Open Access
Add to BibSonomy
Abstract
Corneal biomechanics play a critical role in maintaining corneal shape and thereby directly influence visual acuity. However, direct corneal biomechanical measurement in-vivo with sufficient accuracy and a high spatial resolution remains an open need. Here, we developed a three-dimensional (3D) motion-tracking Brillouin microscope for in-vivo corneal biomechanics mapping. The axial tracking utilized optical coherence tomography, which provided a tracking accuracy better than 3 µm. Meanwhile, 10 µm lateral tracking was achieved by tracking pupils with digital image processing. The 3D tracking enabled reconstruction of depth-dependent Brillouin distribution with a high spatial resolution. This superior technical performance enabled the capture of high-quality mechanical mapping in vivo even while the subject was breathing normally. Importantly, we improved Brillouin spectral measurements to achieve relative accuracy better than 0.07% verified by rubidium absorption frequencies, with 0.12% stability over 2000 seconds. These specifications finally yield the Brillouin measurement sensitivity that is required to detect ophthalmology-relevant corneal biomechanical properties.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Providing approximately 70% refractive power of the human eye, the cornea serves the primary role to clearly focus light on the retina [1]. As corneal refractive power derives from its delicate aspherical profile, to maintain corneal shape under ocular forces, such as in-plane tension and intraocular pressure (IOP) [2], the cornea has evolved to an intricate lattice of hundreds of lamellas intertwined by collagen fibers and proteoglycans to provide sufficient mechanical strength [3,4]. Deterioration of corneal biomechanics is the root cause of corneal ectatic disorders such as keratoconus (KC) [5] and postoperative ectasia after laser vision correction procedures (e.g., laser in situ keratomileusis (LASIK)) [6,7]; in these conditions, the mechanical balance between intraocular forces and corneal resistance is disrupted, resulting in corneal thinning and warpage, which causes dramatic alterations in corneal refractive power. Therefore, in an ideal situation, corneal biomechanical properties would function as the primary clinical data to evaluate reduced corneal elastic modulus, i.e., the measure of tissue’s resistance to deformation under an applied force, and thus identify patients at risk of developing ectasia before morphologic parameters such as thickness and curvature are altered [8]. Despite the critical status of corneal biomechanics in clinical practice, in vivo biomechanical measurements have proven difficult to obtain and have demonstrated significant limitations in early detection of decreased corneal modulus.
There are two commercial devices available to measure some aspects of corneal biomechanics currently: the ocular response analyzer (ORA, Reichert Technologies Inc., USA) and the corneal visualization Scheimpflug technology (Corvis ST, OCULUS Optikgeräte GmbH, Germany). Using air puff as stimuli and monitoring corneal deformation with an electro-optical collimation detector or an ultra-high-speed camera, the ORA and the Corvis ST can produce a series of viscoelastic parameters [9,10]. Evaluating these viscoelastic parameters has some utility in the early detection of keratoconus and ectasia after LASIK [11–13]. However, numerous studies have shown the benefit of these devices to be quite limited currently [14–16]. The accuracy of the extracted mechanical parameters is limited because: 1) the parameters measured are impacted by the whole eye rather than being specific to the cornea; 2) corneal response to external stimuli is not only determined by its inherent mechanics but also dependent on corneal geometry and boundary conditions, such as mechanical fixation, IOP, and corneal thickness [17]; 3) most importantly, 3D biomechanical distribution cannot be accessed directly through these techniques due to limited spatial resolution [18]. To improve resolution, a promising detecting method, optical coherence elastography (OCE), is under development [19,20]. Benefiting from the high resolution of optical microscopy, OCE can distinguish depth-dependent corneal modulus variation [21–23]. However, the accuracy is still hindered by boundary conditions because OCE is also based on the stimuli-response principle. Further, decoupling corneal deformation from patient movement when applying OCE in-vivo remains challenging.
Unlike current methods, Brillouin microscopy can probe the longitudinal modulus locally, with three-dimensional, micron-scale resolution, and without tissue deformation [24,25]. Brillouin microscopy can thus provide a 3D map of corneal modulus [26,27], which makes it highly promising for in-vivo clinical applications [28–30]. However, Brillouin ophthalmic instruments have been limited in both spatial and mechanical accuracy [30]. Spatially, as 10-20 seconds are required per axial scan at a single point on the cornea, the measurement is prone to significant artifacts due to patient’s eye movement. Mechanically, the measurement relies on calibration using materials of known Brillouin properties, which can be impacted by environmental conditions [25,30].
Here, we designed and implemented a motion-tracking Brillouin instrument with superior mechanical and spatial accuracy by three-fold improvement in measurement speed, leveraging 3D motion-tracking and absolute calibration to atomic absorption lines. We demonstrate that these improvements finally enable mapping patients’ corneas in-vivo with sufficient spatial and mechanical accuracy to reveal biomechanical distribution regardless of patient moving patterns.
2. Materials and methods
In this section, we first describe the strategy of combining tracking methods with a Brillouin microscope. The specifications of the setup were exhibited via several ex-vivo tests using enucleated porcine eyes. The porcine eyes were collected from a local slaughterhouse and were used within 12 hours after collection. Next, operation rules of the setup in clinic were introduced. The recruited patient signed informed consent forms and the study was approved by the Cleveland Clinic Institutional Review Board. The overall optical power in the in-vivo measurement was controlled to meet the requirements from American National Standard Institutes (ANSI Z136.1-2007).
2.1 Brillouin longitudinal modulus
Spontaneous Brillouin scattering arises from the interaction between incident light and acoustic phonons generated by inherent density/pressure fluctuation, which results in a frequency shift of the scattered light [31]. The relation between the Brillouin shift ΔνB and the longitudinal modulus M’ can be expressed as
Longitudinal modulus determined by Brillouin microscopy is fundamentally different from the widely used Young’s modulus, and a universal conversion from the longitudinal modulus to the Young’s modulus has not been established because of frequency-dependent modulus properties [33] and near incompressibility of biological tissues [34,35]. However, for specific tissues, including the cornea, a strong correlation between the two moduli has been observed [36]. Thus, we can interpret increase (or decrease) of the longitudinal modulus as an indication of increase (or decrease) in Young’s modulus. In the following, we will report the Brillouin shift, which is the quantity that is directly measured experimentally. To evaluate any uncertainty metric of the measurement in terms of longitudinal modulus, a derivative of Eq. (1) can be computed and it shows that dM’/M’=2dΔνB/ΔνB, where dM’ is the variation of longitudinal modulus measurement, and dΔνB is the variation of Brillouin shift measurement.
2.2 Experimental setup
The motion-tracking Brillouin microscope consists of a Brillouin spectrometer for corneal modulus measurement, an optical coherence tomography (OCT) device for axial tracking and a single-lens imaging system for in-plane pupil tracking (Fig. 1). The components in the dashed square were mounted on a breadboard sitting on an ophthalmic slit lamp frame, serving as the human interface, which could be moved in three dimensions via a joystick.
Fig. 1. Schematic diagram of the in-vivo setup. SLD: superluminescent diode; PC: polarization controller; ND: neutral density filter; DM: dichroic mirror; QWP: quarter wave plate; HWP: half wave plate; PBS: polarizing beam splitter; P: polarizer; L1-L7: lenses.
2.2.1 Brillouin microscope
To provide spatially-resolved Brillouin shifts across the cornea, a Brillouin microscope was built. This was comprised of a Brillouin spectrometer and a confocal microscope. The beam paths of this instrument are drawn in red. The light source was a continuous wave laser (DL pro, Toptica) centered at 780 nm and locked to a Rb absorption line. The locking not only stabilized the laser frequency but also allowed to reduce reflection by adding an extra Rb gas cell before coupling scattered light into the spectrometer through objective lens L7. For better frequency performance, besides the locking, two Bragg filters (BP-780-90, OptiGrate) were placed after the laser head to suppress amplified spontaneous emission (ASE). After filtering, the free space laser was coupled into a polarization-maintaining fiber and directed to the human interface. On the human interface, the output beam from the fiber collimator (TC12APC-780, Thorlabs) was enlarged by a telescope made up by two lenses L1 (f = 20 mm) and L2 (f = 75 mm) to get an effective 0.1 numeric aperture (NA) on lens L3 (f = 50 mm). L3 was mounted on a translational stage (T-LSM025A, Zaber) to fulfill axial scan through the corneal depth. The power after L3 was 7 mW.
To detect the frequency shift ΔνB, the scattered light was first redirected by the PBS to go through the Rb gas cell to remove the unshifted frequency. Then, an objective lens L7 (PLN10X, Olympus) was used to couple the light into a fiber connected to a Brillouin spectrometer. In the Brillouin spectrometer, two 15 GHz virtually imaged phase array (VIPA) etalons were used as the core dispersion components to distinguish Brillouin shifts from the excitation laser frequency [37,38]. As the VIPA played a similar role as a grating, a configuration conceptually similar to blazing was adopted (input beam diameter = 1.31 mm, f = 200 mm lenses before and after the VIPA) to maximize the intensity of Brillouin scattering according to the dispersion law of a VIPA [39]. The exposure time of the spectrometer was set to 0.05 s.
2.2.2 Optical coherence tomography (OCT)
To track axial movement of the measured point, the sampling arm of a frequency-domain OCT shared the same optical path with the 780 nm laser through a dichroic mirror, as shown in Fig. 1. The light source of the OCT was a superluminescent diode (SLD-mCS-371-HP1-SM, Superlum) with a bandwidth of 50 nm and a central wavelength of 840 nm. The corresponding axial resolution was 6.2 µm. The power of the SLD was equally split into the reference and sampling arms by a fiber coupler. The collimated beam diameters at the outputs of the fiber coupler were 1.31 mm. As the focal lengths of L3 and L4 were 50 mm, the transvers resolution was 40.8 µm and the depth of focus (DOF) was 3.1 mm. After focusing 1.5 mW OCT light onto the cornea, the scattered light was interfered with the reflection from the reference arm and analyzed by a custom-made spectrometer. Within the spectrometer, incident light was diffracted by a volume phase holographic grating (1800 lp/mm, Wasatch Photonics) and imaged onto a line camera (raL4096-24gm, Basler) through an achromatic lens (f = 100 mm). Taking the 0.05 s Brillouin exposure time into account, the exposure time of the OCT was set to 0.5 ms to reduce data amount, even though the OCT could work much faster than that. The sensitivity at this exposure time was 116 dB, and the 10 dB sensitivity roll-off distance was 2 mm.
2.2.3 En Face imaging
Besides potential axial movement during an A-scan, patients’ eyes could also exhibit lateral movement. To monitor this movement, an imaging system was placed next to the objective lens, as shown in Fig. 1, to perform pupil tracking. The light source for this imaging system was an LED working at 970 nm (M970L4, Thorlabs). The illumination power on the whole eye was 2 mW and the exposure time of the imaging camera (SCE-B013-U, Mightex) was 20 ms. A long-pass filter (FEL0850, Thorlabs) was inserted before the camera to reject light from the 780 nm laser totally and keep part of the light from the SLD to annotate the measured points. The ratio of distance/pixel was calculated by imaging a standard grid (R1L3S3P, Thorlabs) on the focal plane. The pupil tracking algorithm mainly contained binarization and edge detection.
2.3 Instrument specifications
As the setup was designed to measure corneal Brillouin shifts and concurrent eye movement, key specifications were frequency performance of the Brillouin spectrometer and tracking accuracy. To understand these specifications, these two parts were tested separately.
2.3.1 Brillouin spectrometer calibration
Traditionally, a VIPA-based Brillouin spectrometer uses two different materials with known Brillouin shifts and linear fitting to estimate frequencies along the dispersion axis [25,30]. After getting the ratio of GHz/pixel, the Brillouin shift of an unknown material can be calculated by counting pixels between its Stokes and Anti-Stokes. The accuracy of this calibration strongly relies on the accuracy of the known Brillouin shifts, but the Brillouin shifts of the standard materials vary over time because of the change of the room temperature. For example, the Brillouin shift of the first standard material, water, changes 7.4 MHz/°C, while the Brillouin shift of the second standard material, polystyrene, suffers a nonlinear decrease when the temperature increases [30]. To overcome this drawback and improve calibration accuracy, in our setup we rely on atomic gas vapor (Rubidium), which offers highly conserved frequency standard at its narrow absorption lines. This enabled us to use three locked laser frequencies (ν1 = 384229.1497 GHz, ν2=ν1-1.1678 GHz, ν3=ν1 + 2.9607 GHz) to excite the Brillouin signal of water. Also, polynomial fitting was substituted for the linear fitting to calculate frequencies along the dispersion axis. After the calculation, the water was excited by another two locked laser frequencies (ν4=ν1 + 31.7 MHz, ν5=ν1 + 92.0 MHz) to verify the accuracy. Water was selected as the calibration sample based on the following two reasons. First, water occupies 78% of the cornea by volume [40,41], so that it has a Brillouin shift in the same spectral region as the cornea. Second, the speed of sound in water under different temperature has been examined thoroughly [42,43].
The raw Brillouin spectra of water excited by ν1, ν2 and ν3 were combined and shown in Fig. 2(a). Stokes (S) and Anti-Stokes (AS) of ν4 and ν5 are not shown in Fig. 2(a) because they are too close to the Brillouin signals excited by ν1. For each excitation frequency, 100 spectra were acquired for average. Pixel positions of these frequency components were extracted using Lorentzian-envelop fitting. The relation between the pixel position and the frequency shift is shown in Fig. 2(b). The frequency of S(ν3) was set to be the largest because the dispersion order of Stokes was higher. The excitation laser frequencies for Stokes were on the left of S(ν3), while those for Anti-Stokes were on the right of AS(ν2).
Fig. 2. Spectrometer frequency calibration. (a) Stokes and Anti-Stokes of water excited by ν1, ν2, and ν3. (b) Relation between the pixel position and the Brillouin shift. The blue dots were Brillouin shifts of ν1, ν2, and ν3, used to fit the frequency distribution, while the red squares were Brillouin shifts of ν4 and ν5, used to evaluate the fitting accuracy.
The relation between the frequency, ν, and the pixel position, x, was assumed as
where ai was the coefficient. The origin of x was set to be the first pixel on the left of Fig. 2(a). According to the known frequency differences, Δν1-2 = 1.1678 GHz and Δν1-3 = 2.9607 GHz, the coefficients a1 to a4 were calculated through the positions of the Stokes and Anti-Stokes.In theory, the difference between S(ν3) and S(ν1) does not equal Δν1-3 precisely because of the change of excitation frequency, according to Eq. (1). As the speed of sound is much smaller than that of light, the error introduced by this approximation was limited to tens of kHz, which could be ignored. To determine a0, the Brillouin shift of water with known temperature was assigned to AS(ν1) because ν1 would be the excitation frequency in the in-vivo test. According to Eq. (2), a0 could be calculated once a1 to a4 were determined. Polynomial fitting could not be used directly here to find the coefficients because the free spectral range (FSR) of the spectrometer was unknown so that the frequency difference between S(ν2) and AS(ν3) could not be achieved. After getting coefficients a0 to a4, the Brillouin shift of an unknown material, ΔνB_unknown, was calculated through
To verify the accuracy of the calculated coefficients in Eq. (2), the laser was locked to ν4 and ν5 sequentially to generate the test frequencies, marked as the red squares in Fig. 2(b). As ν4=ν1 + 31.7 MHz and ν5=ν1 + 92.0 MHz, comparing the calculated results from Eq. (2) with the well-defined 31.7 MHz and 92.0 MHz showed that the errors and the standard deviations of the Stokes and Anti-Stokes excited by ν4 were 3.04 ± 6.51 MHz and 3.50 ± 6.37 MHz, while those excited by ν5 were 0.31 ± 7.02 MHz and 4.12 ± 6.61 MHz. As the frequency accuracy only related to fitting accuracy, change of the Brillouin shift did not influence this accuracy. Therefore, taking the largest error of 4.12 MHz into account, the relative accuracy in the 5.7 GHz region, typical of cornea values [30], was better than 0.07%.
After demonstrating the accuracy, the stability of the spectrometer was optimized and tested. Even though the laser frequency was locked to Rb, mechanical drift inside the Brillouin spectrometer could also lead to instability. In the short-term, the spectrometer can be assumed to be stable. Thus, to evaluate the long-term stability, standard deviations of a 10-second and a 2000-second running were compared. As shown in Fig. 3, the standard deviation of the Brillouin shift of water maintained 6.9 MHz (0.12% relatively) during the 2000-second test, which indicated that the Brillouin spectrometer remained stable during this period.
Fig. 3. Measurement of short-term and long-term stability in Brillouin shift measurements. (a) Continuous acquisition of Brillouin shift of water for 10 s with an exposure time of 0.05 s. (b) Brillouin shift of water for 2000 s with an exposure time of 0.05 s and an acquisition interval of 1 s.
2.3.2 3D tracking accuracy
The tracking accuracy was tested using porcine eyes to simulate clinical imaging scenarios. To test the accuracy of axial tracking in clinic, the OCT light was focused on the corneal edge, avoiding the central reflection region to study the tracking performance in a scattering region. A porcine eye was mounted on a translational stage (T-LSM025A, Zaber) to move along 1 mm with a step size of 100 µm. At each place, 100 frames were acquired for averaging. The measured position of the porcine eye was determined by the position of the reflection peak from its anterior surface, as shown in Fig. 4(a). In Fig. 4(b), the errors and the standard deviations of the distances measured by OCT were plotted along with the known distances read from the translational stage. Within the 1 mm moving range, the maximum error was about 3 µm. Even though the frequency domain OCT could reach an accuracy of tens of nanometers when using mirrors as targets on both arms, 3 µm was achieved because the porcine eye had a much lower scattering ratio, which was close to clinical scenarios.
Fig. 4. Axial tracking test using a porcine eye. (a) Distance determination by measuring the position of the reflection peak from the anterior corneal surface. (b) Measurement errors and standard deviations by comparing with the known displacements read from the translational stage.
Like the test of axial accuracy, the test of lateral accuracy was also based on tracking the movement of a porcine eye, thereby imitating an in-vivo situation. The porcine eye was mounted on a translational stage (T-LSM025A, Zaber), which was moved horizontally. The moving range was 9 mm (from -4.5 mm to 4.5 mm) with a step size of 0.5 mm. At each position, 100 frames were acquired for averaging. Measured distances were calculated by counting pixel changes of the center of the pupil and multiple them with the ratio of distance/pixel. The error and standard deviation at each position was plot in Fig. 5. As shown in this figure, the error between tracking algorithm and known displacements were within 10 µm.
Fig. 5. Measurement errors and standard deviations of the lateral tracking test using a porcine eye, compared with the known displacements read from the translational stage.
2.4 Brillouin clinical instrument
After individual tests of the Brillouin spectrometer and the 3D tracking, the Brillouin microscope was configured for in-vivo measurements. A picture of the setup was shown in Fig. 6(a). To operate in clinical settings, the in-vivo Brillouin microscope features a human interface adapted from an ophthalmic slit lamp frame, enabling human-operated motion in three dimensions via a joystick. During the measurement, the patient was asked to sit in front of the human interface, rest the head on a chinrest and maintain their position by biting a fixed bite bar. The bite bar was enclosed by an inside-out individual-use plastic bag for sanitation. A green LED bead was placed right above the objective lens, as shown in Fig. 6(b), which served as patient’s fixation target.
Fig. 6. Operating the motion-tracking Brillouin microscope in clinic. (a) The human interface and patient fixation mount. (b) Close-up shot of the setup when running a scan on a volunteer. (c) Software panel of the setup.
To operate all functional parts simultaneously, a software interface was designed in LABVIEW, as shown in Fig. 6(c). Synchronization among different parts was fulfilled by using a universal clock. In the pupil tracking window, the small white dot near the center of the pupil identified the location of the laser dot, while the large white dot at the edge of the pupil was the reflected image of the 970 nm lamp. The diameters of the concentric red circles, d, equaled 2 mm, 5mm, 7 mm, 8 mm and 9 mm. When picking a point of interest to measure, the operator moved the cursor to the laser dot and clicked. Then, an A-scan would start. After each scan, the measured point was marked as a green dot in the pupil tracking window. Even though the thickness of human cornea is around 500∼600 µm, the scanning range was set to 1.8 mm with a step size of 15 µm to account for unpredictable patient movement. The start point of a scan was determined by the OCT. During each scan, the acquired Brillouin spectra were fitted using the Lorentzian function to calculate Brillouin shifts. The Brillouin profile window showed these roughly calculated Brillouin shifts along the depth without position correction to give an idea of the scan quality. Fine data processing was conducted afterwards.
For an overall analysis of the instrument performance, in Table 1 we listed the specifications obtained while operating the instrument with parameters for in vivo measurements. The optimized configuration in the Brillouin spectrometer enabled a much shorter exposure time than the traditional (0.05 s vs 0.2 s) with similar frequency stability and only slight increase of laser power (7 mW vs 5 mW), which was still well below safety limits [30].
Table 1. Specifications of motion-tracking Brillouin microscope
3. Results
To demonstrate the in-vivo performance of the setup, in this section, we first show the ability of motion artifact correction by measuring the Brillouin depth profile at the center of the pupil under different motion patterns. Brillouin maps of a normal cornea were then generated.
3.1 Brillouin depth profiles free of motion artifacts
To verify the tracking performance and use tracking information to compensate for motion errors of measured Brillouin shifts, the center of the pupil, which was determined by the pupil tracking, was scanned twice under different breath conditions. The patient was asked to hold his breath during the first scan and breathe regularly during the second. During each scan, Brillouin shifts, axial, and lateral movements were recorded simultaneously.
In terms of lateral motion, the moving pattern when breathing regularly was shown in Fig. 7(a), while the moving pattern when holding breath was shown in Fig. 7(b). From these two figures, no obvious difference could be observed. The in-plane eye movement remained within a range of approximately 40 µm during the scan regardless of the breathing condition. Taking the 40.8 µm transverse resolution of the OCT, the eye in-plane movement could be ignored, showing the effectiveness of the green LED fixation target.
Fig. 7. Brillouin correction by tracking information under different breathing conditions. (a) Eye in-plane movement along with time when the patient breathed regularly. Movements in x represent moving left (negative values) or right (positive values). Movements in y represent moving up (negative values) or down (positive values). (b) Eye in-plane movement along with time when the patient held his breath. (c) Corneal axial movement along with time when the patient breathed regularly. (d) Corneal axial movement along with time when the patient held his breath. (e) Correction of the Brillouin profile measured during regular breath by being paired with coordinates measured by the OCT. The red shadow represents the cornea, while the rest of the curve is the aqueous. (f) Correction of the Brillouin profile measured during breath holding by being paired with coordinates measured by the OCT.
Unlike the stable patterns of in-plane movement, the axial movement showed a clear difference under different breathing conditions, as shown in Fig. 7(c) and (d). When the patient breathed regularly, as in Fig. 7(c), derivative oscillation was superimposed on a continuous movement away from the objective lens. The total drift was about 200 µm, modulated by a period of 2-3 seconds. In contrast, when the patient held his breath, as demonstrated in Fig. 7(d), only a 200 µm continuous drift was observed.
Even though the motion pattern was influenced by breathing, the axial tracking indicated that the patient tended to move hundreds of micrometers during an A-scan. Because of 500-600 µm corneal thickness, this movement led to distortion of the measured Brillouin distribution from the cornea to the aqueous humor. To correct the distortion and recover the Brillouin profile, Brillouin shifts were paired with their axial coordinates measured by the OCT. By reaccommodating the Brillouin shifts according to their axial coordinates, the corrected Brillouin profiles showed smooth transition from the cornea to the aqueous, as shown in Fig. 7(e) and (f). The cornea is marked with red shadow, while the aqueous humor is the tail part with a Brillouin shift of 5.24 GHz. In the situation of breath-holding, Fig. 7(f), corneal thickness was corrected to around 500 µm, instead of the original 700 µm. For the more complicated motion pattern caused by regular breathing, the corrected Brillouin profile in Fig. 7(e) shared a same shape as the corrected one in Fig. 7(f), indicating the efficacy of tracking-based correction. Moreover, the same Brillouin shift of the aqueous humor in different breathing cases served as the last calibration check, proving the feasibility of using the aqueous humor as a baseline.
3.2 Brillouin map
After validating the efficacy of 3D tracking in Brillouin profile correction, the in-vivo setup was used to plot a map of modulus distribution across the cornea. The right eye from a patient with a normal cornea was measured. As the Brillouin shift is also sensitive to the temperature [30], the room temperature was kept at 23 °C. During the exam, 29 A-scans were performed throughout the whole cornea. Currently, 2D geometrical maps are mainly used in ophthalmological diagnosis, so 2D Brillouin maps were created to resemble clinical maps. To do so, we took advantage of the relatively flat Brillouin shifts in the anterior corneal region, as shown in Fig. 7(e) and (f); therefore, the mean value of the plateau part of a Brillouin profile was used to represent the Brillouin shift at that measured point. The end point of the plateau was determined by finding the intersection of the flat and the steep slopes using linear fitting. To connect scattered measured points, 2D interpolation was used. This plateau averaging strategy was feasible because in-plane eye movement was proved to be within 40 µm, which was smaller than the transverse resolution of the OCT. If in-plane eye movement was larger, the results could not be marked as 2D points on the cornea. Instead, 3D map would be required. When conducting plateau averaging in a recovered Brillouin profile, it appears that reaccommodating was not necessary in some cases, such as the case shown in Fig. 7(f), because the values in the plateau region were the same without reaccommodating. However, without correction, unpredictable irregular Brillouin profiles would make it difficult to select the plateau region.
The Brillouin map of the normal cornea is shown in Fig. 8(a) whose coordinate origin was set at the center of the pupil. In this map there is a softer region within the d = 4 mm circle, where Brillouin shifts were about 5.72 GHz. The periphery outside the d = 4 mm circle shows a larger Brillouin shift, varying from 5.74 to 5.76 GHz. The superior part was about 0.02 GHz stiffer than the inferior. The thickness and curvature maps measured by a commercial instrument (Pentacam HR, Oculus Inc) are shown in Fig. 8(b) and (c) as reference. To further investigate the change of Brillouin shifts along the depth and have a better understanding of the variance of modulus across the cornea, two representative Brillouin profiles at the center (X = -0.40 mm, Y = -0.03 mm) and the periphery (X = -1.80 mm, Y = 2.65 mm) were plotted in Fig. 8(d) after correction by tracking information. The positive X coordinate means the point is on the right of the origin, while the negative Y coordinate means the point is on top of the origin. From Fig. 8(d), it can be seen that the corrected Brillouin profiles had similar thickness as that measured by the commercial instrument. The central cornea had lower modulus than the periphery as expected [29]. The modulus decreased from the anterior to the posterior could be observed in both profiles.
Fig. 8. Motion-tracking Brillouin microscope results of the normal cornea. (a) Brillouin map of the normal cornea. The color bar represents Brillouin shifts in GHz. Coordinates on the X and Y axes annotate diameters of a series of concentric circles. (b) Thickness map of the cornea measured by a commercial instrument. (c) Curvature map of the cornea measured by a commercial instrument. (d) Representative Brillouin profiles at the center and the periphery.
4. Discussion
Our next-generation motion-tracking Brillouin ophthalmic instrument shows promising results from patients in the clinic for mapping focal, depth-dependent biomechanical properties of the cornea. Compared with existing clinical techniques, Brillouin microscopy provides a non-contact option to measure modulus without application of an external force and with high spatial resolution rather than global averaging. Given the expectation that relevant corneal abnormalities are focal in nature in their earliest manifestations, this feature makes Brillouin microscopy a leading candidate for corneal biomechanical measurement. When bringing Brillouin microscopy into clinical use, patients’ comfort is an important factor and its improvement will translate in more reliable and efficient measurements than previous generation instruments. The measurement time was shortened three-fold by adopting an optimal VIPA configuration. The acquisition time of a Brillouin spectrum was decreased from 0.2 s to 0.05 s, allowing an A-scan to be completed within 5 s, which is much faster than the previous 10-20 s and well-tolerated by patients. For a Brillouin map built from 30-40 A-scans, the total measurement time was less than 20 min. Besides the faster acquisition, 3D tracking also allowed patients to blink or even sit back to relax between adjacent A-scans.
In addition to decreased acquisition time and improved patient comfort, this new generation Brillouin microscope also achieved much improved spatial accuracy. Even though each A-scan only took 5 s, in-vivo tests showed that inevitable patient movement still introduced an error of hundreds of micrometers, which was comparable to the corneal thickness. To overcome this motion artifact, precise 3D tracking was performed along with Brillouin detection. The 3 µm accuracy axial tracking and 10 µm accuracy lateral tracking guaranteed successful reconstruction of disordered Brillouin shifts. Another benefit of combining OCT with Brillouin was taking advantage of the thickness measured by the OCT to determine the corneal region from a prolonged scanning range. Even though only the cornea and the following aqueous humor was shown in our depth-dependent results, the actual scanning range was 1.8 mm, much longer than the shown range, thereby leaving space for patient movement. Thus, scans were started before the focal point reaching the cornea. With the help of the thickness measured by the OCT, the starting point of the cornea was determined by stepping back from the end of the clear posterior corneal edge with a step size of the thickness. Without this OCT-guided approach, noisy Brillouin shifts could be included in plateau averaging or meaningful Brillouin shifts could be eliminated.
Another important improvement is the frequency accuracy of the Brillouin spectrometer. The use of standard Rb absorption frequencies brought certified MHz-level accuracy. Tests showed that the absolute accuracy was better than 4.12 MHz, corresponding to a relative accuracy of 0.07% in the 5.7 GHz corneal region. If converting it to the longitudinal modulus, the resulting accuracy was better than 0.14%. Also, the stability results proved that the system could maintain this accuracy over 2000 s, longer than the measurement for each patient. The improved accuracy and stability should allow for the identification of MHz level changes in post-LASIK or subtle keratoconus, so that corneal biomechanical alterations can be revealed at early stages.
Even though 3 times shorter measurement time was achieved by optimizing VIPA configuration, line-scan could further facilitate corneal scanning because axial scan will be omitted [44]. If a depth-dependent Brillouin profile can be acquired in a single shot. The difficulty in fulfilling this idea may arise from weaker Brillouin scattering because of the use of a less focused beam. To maintain the same acquisition rate, a stronger spectral extinction ratio will be needed, which can be carried out by suppressing laser noise, adding Fabry-Perot etalons in the spectrometer and cooling down the spectral camera to a lower temperature. Besides the improvement of the setup, we will also need to enlarge our recruitment to fully validate our approach in clinical settings; but the Brillouin maps of the cornea already show encouraging and unprecedented sensitivity of our new instrument to clinically-relevant mechanical properties: in the normal cornea, the Brillouin map show a softer center and stiffer periphery, coinciding with the collagen map acquired through ex-vivo X-ray detection [45,46].
Finally, even though here we have presented the Brillouin map as two-dimensional by performing an average of the anterior plateau of corneal Brillouin depth profiles, this was merely a convenient choice to compare to pachymetry and topography maps and provide a quick visual impression of Brillouin-based biomechanical distribution. However, our data contain reliable 3D information which we expect will allow to extract additional data in the future, such as strength decrease from the anterior to the posterior and depth-dependent change of the same cornea before and after surgeries.
5. Conclusion
In conclusion, we combined tracking strategies with a conventional Brillouin microscope to build an in vivo motion-tracking Brillouin microscope, thereby correcting disordered Brillouin profiles caused by patient motion artifact. The efficiency and accuracy of the Brillouin spectrometer were also upgraded by selecting appropriate parameters and using well-defined frequencies for calibration. These improvements made the setup successful in mapping corneal modulus and providing values to evaluate corneal modulus change. This innovative technique opens new perspectives of Brillouin microscopy in in-vivo applications.
Funding
National Eye Institute (R01EY028666, R01EY032537); Research to Prevent Blindness.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data for the characterization of the instrument are available from the corresponding author upon request. Data underlying patient results presented in this paper are not publicly available at this time because of privacy polices but may be obtained from the authors upon reasonable request.
References
1. R. A. Gordon and P. B. Donzis, “Refractive development of the human eye,” Arch. Ophthalmol. (Chicago, IL, U. S.) 103(6), 785–789 (1985). [CrossRef]
2. W. J. Dupps Jr. and S. E. Wilson, “Biomechanics and wound healing in the cornea,” Exp. Eye Res. 83(4), 709–720 (2006). [CrossRef]
3. K. M. Meek and C. Knupp, “Corneal structure and transparency,” Prog. Retinal Eye Res. 49, 1–16 (2015). [CrossRef]
4. C. S. Kamma-Lorger, C. Boote, S. Hayes, J. Moger, M. Burghammer, C. Knupp, A. J. Quantock, T. Sorensen, E. Di Cola, N. White, R. D. Young, and K. M. Meek, “Collagen and mature elastic fibre organisation as a function of depth in the human cornea and limbus,” J. Struct. Biol. 169(3), 424–430 (2010). [CrossRef]
5. Y. S. Rabinowitz, “Keratoconus,” Surv. Ophthalmol. 42(4), 297–319 (1998). [CrossRef]
6. J. B. Randleman, B. Russell, M. A. Ward, K. P. Thompson, and R. D. Stulting, “Risk factors and prognosis for corneal ectasia after LASIK,” Ophthalmology 110(2), 267–275 (2003). [CrossRef]
7. J. B. Randleman, M. Woodward, M. J. Lynn, and R. D. Stulting, “Risk assessment for ectasia after corneal refractive surgery,” Ophthalmology 115(1), 37–50.e4 (2008). [CrossRef]
8. M. J. Girard, W. J. Dupps, M. Baskaran, G. Scarcelli, S. H. Yun, H. A. Quigley, I. A. Sigal, and N. G. Strouthidis, “Translating ocular biomechanics into clinical practice: current state and future prospects,” Curr. Eye Res. 40(1), 1–18 (2015). [CrossRef]
9. D. A. Luce, “Determining in vivo biomechanical properties of the cornea with an ocular response analyzer,” J. Cataract Refractive Surg. 31(1), 156–162 (2005). [CrossRef]
10. R. Ambrósio Jr, I. Ramos, A. Luz, F. C. Faria, A. Steinmueller, M. Krug, M. W. Belin, and C. J. Roberts, “Dynamic ultra high speed Scheimpflug imaging for assessing corneal biomechanical properties,” Rev. bras.oftalmol. 72(2), 99–102 (2013). [CrossRef]
11. R. Vinciguerra, R. Ambrosio Jr., C. J. Roberts, C. Azzolini, and P. Vinciguerra, “Biomechanical characterization of subclinical keratoconus without topographic or tomographic abnormalities,” J Refract Surg 33(6), 399–407 (2017). [CrossRef]
12. P. Kataria, P. Padmanabhan, A. Gopalakrishnan, V. Padmanaban, S. Mahadik, and R. Ambrosio Jr., “Accuracy of Scheimpflug-derived corneal biomechanical and tomographic indices for detecting subclinical and mild keratectasia in a South Asian population,” J. Cataract Refractive Surg. 45(3), 328–336 (2019). [CrossRef]
13. V. S. De Stefano, I. Seven, J. B. Randleman, and W. J. Dupps Jr., “Custom air puff-derived biomechanical variables in a refractive surgery screening setting: Study from 2 centers,” J. Cataract Refractive Surg. 44(5), 589–595 (2018). [CrossRef]
14. G. Labiris, A. Giarmoukakis, Z. Gatzioufas, H. Sideroudi, V. Kozobolis, and B. Seitz, “Diagnostic capacity of the keratoconus match index and keratoconus match probability in subclinical keratoconus,” J. Cataract Refractive Surg. 40(6), 999–1005 (2014). [CrossRef]
15. Y. M. Wang, T. C. Y. Chan, M. Yu, and V. Jhanji, “Comparison of corneal dynamic and tomographic analysis in normal, forme fruste keratoconic, and keratoconic eyes,” J Refract Surg 33(9), 632–638 (2017). [CrossRef]
16. J. Steinberg, M. Siebert, T. Katz, A. Frings, J. Mehlan, V. Druchkiv, J. Buhren, and S. J. Linke, “Tomographic and biomechanical scheimpflug imaging for keratoconus characterization: a validation of current indices,” J Refract Surg 34(12), 840–847 (2018). [CrossRef]
17. S. Kling and S. Marcos, “Contributing factors to corneal deformation in air puff measurements,” Invest. Ophthalmol. Visual Sci. 54(7), 5078–5085 (2013). [CrossRef]
18. V. S. De Stefano and W. J. Dupps Jr., “Biomechanical diagnostics of the cornea,” Int Ophthalmol Clin 57(3), 75–86 (2017). [CrossRef]
19. S. Wang and K. V. Larin, “Optical coherence elastography for tissue characterization: a review,” J. Biophoton 8(4), 279–302 (2015). [CrossRef]
20. M. A. Kirby, I. Pelivanov, S. Song, L. Ambrozinski, S. J. Yoon, L. Gao, D. Li, T. T. Shen, R. K. Wang, and M. O’Donnell, “Optical coherence elastography in ophthalmology,” J. Biomed. Opt. 22(12), 1–28 (2017). [CrossRef]
21. S. Wang and K. V. Larin, “Noncontact depth-resolved micro-scale optical coherence elastography of the cornea,” Biomed. Opt. Express 5(11), 3807–3821 (2014). [CrossRef]
22. V. S. De Stefano, M. R. Ford, I. Seven, and W. J. Dupps Jr., “Depth-dependent corneal biomechanical properties in normal and keratoconic subjects by optical coherence elastography,” Trans. Vis. Sci. Tech. 9(7), 4 (2020). [CrossRef]
23. A. Ramier, A. M. Eltony, Y. Chen, F. Clouser, J. S. Birkenfeld, A. Watts, and S. H. Yun, “In vivo measurement of shear modulus of the human cornea using optical coherence elastography,” Sci. Rep. 10(1), 17366 (2020). [CrossRef]
24. G. Scarcelli and S. H. Yun, “Confocal Brillouin microscopy for three-dimensional mechanical imaging,” Nat. Photonics 2(1), 39–43 (2008). [CrossRef]
25. G. Scarcelli, W. J. Polacheck, H. T. Nia, K. Patel, A. J. Grodzinsky, R. D. Kamm, and S. H. Yun, “Noncontact three-dimensional mapping of intracellular hydromechanical properties by Brillouin microscopy,” Nat. Methods 12(12), 1132–1134 (2015). [CrossRef]
26. G. Scarcelli, R. Pineda, and S. H. Yun, “Brillouin optical microscopy for corneal biomechanics,” Invest. Ophthalmol. Vis. Sci. 53(1), 185–190 (2012). [CrossRef]
27. G. Scarcelli, S. Kling, E. Quijano, R. Pineda, S. Marcos, and S. H. Yun, “Brillouin microscopy of collagen crosslinking: noncontact depth-dependent analysis of corneal elastic modulus,” Invest. Ophthalmol. Vis. Sci. 54(2), 1418–1425 (2013). [CrossRef]
28. G. Scarcelli and S. H. Yun, “In vivo Brillouin optical microscopy of the human eye,” Opt. Express 20(8), 9197–9202 (2012). [CrossRef]
29. G. Scarcelli, S. Besner, R. Pineda, P. Kalout, and S. H. Yun, “In vivo biomechanical mapping of normal and keratoconus corneas,” JAMA Ophthalmol 133(4), 480–482 (2015). [CrossRef]
30. P. Shao, A. M. Eltony, T. G. Seiler, B. Tavakol, R. Pineda, T. Koller, T. Seiler, and S. H. Yun, “Spatially-resolved Brillouin spectroscopy reveals biomechanical abnormalities in mild to advanced keratoconus in vivo,” Sci. Rep. 9(1), 7467 (2019). [CrossRef]
31. J. M. Vaughan and J. T. Randall, “Brillouin scattering, density and elastic properties of the lens and cornea of the eye,” Nature 284(5755), 489–491 (1980). [CrossRef]
32. D. W. Leonard and K. M. Meek, “Refractive indices of the collagen fibrils and extrafibrillar material of the corneal stroma,” Biophys. J. 72(3), 1382–1387 (1997). [CrossRef]
33. R. Prevedel, A. Diz-Munoz, G. Ruocco, and G. Antonacci, “Brillouin microscopy: an emerging tool for mechanobiology,” Nat. Methods 16(10), 969–977 (2019). [CrossRef]
34. P. J. Wu, I. V. Kabakova, J. W. Ruberti, J. M. Sherwood, I. E. Dunlop, C. Paterson, P. Torok, and D. R. Overby, “Water content, not stiffness, dominates Brillouin spectroscopy measurements in hydrated materials,” Nat. Methods 15(8), 561–562 (2018). [CrossRef]
35. G. Scarcelli and S. H. Yun, “Reply to ‘Water content, not stiffness, dominates Brillouin spectroscopy measurements in hydrated materials’,” Nat. Methods 15(8), 562–563 (2018). [CrossRef]
36. J. N. Webb, J. P. Su, and G. Scarcelli, “Mechanical outcome of accelerated corneal crosslinking evaluated by Brillouin microscopy,” J. Cataract Refractive Surg. 43(11), 1458–1463 (2017). [CrossRef]
37. G. Scarcelli and S. H. Yun, “Multistage VIPA etalons for high-extinction parallel Brillouin spectroscopy,” Opt. Express 19(11), 10913–10922 (2011). [CrossRef]
38. J. Zhang and G. Scarcelli, “Mapping mechanical properties of biological materials via an add-on Brillouin module to confocal microscopes,” Nat Protoc 16(2), 1251–1275 (2021). [CrossRef]
39. S. Xiao, A. M. Weiner, and C. Lin, “A dispersion law for virtually imaged phased-array spectral dispersers based on paraxial wave theory,” IEEE J. Quantum Electron. 40(4), 420–426 (2004). [CrossRef]
40. Z. D. Taylor, J. Garritano, S. Sung, N. Bajwa, D. B. Bennett, B. Nowroozi, P. Tewari, J. Sayre, J. P. Hubschman, S. Deng, E. R. Brown, and W. S. Grundfest, “THz and mm-wave sensing of corneal tissue water content: electromagnetic modeling and analysis,” IEEE Trans. THz Sci. Technol. 5(2), 170–183 (2015). [CrossRef]
41. M. J. Doughty and M. L. Zaman, “Human corneal thickness and its impact on intraocular pressure measures: a review and meta-analysis approach,” Surv. Ophthalmol. 44(5), 367–408 (2000). [CrossRef]
42. M. Greenspan and C. Tschiegg, “Speed of sound in water by a direct method,” J. Res. Natl. Bur. Stan. 59(4), 249–254 (1957). [CrossRef]
43. W. Marczak, “Water as a standard in the measurements of speed of sound in liquids,” J. Acoust. Soc. Am. 102(5), 2776–2779 (1997). [CrossRef]
44. J. Zhang, A. Fiore, S. H. Yun, H. Kim, and G. Scarcelli, “Line-scanning Brillouin microscopy for rapid non-invasive mechanical imaging,” Sci. Rep. 6(1), 35398 (2016). [CrossRef]
45. K. M. Meek and C. Boote, “The use of X-ray scattering techniques to quantify the orientation and distribution of collagen in the corneal stroma,” Prog. Retinal Eye Res. 28(5), 369–392 (2009). [CrossRef]
46. K. M. Meek, S. J. Tuft, Y. Huang, P. S. Gill, S. Hayes, R. H. Newton, and A. J. Bron, “Changes in collagen orientation and distribution in keratoconus corneas,” Invest. Ophthalmol. Vis. Sci. 46(6), 1948–1956 (2005). [CrossRef]
