1.1.

State-of-the-Art

The most used noninvasive methods to assess arterial elasticity conditions are based in photoplethysmography, ultrasound, and arterial tonometry.³ Photoplethysmography uses infra-red optical signals to measure the volumetric pulsation of the blood flow. Even though being considered the most user-friendly method, this sensor can mainly be applied in the body extremities, like fingers and earlobes.⁵ Ultrasound systems are also used to record arterial volume changes, being able to monitor the central arteries directly. Even though they are considered the most accurate assessment methods, their main drawbacks are the high expertise technician requirement and constant control of the ultrasound beam direction.⁶ A third method uses applanation tonometry with piezoresistive pressure sensors, which is performed by placing the sensor over the artery, depressing the sensor to flat the artery.³ To access data the sensor has to be moved until a clear pulse is found and the pressure imposed by the operator has to be carefully determined. An example of the arterial tonometry state-of-the-art is the SphygmoCor (AtCor—Australia) device.³ The main disadvantage of this device is the difficulty to assess the pulse in central arteries, such as the carotid. Usually, obtaining a steady pulse wave is only possible in the radial artery, on which inverse transfer functions are applied to access the central pulse waveform.

The use of FBGs in noninvasive CV monitoring is not largely reported. In 2005, A. van Brakel, et al., described a twin Bragg grating Fabry-Perot interferometer as blood pressure manometer used in the radial artery.⁷ The proposed sensor can be considered a tonometer based on fiber Bragg gratings (FBGs). Having a strain high sensitivity, the FBGs can be an alternative to the present sensors in order to ease the pulse wave acquisition directly in the carotid artery. The presented sensor uses one pretensioned optical fiber with a FBG to detect the pulse waveform through the deformation resulting from the carotid artery micromovements.

This paper is organized as follows: after the introduction, the sensor design is presented in Sec. 2, where the sensor calibration and implementation methods are also explained. In Sec. 3 are shown the results obtained in human carotids. Finally, the work main conclusions are presented.

1.2.

Carotid Pulse Waveform Analysis

The central arterial pulse is a combination between the forward wave (FW) created by the left ventricle and its reflection [reflected wave (RF)]. The FW is reflected in arterial impedance mismatches, such as in the transitions between elastic and muscular arteries; in arterial bifurcations; in vasomotor tone and geometry chances of small arteries.³ The carotid pulse waveform (CPW) reproduces the central hemodynamic conditions due to its elastic nature like aorta, and being its first bifurcation.

When the central arteries maintain its healthy structure, the pulse wave is spread through the arteries at a low velocity and the reflected wave tends to arrive back in the aorta during the late systole or early diastole. Conversely, in stiff arteries pulse velocity rises and the RW arrives in the central arteries in the early systole adding to the FW, increasing the systolic pressure (SP). The RW effects in the pulse pressure (PP) increasing are usually quantified by the Augmentation Index (AIx), which relates the augmentation pressure (AP) and the PP presented in the Fig. 1.³

Fig. 1

Carotid pulse waveform schematics. (F-wave-foot, IP-inflection point; DN-dicrotic notch; DW-dicrotic wave; PP-pulse pressure).

In young subjects the inflection point (IP) occurs after the SP, due to the late arrival of the RW in the central arteries, and the SP is defined by the FW. In this case, the AIx is negative. As shown in Fig. 1, in stiffer arteries the IP occurs before SP as a consequence of an early arrival of the RW overlapping the FW, which is characteristic of a stiffer arterial tree, resulting in positive AIx.³

The early reflection of the pulse wave could be responsible for: (i) central PP increasing and consequent increase of the risk for stroke and renal failure; (ii) load increasing in the left ventricle during systole, which could increase the ventricular mass and lead to hypertrophy and cardiac failure; iii) decreasing of coronary arteries irrigation, increasing the myocardium ischemia and angina risk.⁸ The AIx has been correlated with several CV conditions, being an important index of CV risk. Previous studies have related the AIx with all-cause mortality in patients with end-stage renal failure, with the degree of coronary artery disease and global CV risk. Moreover, the prognostic value of APW analysis evolution has been studied during anti-hypertensive pharmacological and congestive heart failure treatments.⁹

Since the elastic arteries suffer more stiffness effects during life, it is crucial to assess the pulse waveform in the central arteries.³ The carotid location was chosen due to its closer location to aorta, presenting the same elastic nature.

1.3.

Fiber Bragg Grating Technology

The increasing application of optical fiber technology in the sensing field is mainly due to the wide operation range, reduced losses, electrical isolation, wavelength codification (resulting in high signal-to-noise ratio) and immunity to electromagnetic interferences. This last feature is highly desirable in medical devices, allowing the application of these sensors in high electromagnetic fields environments, unlike the electronic sensors, such as found during magnetic resonance imaging medical exams.

In this work the sensor is based in an FBG, that can be described as a periodical modulation of the optical fiber core refractive index, with an approximately amplitude of 0.1%. When a spectral wide optical signal is guided through the optical fiber, a consecutive reflection of the signal will occur in the recorded grating. The wavelength, for which all the reflected waves are in phase, is known as the Bragg wavelength (${\lambda }_B$). In a fiber with an effective refractive index in the guided mode, $n_{\mathrm{ef}}$, the Bragg wavelength is given by Ref. 10:

In the FBG based sensors, the changes of temperature and/or strain applied to the fiber impose a shift of the Bragg wavelength (${\lambda }_B$), represented by Ref. 10:

2.1.

Sensor Assembly and Signal Detection

The used FBG sensors were produced in photosensitive single mode optical fiber FiberCore PS1250/1500. The FBG, with a 4 mm length, was inscribed by an automated phase-mask method, using a UV excimer laser peaking at 248 nm, presenting a Bragg wavelength of 1550.2 nm at 23°C.

The proposed sensor consists in a small outer plastic box ($35\times 35\times 20\mathrm{mm}$) crossed by an optical fiber with the FBG placed in the middle. The fiber was pretensioned to become sensitive to elongation and compression. As shown in Fig. 2, a latex membrane was placed between the box and the stainless steel sphere (with a diameter of 9 mm), allowing vertical movements with the carotid arterial distensions. During the acquisition process, the sensor is placed over the carotid artery and the obtained signal is monitored in real time using a computer.

Fig. 2

Fibre Bragg probe configuration: (a) Schematics; (b) prototype; and (c) placement over carotid artery.

For the sensor interrogation the signal from a broadband optical source, $C+L$ amplified spontaneous emission (ASE) source from Amonics-China, is injected into the optical fiber, reflecting in the Bragg grating, and directed by an optical circulator to an interrogation system as schemed in Fig. 3. The reflected signal was analyzed with the interrogation monitor I-MON 512E-USB from Ibsen-Denmark with an acquisition sampling rate of 950 Hz.

Fig. 3

Interrogation setup schematics.

2.2.

Sensor Calibration

The system was calibrated with a controlled vertical movement imposed on the sphere. For this task it was used a micrometric translation stage to displace the metallic sphere simultaneously with the Bragg wavelength monitoring.

The sensor response to standard mechanical waveforms created by an arbitrary waveform generator from Agilent Technologies-USA, model 33220A, was studied and used for calibration. The scheme of this setup is shown in Fig. 4. The electric signal from the waveform generator was amplified by a High-Voltage PieZoelecTric (HVPZT) amplifier, model E-508.00 from Physic Instrumente (PI-Germany). The used actuator stage (Z/Tilt Piezoelectric Flexure Stage model P287.70 from PI) was chosen due to its movement range being close to the carotid movements, with a maximum displacement of 700 μm. The sensor was placed with the top of the sphere in contact with the moving part of the flexure stage. The movement of this stage was controlled using several generated waveforms, we only present the sensor response to “type A” and “C” Murgo waveforms.¹² The generated pressure waves were monitored with an Oscilloscope from Tektronix-USA, model TK-220.

Fig. 4

Setup of the sensor mechanical characterization with arbitrary waveforms.

2.3.

Data Acquisition in Human Carotids

3.1.

Sensor Calibration

The sensor system has a sphere that moves vertically following the carotid distention pattern. The calibration of the sensor response to a controlled vertical movement is shown in Fig. 5.

Fig. 5

Bragg wavelength variation with the vertical sphere movement. Points are experimental data and line is the linear fit ($r^2>0.996$), $\text{slope}=2.04\times {10}^{-3}\mathrm{nm}/\mu \mathrm{m}$, $\text{interception}=1550.36\mathrm{nm}$.

The Bragg wavelength dependence with the sphere displacement is described by a linear function, with a fit correlation factor ($r^2$) of 0.98. Considering this approach, the pulse waveform assessed with the sensor is similar to the real pulse waveform, changing only the baseline.

Before the developed sensor application in human carotids, its response to a mechanical actuator with a reference pattern was studied. For this test several waveforms were considered, namely for “A” and “C type” waveforms. As it can be seen in Fig. 6, the FBG sensor detects and follows the piezoelectric actuator path, allowing the identification of all characteristic points of the waves, even the ones with relatively low amplitude differences. These results supported the application of this sensor to assess the pulse wave in human carotids.

Fig. 6

Bragg sensor response of the actuator motion (type A and C patterns, respectively). (Dashed line—actuator motion, full line—sensor response).

To measure the accuracy of the sensor response, it was considered the root mean square deviation (RMSD) of normalized amplitude measured values, which is given by:

3.2.

Data Processing

The sensor was tested in four subjects named as subjects I, II, III and IV. In Fig. 7 it can be seen the raw and filtered data acquired in subject I. This data demonstrates that the applied FFT filter reduces substantially the increasing tendency of the signal, which was due to temperature changes in the sensor when placed in contact with the skin.

Fig. 7

Raw data (left) and filtered data (right) acquired of the subject I.

From the filtered data, 10 consecutive pulses were selected and normalized. From these pulses an average pulse is calculated, which can be analyzed in terms of temporal and amplitude characteristics. In Fig. 8, 10 cycles of the processed data acquired in the four subjects are presented. It can be seen that for the same subject the wave pattern is similar in time. For all subjects, the five relevant identification points of the PP wave in the carotid artery were calculated with the 1st derivative of the average pulse wave.

Fig. 8

Acquisition of pulse waveform in the carotid artery of four human subjects.

The subjects I, III, and IV presented type C waves, corresponding to negative ${\mathrm{AI}}_{\mathrm{x}}$, with values of ${\mathrm{AI}}_{\mathrm{xI}}=-0.10\pm 0.04$; ${\mathrm{AI}}_{\mathrm{xIII}}=-0.28\pm 0.05$ and ${\mathrm{AI}}_{\mathrm{xIV}}=-0.13\pm 0.03$. The subject II had a slightly positive ${\mathrm{AI}}_{\mathrm{xII}}=0.04\pm 0.04$. All subjects had a unique identification curve, presenting different ${\mathrm{AI}}_{\mathrm{x}}$ values. In the Fig. 9 it can be seen the RMSD of each pulse related to the maximum deviation. Small deviations between pulses can be observed with values never reaching a 10% deviation.

Fig. 9

RMSD values of ten consecutive pulses to the average pulse.

Finally, repeatability tests were carried out with subject IV, in five consecutive days, within the same data acquisition period. The preliminary repeatability tests are presented in Fig. 10. Each graphic represents eight periods of the accessed waveforms. It can be seen that a low noise pulse wave was detected with a clear distinction of the identification points of the CPW. As previously done, the average pulse for each day was calculated [Fig. 11(a)] as well as the corresponding average pulse for the five days, presenting a value of $\mathrm{AIx}=-0.19\pm 0.05$ for the week.¹³ As expected, this value is close to the value assessed earlier for the same subject (${\mathrm{AIx}}_{\mathrm{IV}}=-0.13\pm 0.05$), showing a good repeatability with a maximum RMSD of 7% [Fig. 11(b)].

Fig. 10

Acquired data in 5 consecutive days for the subject IV (HR—heart rate; bpm—beats per minute).

Fig. 11

(a) Average pulse of five consecutive days and the identifications points determination through the 1st derivative; (b) RMSD values of the average pulse of each day (dots) to the average pulse of the week (solid line).