Keywords

1 Introduction

Assessment of surgical pain has been attempted with difficulties because it is not easy to describe degree of pain objectively. There have been many attempts to quantify pain such as skin conductance-based method [1, 2], analgesia nociception index (ANI, Mdoloris Medical System, Lille, France) [3, 4] and surgical plethysmographic index (SPI, GE Healthcare, Chicago, USA) [5]. Skin conduction was reported to correlate significantly with a numeric rating scale (NRS) in postoperative pain [6] and ANI seems more sensitive than heart rate and systolic blood pressure to moderate nociceptive stimuli in propofol-anaesthetized patients [7]. Especially, SPI which based on continuous cardiovascular variables like heartbeat intervals (HBI) and photoplethysmography amplitude (PPGA) was evaluated as a promising index of postoperative pain assessment [8]. The result of SPI could be interpreted as changes of heart rate and blood volume with sympathetic activation [9, 10]. In this study, we consider the possibility of assessing surgical pain by analyzing PPG amplitude variability (PPGV) as another pain indicator. For this, we assumed that the PPGV has relevance with blood pressure variability (BPV), which is a method to measure systolic blood pressure and detect changes in the short or long term and estimates the physiological state via the mean of the measured blood pressure or a standard deviation [11], because blood volume could be proportionally changed with blood pressure excluding the consideration of vascular property. In analyzing PPGV, we adapted frequency domain analysis metric of heart rate variability (HRV).

2 Methods

2.1 Dataset

Sixty surgical patients (29 males, 31 females, 52.1 ± 11.4 years old) were enrolled to the experiment, and 58 pair of dataset was used except 2 pair of dataset which has recording error. According to the protocol, PPG data was recorded before and after surgery, and dataset recorded before and after surgery were regarded as a painless and pain dataset, respectively. Every record has 6-min length of 300 Hz sampling frequency. To calculate the area of PPG waveform, every upper and lower peak location of the PPG waveform was detected and validated by proficient researchers.

2.2 Parameter Extraction

Derived parameters were defined as the amplitude from baseline to systolic peak (ACAbaseline), the amplitude from diastolic peak to systolic peak (ACAdia), the amplitude difference of adjacent systolic peaks (PPGAVsys), the amplitude difference of adjacent diastolic peaks (PPGAVdia), the ratio of PPGAVsys to ACAbaseline (PPGAVsys/ACAbaseline), the ratio of PPGAVsys to HPH (PPGAVsys/HPH), the ratio of PPGAVdia to ACA (PPGAVdia/ACA), and the ratio of PPGAVdia to HPH (PPGAVdia/HPH). Figure 1 shows the graphical representation of parameters, and Eqs. (1)–(5) represents mathematical equation of ACAdia, BP, ACAbaseline, PPGAVsys, PPGAVdia, respectively (see Table 1).

Fig. 1
figure 1

PPG waveform and derived parameters

Full size image
Table 1 Extracted parameters
Full size table
$$ ACA_{dia} = PPGA_{sys} \left( k \right) - PPGA_{dia} \left( k \right) $$
(1)
$$ \begin{aligned} BP\left( k \right) = & \,\frac{{PPGA_{dia} \left( {k + 1} \right) - PPGA_{dia} \left( k \right)}}{{Time_{dia} \left( {k + 1} \right) - Time_{dia} \left( k \right)}} \\ & \,\left\{ {Time_{sys} \left( k \right) - Time_{dia} \left( k \right)} \right\} + PPGA_{dia} \left( k \right) \\ \end{aligned} $$
(2)
$$ ACA_{baseline} = PPGA_{sys} \left( k \right) - BP\left( k \right) $$
(3)
$$ PPGAV_{sys} = PPGA_{sys} \left( {k + 1} \right) - PPGA_{sys} \left( k \right) $$
(4)
$$ PPGAV_{dia} = PPGA_{dia} \left( {k + 1} \right) - PPGA_{dia} \left( k \right) $$
(5)

2.3 Validation

Every parameter was derived in pain and painless dataset, and it was verified whether there is significant different using paired t-test.

3 Result and Discussion

Table 2 shows that the mean, standard deviation, coefficient of variation and significance of difference of the result according to the nociceptive pain. As a result, statistical difference was not found in TP, VLF, LF, HF, nLF of ACAbaseline, TP, LF, HF, nLF of ACAdia, TP, VLF, KF, HL, nHF of PPGAVsys, TP, VLF, LF, HF, nLF, nHF of PPGVdia, nHF of PPGAVsys/ACAbaseline, nHF, LF/HF of PPGAVsys/ACAdia and nLF, nHF of PPGAVdia/ACAdia, nLF, nHF of PPGAVdia/ACAbaseline. LF component of BPV reflects sympathetic vasomotor activity which is known as Mayer wave [12,14,15,16,17,17]. Otherwise, high frequency component of BPV reflects mechanical function of hemodynamics according to the pressure change inside thoracic cage [15]. In the result of this study, LF value of ACAbaseline and ACAdia was decreased but there is no statistical significance, moreover, PPGAVsys and PPGAVdia also shows increases with no significance. However, in every normalized parameter, LF value is increased with statistical significance (p < 0.05). This result could be interpreted that normalization reduced the ambiguity of PPG amplitude which is measured as an arbitrary unit. In terms of parameters, amplitude variability parameter, PPGAVsys, PPGAVdia shows clear classification result compared with simple amplitude-based variables, ACAbaseline and ACAdia. This result suggests that the frequency domain variables of PPGV has a possibility in pain quantification and that the significant changes were reflected better in normalized variables.

Table 2 Statistical changes of derived parameters
Full size table

4 Conclusion

In this study, we investigated PPGV in surgical pain quantification. We derived frequency domain variable of PPGV before and after surgery and found that there is a significant difference (p < 0.05) in PPGV variables according to the pain stimuli. Especially, result indicated that a clear difference was found in normalized PPGV variables. Consequently, the result of this study suggests that the possibility of frequency domain analysis of PPGV as a novel indicator of surgical pain quantification.