，非抑郁组与重度抑郁组之间以及中度抑郁组与重度抑郁组之间也存在显著差异。此外，HDRS评分与RR间期的FMEn（= 0.24，< 0.05）、4Hz（= 0.26，< 0.05）或2Hz（= 0.28，< 0.05）重采样时的CSEn以及4Hz（= 0.31，< 0.01）或2Hz（= 0.30，< 0.05）重采样时的CFMEn之间也存在显著相关性。不同抑郁阶段之间心肺耦合参数的显著差异以及熵测量与抑郁严重程度之间的显著相关性均表明抑郁症患者存在中枢自主神经调节异常，并反映了不同抑郁水平之间迷走神经调节降低的不同程度。基于替代数据的分析证实，生理信号的非线性特性在抑郁症识别中起主要作用。
本研究证明了基于熵方法的心肺耦合在抑郁症辅助诊断中的潜力。
