Key Developments for Partial Coherence Biometry and Optical Coherence Tomography in the Human Eye Made in Vienna

Abstract

Purpose: To describe key developments of optical biometry and optical coherence tomography (OCT) for ophthalmic applications made by one of the pioneering research groups.

Methods: Partial coherence interferometry (PCI) as the basic ranging technology for modern optical biometry and for OCT was introduced for biomedical applications in the 1980s. Later, Fourier domain (FD) OCT was introduced and demonstrated to provide superior sensitivity as compared to time domain OCT. Further developments comprised ultrahigh-resolution OCT and deep-penetration OCT at wavelengths of approximately 1050 nm. Important functional extensions comprise Doppler OCT/OCT angiography, polarization-sensitive OCT, and adaptive optics OCT.

Results: High-precision PCI biometry has found extensive applications in cataract surgery and in research on intraocular lens design. Optical coherence tomography, especially in the second-generation variant of FD OCT, is now indispensable for ocular diagnostics in general and for retinal diagnostics in particular; 1050 nm OCT shows improved penetration into deeper layers like the choroid.

Conclusions: The contributions of the Vienna research group helped to establish PCI biometry and FD OCT as the gold standards in their respective fields.

Optical methods played an important role in ocular diagnostics ever since the introduction of the ophthalmoscope by Hermann von Helmholtz in 1850/1851. In the following 140 years, several other important optics-based instruments were introduced for ocular diagnostics and imaging, like the slit lamp biomicroscope, the Scheimpflug camera, fundus photography, or the scanning laser ophthalmoscope. However, it was not until the introduction of optical coherence tomography (OCT) in 1991¹ that three-dimensional (3D) imaging with micron scale resolution became available, enabling a contact-free, noninvasive “in vivo biopsy” that truly revolutionized ocular diagnostics. 

Although the term “optical coherence tomography” was introduced in 1991, the history of this technology is older. Its predecessor and basic ranging technology, partial (or low) coherence interferometry (PCI or LCI) was first applied to turbid media in the 1970s² and, after dormancy of several years, was re-introduced by Fercher and coworkers^3–5 for the purposes of tissue interferometry and ocular biometry. Partial coherence interferometry found two applications in ophthalmic diagnostics that, over time, revolutionized their fields: high-precision ocular biometry largely replaced the previous ultrasound biometry for applications in cataract surgery, and OCT, in the modified form of Fourier (or spectral) domain (FD or SD) OCT, is now indispensable for retinal diagnostics. Partial coherence interferometry biometry and FD OCT are nowadays the gold standards in their respective fields, with tens of thousands of instruments installed worldwide. Although many research groups are nowadays working in these fields, thousands of scientific papers using these technologies are published each year, and millions of PCI and OCT scans are performed in ophthalmic offices and clinics per year, only a few research groups contributed to the early developments. One of these groups that was involved from the very beginning (founded by one of the authors: AFF), is our group at the Medical University of Vienna. Our group contributed some of the key technologies of PCI and OCT that led to technologies that are now industry standard, like PCI biometry, FD OCT, or 1050-nm OCT as emerging standard. 

This article provides an overview of key developments to PCI and OCT contributed by our group. The article is not intended to be a comprehensive overview of PCI and OCT history—we acknowledge that several other groups contributed fundamental and important research, development, and technology to the field—but a personal view of our contributions. We are grateful to the editors of this feature issue, Jim Fujimoto and David Huang, well known as pioneers in our field, for giving us the opportunity to publish this personal view. 

Basics and First Steps of PCI in Ophthalmology

Optical coherence tomography images consist, analogous to ultrasound images, of optical depth (or A-) scans reassembled into a transverse scan yielding cross-sectional images, or B-scans, of the sample. Ultrasonography uses sound waves that are transmitted from a transducer into the sample to perform the A-scan; the delay of waves reflected back to the transducer yields the depth position of reflecting sample sites. In optics, due to the high speed of light, delay times cannot be directly measured; instead, interferometric techniques are used. Hence, a prerequisite for OCT and interferometric imaging is an interferogram. 

Early biomedical interferograms have been demonstrated by Fercher et al.⁶ A long-coherence Helium-Neon laser beam was directed at the pupil of a subject's eye. The beam was reflected at the cornea and at the ocular fundus, giving rise to an interferogram consisting of concentric interference fringes (Fig. 1). These interferogram fringes pulsate with the subject's heartbeat. 

Figure 1

 

High space- and time-coherence interferogram at eye pupil.

This interference phenomenon offered two basic measurement methods; both techniques require an illuminating beam of high spatial coherence: 

•  
  Using a beam of high temporal coherence enabled the measurement of distance variations between cornea and retina. This provides access to blood pulse–induced dilatations of ocular tissues⁷ and enabled research on ocular blood flow in diseases like glaucoma and diabetic retinopathy.^8,9
•  
  Using a sampling beam of low temporal coherence enabled the measurement of distances between light-reflecting sites, for example the axial length of the eye (see details below).^4,10

The requirements for light sources used for the latter application (for PCI and OCT) are a high spatial coherence (i.e., transversal monomode, Gaussian beam profile) and a very low temporal coherence (broad emission bandwidth, ideally with a Gaussian spectral shape). These requirements were difficult to meet in the early times. The early experiments up to approximately 1985 were performed with dye lasers; later, multimode semiconductor lasers were used. These light sources suffered from problems like beam instabilities (dye lasers) and low spectral bandwidth (multimode diodes). Progress in laser technology (Ti:Al₂O₃ lasers, superluminescent diodes [SLDs]) improved the situation. 

The first measurement of an intraocular distance by PCI was the measurement of the axial eye length, from the corneal apex to the fundus of the eye. A dual-beam illumination scheme was used.^4,10 Figure 2A depicts the basic optical dual-beam scheme: An external Michelson interferometer splits a beam of short coherence length into two components with a temporal delay, indicated as wavelets WL 1 and WL 2, illuminating the eye along a coaxial path. If the Michelson arm length difference A equals the (optical) eye length (L), wavelets WL 1′ and WL 2′, reflected at fundus and cornea, respectively, travel the same total path length and will overlap at the exit of the beam splitter in front of the eye and generate an interferogram at the observation plane (or photodetector). The presence of the interferogram indicates the path length match within the coherence length. Hence, the easily measurable arm length difference (A) of the Michelson interferometer provides the eye length (L). 

Figure 2

 

Basic sketches of PCI. (A) Dual-beam PCI. The sample (eye) is illuminated by a dual beam generated by a Michelson interferometer. (B) Reflectometer PCI. The sample embodies one mirror of a classic Michelson interferometer.

Advantages of the dual-beam technology for eye length measurement are that it requires only one interferometric measurement to identify path length matching, and it is insensitive to small axial eye movements. Hence, it is still in use for measurements of larger distances (eye length for cataract surgery). Disadvantages are its complex interferogram, the need for additional components, and lower sensitivity. With the availability of high-speed path length scanning techniques, the reflectometer technology,¹¹ where the sample is directly placed in one of the interferometer arms, became feasible for rapid OCT imaging of shallow structures like the retina (cf. Fig. 2B). 

Electronic Detection via Heterodyne PCI: Introduction of A-Scans

The detection of the interference phenomenon between beams reflected at the cornea and the ocular fundus and its use for measuring the eye length can be regarded as the foundation of ocular biometry by PCI and as an important step toward OCT. However, the first experiments still suffered from drawbacks that prevented their application in real clinical situations: the laborious procedure of measurement and the rather low sensitivity. 

The operator had to shift interferometer plates in tiny steps over a distance of several millimeters. At each step, the operator had to look for an interference pattern of low contrast. This took approximately 15 minutes for a single measurement, too slow to be practical in patients. Although sensitivity measurements according to present standards were not carried out, the fact that fringes were visible only for the strongest reflecting layer (photoreceptor/Bruch's membrane complex) allows an estimation of the sensitivity of approximately 50 to 60 dB. 

To improve speed and sensitivity, the method was modified: instead of looking for static fringes, a dynamic approach based on the heterodyne detection principle was used.^5,12 The light emitted by a multimode laser diode (MMLD) was split by a Fabry-Perot interferometer into a direct and a delayed beam (path difference 2d, cf. Fig. 3). Both beams were reflected at cornea and retina and superimposed on a photodetector. One of the interferometer plates was moved with constant speed v, causing a Doppler shift f_D = 2v/λ (λ = 780 nm, center wavelength of MMLD) of beam 2. In case of path length matching, an interference signal is observed at the Doppler frequency, which can be extracted by a band pass filter. Thereby, the signal is shifted into a regime in which the low-frequency 1/f noise can be neglected, and the band pass filter greatly reduces other noise. This technique is now known as (time-domain) optical A-scan technique and is the basic ranging technique of time-domain (TD) OCT. (Notes: “Time domain” indicates that the movement of the reference mirror is translated into a time-dependent interferogram; and the use of an MMLD and the extraction of the signal at the Doppler frequency led to the term “laser Doppler interferometry” [LDI], which, however, does not include information on the use of a short-coherence laser and was, therefore, later abandoned.) 

Figure 3

 

Sketch of dual-beam heterodyne PCI. AMP, amplifier; BSC, beam splitter cube; FPI, Fabry Perot interferometer; HeNe, Helium Neon laser; IR, infrared scope; PC, personal computer; PD, photo detector; SM, stepper motor; SMLD, single mode laser diode. Reprinted with permission from Hitzenberger CK. Optical measurement of the axial eye length by laser Doppler interferometry. Invest Ophthalmol Vis Sci. 1991;32:616–624. © 1991 The Association for Research in Vision and Ophthalmology, Inc.⁵

The use of the heterodyne principle reduced the time to measure the eye length into the seconds regime and improved the sensitivity to approximately 75 to 80 dB. For imaging of shorter distances (e.g., for retinal OCT), the complex dual-beam approach with its losses due to the use of the corneal reflection as a reference was not required. Instead, the sample could be directly placed in one arm of a Michelson interferometer (cf. Fig. 2B),¹ which provided a further increase of sensitivity.¹³ 

Figure 4 shows an A-scan obtained with the first dual-beam heterodyne PCI instrument in a healthy eye in vivo.⁵ The signal peak corresponds to an optical length of 33.55 mm. A division by the group refractive index of the ocular media yields the geometric length of 24.78 mm. 

Figure 4

 

Optical A-scan for axial eye length measurement obtained with dual-beam PCI instrument. The envelope of the interferometric heterodyne signal intensity is shown as a function of optical distance to the anterior corneal surface. The signal peak position indicates the optical length of the eye. Reprinted with permission from Hitzenberger CK. Optical measurement of the axial eye length by laser Doppler interferometry. Invest Ophthalmol Vis Sci. 1991;32:616–624. © 1991 The Association for Research in Vision and Ophthalmology, Inc.⁵

After successful demonstration of eye length measurements in healthy subjects, a prospective trial in 196 cataract eyes of 100 patients was carried out,¹⁴ where PCI was compared with ultrasound measurements. A total of 177 (90.5%) of the 196 eyes were measurable by PCI. Figure 5 shows the result of axial eye lengths measured by PCI versus applanation ultrasound (Fig. 5A) and immersion ultrasound (Fig. 5B). An excellent correlation between eye lengths was found (r = 0.97 and 0.99, respectively). Although the axial resolution of this early PCI instrument was limited to approximately 100 μm (by the low bandwidth of the MMLD), the repeatability (SD) of the geometric eye length was much better: 19 μm on average in the cataract eyes, approximately an order of magnitude better than the 150 to 200 μm that had been reported in literature for ultrasound measurements.^15,16 In conjunction with the higher convenience for the patient (no contact or anesthesia needed), this provided the basis for the commercial success of this technology for ocular biometry in cataract surgery (e.g., Carl Zeiss IOL Master; Carl Zeiss Meditec, Jena, Germany), where PCI nearly completely replaced ultrasound-based measurements. 

Figure 5

 

Comparison of eye lengths measured by PCI versus ultrasound in cataract eyes. (A) Partial coherence inferometry versus applanation ultrasound; (B) PCI versus immersion ultrasound. Reprinted with permission from Hitzenberger CK, Drexler W, Dolezal C, et al. Measurement of the axial length of cataract eyes by laser Doppler interferometry. Invest Ophthalmol Vis Sci. 1993;34:1886–1893. © 1993 The Association for Research in Vision and Ophthalmology, Inc.¹⁴

In a further step, the resolution was improved to approximately 15 μm by replacing the MMLD by a broadband SLD, providing a precision in the micron range for corneal and anterior chamber measurements.^17–19 

From A-Scan to B-Scan

Although the initial aim of PCI biometry was the measurement of just a single distance, the axial eye length, it soon became clear that the technology provides more information on ocular tissues. This was already demonstrated in 1990²⁰ and in the first article on heterodyne PCI.⁵ Under certain conditions, two signal peaks were visible whose distance equals the retinal thickness (cf. Fig. 6A). Due to the limited sensitivity of the dual-beam PCI technique, the first peak, corresponding to the inner limiting membrane (ILM), was visible only at proper orientation of the retina (specular reflection from ILM). 

Figure 6

 

Partial coherence interferometry measurements of retinal thickness and fundus profile. (A) Optical A-scan obtained with dual-beam PCI instrument. The first signal peak corresponds to the position of the ILM; the second peak marks the position of the RPE. The peak separation equals the optical thickness of the retina. (B) Fundus profile. The signal peak position is shown as a function of the angle between vision axis (VA) and measurement axis (MA). The excavation of the optic disk is visible at an angle of −12 to −14°. Reprinted with permission from Hitzenberger CK. Optical measurement of the axial eye length by laser Doppler interferometry. Invest Ophthalmol Vis Sci. 1991;32:616–624. © 1991 The Association for Research in Vision and Ophthalmology, Inc.⁵

To analyze the surface contour of the ocular fundus, measurements at different incident angles to the vision axis were carried out, providing fundus profiles. Figure 6B shows an example. Here, as a step toward reassembling full A-scans to generate OCT tomograms, we reassembled signal peaks from adjacent A-scans corresponding to a single anatomical structure, generating a two-dimensional (2D) topogram. The deviation of the measurement points from the theoretic profile (solid line) at approximately 12 to 14° nasal indicates the excavation of the optic disc. 

Huang et al.¹ were the first to convert the signal intensities recorded at adjacent A-scan positions into gray or false color values and to mount the signals to form a cross-sectional tomogram: the first B-scan, recorded in a human retina in vitro. In their seminal paper, Huang et al.¹ also introduced the term “optical coherence tomography” for this technology. 

Because scanning speeds were slow at that time (approximately 1.5 mm/s), we decided to use the dual-beam technique for our first in vivo demonstrations of OCT imaging, as this technique is insensitive to axial sample motions. Figure 7 shows our first in vivo OCT image of a human retina.²¹ Due to the slow scanning speed, we recorded only 21 A-scans for the B-scan. This limited the transverse resolution to approximately 160 μm. Nevertheless, structural details like the retinal thickness, optic disk excavation, and lamina cribrosa are visible. In parallel work, the MIT group demonstrated in vivo retinal OCT images using the reflectometer configuration where the sample was directly placed in the sample arm.^22,23 They used a fiberoptic interferometer design in combination with an x-y scanner and a higher scanning speed (160 mm/s), which reduced the imaging time of a single B-scan (consisting of 100 A-scans) to 2.4 seconds, useful for 2D clinical imaging and enabling commercialization by Carl Zeiss Meditec (Dublin, CA, USA). 

Figure 7

 

Early in vivo optical coherence tomogram. Vertical axis: optical distance to anterior corneal surface. Horizontal axis: angle between vision axis and measurement axis. LC, lamina cribrosa; PE, pigment epithelium; RE, retina. Reprinted with permission from Fercher AF, Hitzenberger CK, Drexler W, Kamp G, Sattmann H. In vivo optical coherence tomography. Am J Ophthalmol. 1993;116:113–114. © 1993 Elsevier Inc.²¹

Noninvasive High-Precision Optical Biometry of the Human Eye

High-Precision Optical Biometry of the Anterior Eye Segment

The first clinically viable instrument enabled fast, noninvasive high-resolution (10 μm) measurements of corneal and lens thickness as well as anterior chamber depth with unique precision (<10 μm; cf. Fig. 8A),^24,25 more than one order of magnitude better than that of state-of-the-art ultrasound and optical techniques. Corneal thickness could be determined with submicrometer precision,²⁶ enabling the investigation of the effect of viscoelastic material during phacoemulsification,²⁷ dorzolamide,²⁸ and small-incision cataract surgery²⁹ on corneal thickness. 

Figure 8

 

Noninvasive high-precision PCI biometry of the human eye. Biometry of the dynamics of the human eye during accommodation: changes of the anterior eye segment (A); movement of the anterior and posterior lens pole (B); axial eye length change (C); significant difference in axial eye elongation in emmetropes and myopes (D). (B) Reprinted with permission from Drexler W, Baumgartner A, Findl O, Hitzenberger CK, Fercher AF. Biometric investigation of changes in the anterior eye segment during accommodation. Vis Res. 1997;37:2789–2800. © 1997 Elsevier Science Ltd.³⁰ and (A, C, D) reprinted with permission from Drexler W, Findl O, Schmetterer L, Hitzenberger CK, Fercher AF. Eye elongation during accommodation in humans: differences between emmetropes and myopes. Invest Ophthalmol Vis Sci. 1998;39:2140–2147. © 1998 The Association for Research in Vision and Ophthalmology, Inc.³³

High-Precision Optical Biometry During Accommodation and IOP Changes

Dual-beam PCI biometry also enabled a careful investigation of the mechanism of human accommodation by precise determination of the anterior chamber depth, thickness of crystalline lens, their changes during accommodation (cf. Fig. 8A) and the movement of the anterior and posterior lens pole (cf. Fig. 8B).³⁰ Another study demonstrated that pilocarpine acts “physiologically” in young phakic subjects, but is a “superstimulus” in presbyopic phakic subjects.³¹ Furthermore, the investigation of implantable contact lens (ICL) dynamics during accommodation revealed that under photopic conditions, with constriction of the pupil, the distance between the ICL and the crystalline lens was significantly reduced, which might be one of the causes of subcapsular opacification in some of the eyes with ICLs.³² 

For the first time, eye elongation during accommodation in emmetropes and myopes was demonstrated (cf. Figs. 8C, 8D).³³ Exaggerated longitudinal eye growth is assumed to play an important role in the development of myopia. Therefore, it is noteworthy that all investigated eyes elongated during accommodation (cf. Fig. 8D), being more pronounced in emmetropes than in myopes. In a similar study using PCI biometry, axial eye length increase and decrease due to IOP elevation and reduction could be detected.³⁴ 

High-Precision Optical Biometry in Pseudophakic Eyes and Pseudo-accommodation

Partial coherence interferometry biometry also had significant impact in pseudophakic eyes, enabling the measurement of the postoperative anterior chamber depth, a hallmark for cataract surgery's IOL power calculation formulae, with high precision (<5 μm) and high resolution (10-12 microns), more than a factor 20 better than conventional ultrasound (cf. Fig. 9A, bottom). For the first time, positive posterior lens capsule to IOL distances (lens capsule distance [LCD]), a possible risk factor for posterior capsule opacification, could be detected and quantified.³⁵ As a consequence, numerous novel IOL designs were extensively investigated with this technique in terms of postoperative anterior chamber depth as well as LCD.³⁶ 

Figure 9

 

Partial coherence interferometry biometry in cataract surgery. Biometry of preoperative phakic cataract eye (A, top) and postoperative pseudophakic eye (A, bottom) with more than 10 times higher precision than applanation ultrasound (B) resulting in improved postoperative refractive outcome (C) using different IOL power formulae (D). OB, optical biometry; US, ultrasound biometry. (A) Reprinted with permission from Findl O, Drexler W, Menapace R, Hitzenberger CK, Fercher AF. High precision biometry of pseudophakic eyes using partial coherence interferometry. J Cataract Refract Surg. 1998;24:1087–1093. © 1998 American Society of Cataract and Refractive Surgery and European Society of Cataract and Refractive Surgeons³⁵ and (D) reprinted with permission from Findl O, Drexler W, Menapace R, Heinzl H, Hitzenberger CK, Fercher AF. Improved prediction of intraocular lens power using partial coherence interferometry. J Cataract Refract Surg. 2001;27:861–867. © 2001 ASCRS and ESCRS.⁴¹

In addition to the exact static postoperative IOL position, PCI biometry could also successfully quantify any IOL dynamics (e.g., before and after Nd:YAG capsulotomy).³⁷ Several IOL designs emerged, promising pseudo-accommodation for pseudophakic patients enabled by ciliary muscle–induced movements of IOLs. Hence, stimulus-driven as well as pharmacologically induced IOL movements of several designs have been investigated with PCI biometry.³⁸ The overall conclusion of numerous investigated pseudo-accommodating IOLs was that on average not even pilocarpine was able to induce a forward movement to result in a refractive change of 0.5 diopters. 

High-Precision Biometry in Cataract Surgery

The most significant clinical impact of PCI biometry was in the field of cataract surgery (cf. Fig. 9A, top). Early studies already demonstrated that PCI was more than 10 times more precise than applanation ultrasound, the gold standard at this time (cf. Fig. 9B).^39,40 This resulted in a possible mean absolute error for postoperative refraction achieved with PCI biometry of 0.49 diopters, resulting in an improvement of 27% (cf. Fig. 9C). Partial coherence interferometry biometry applied to several widely used IOL power formulae yielded significantly better IOL power prediction and therefore refractive outcome in cataract surgery than ultrasound biometry. Further improvement could be achieved by applying PCI to a modified SRK/T formula that predicts the postoperative ACD using PCI biometry data (cf. Fig. 9D).⁴¹ Additional convincing studies investigating the impact of operator experience⁴² on the performance of optical biometry resulted in first commercial prototypes for anterior eye segment⁴³ and axial eye length biometry.⁴⁴ Today this optical biometry based on dual-beam PCI has become the gold standard for biometry in cataract surgery, the most frequently performed ophthalmic surgery. 

Ultrahigh-Resolution OCT

In OCT, transverse resolution as well as depth of focus are governed by the focal spot size, whereas the axial resolution is mainly governed by the product of coherence gating and confocal gating. In low numerical aperture situations, as in retinal imaging, coherence gating dominates, the width of the coherence gate is inversely proportional to the light source bandwidth.⁴⁵ Based on the relationship between source bandwidth and axial resolution, ultrahigh-resolution (UHR) OCT was developed.^46–49 By combining a state-of-the-art TiAl₂O₃ laser that generated pulses of less than 5.5 fs duration (corresponding to a bandwidth of >350 nm at 800 nm center wavelength) with an OCT system optimized to support 260-nm optical bandwidth, an axial resolution of 2 to 3 μm was achieved. This was the highest in vivo ophthalmic resolution at that time, enabling a noticeably superior visualization of all major intraretinal layers⁵⁰ (cf. Figs. 10A, 10B). This first UHR OCT work was accomplished when one of the authors (WD) was at Massachusetts Institute of Technology. 

Figure 10

 

In vivo ultrahigh-resolution retinal OCT. First ultrahigh axial resolution OCT of a normal human fovea (A) and along papillomacular axis (B). Comparison with histology (A, C), with standard resolution (B) and first clinical ultrahigh-resolution OCT study in different retinal pathologies (D). CNV, choroidal neovascularization; CSC, central serous chorioretinopathy; GCL, ganglion cell layer; INL, inner nuclear layer; IPL, inner plexiform layer; IS/OS PR, junction between the inner and outer segment of the photoreceptors; NFL, nerve fiber layer; ONL, outer nuclear layer. (A, B) Reprinted with permission from Drexler W, Morgner U, Ghanta RK, Kärtner FX, Schuman JS, Fujimoto JG. Ultrahigh-resolution ophthalmic optical coherence tomography. Nat Med. 2001;7:502–507⁵⁰; (C) reprinted with permission from Gloesmann M, Hermann B, Schubert C, Sattmann H, Ahnelt PK, Drexler W. Histologic correlation of pig retina radial stratification with ultrahigh-resolution optical coherence tomography. Invest Ophthalmol Vis Sci. 2003;44:1696–1703. © 2003 The Association for Research in Vision and Ophthalmology, Inc.⁵¹; and (D) reprinted with permission from Drexler W. Ultrahigh-resolution optical coherence tomography. J Biomed Opt. 2004;9:47–74. © 2004 Society of Photo-Optical Instrumentation Engineers.⁴⁷

It is noteworthy that the labeling of the intraretinal layers in Figure 2 of Reference 50 is, according to state-of-the-art knowledge, incorrect in the RPE complex. This was due to limited experience in in vivo retinal visualization with this high axial resolution at that time. As a consequence, ex vivo pig (cf. Fig. 10C) and monkey retinal specimens were acquired to correlate UHR OCT images with histology and to provide a basis for improved interpretation of in vivo ophthalmic OCT tomograms of high clinical relevance.⁵¹ In a next step, a compact, clinically viable ultrahigh-resolution (3 μm) ophthalmic OCT system had been developed and used in clinical imaging (cf. Fig. 10D) for the first time.⁵² In this study, a compact, robust, commercially available TiAl₂O₃ laser (Compact Pro; FEMTOLASERS, Vienna, Austria) with up to a 165-nm bandwidth was used in combination with the commercially available OCT 1 system (Carl Zeiss Meditec). Optical coherence tomography imaging was performed with axial scan rates up to 250 Hz using up to 800 μW of incident power in more than 250 eyes of 160 patients with different macular diseases (cf. Fig. 10D). Numerous successful clinical 2D UHR OCT studies on patients with different retinal pathologies followed.^53,54 Because this approach was based on time-domain OCT, only 2D UHR retinal imaging was possible. Since then, ultrahigh axial OCT resolution ≤5 μm became ophthalmic standard. 

Deep-Penetration OCT

Time-domain OCT and the first generation of FD OCT systems operate at a central wavelength of approximately 800 nm. Although 800-nm OCT systems can resolve all major intraretinal layers, they enable only limited penetration beyond the retina due to strong absorption and scattering in the melanin-rich RPE. Furthermore, in clinical OCT imaging, cataract represents a significant challenge when imaging the retina. Optical coherence tomography imaging at different wavelengths can be used to enhance tissue contrast and penetration, as well as to measure absorbing or scattering properties of various pigments and structures. In the 600- to 1200-nm region, scattering decreases monotonically with increasing wavelength. For that reason, OCT imaging at 1050 to 1060 nm can deliver deeper tissue penetration beneath the RPE into the choroid. 

Although the use of a light source at a wavelength of approximately 1100 nm was already suggested in 1991,⁵ it took more than a decade until first in vitro retinal OCT of pig retinas demonstrated enhanced penetration into the choroid.⁵⁵ First in vivo time domain OCT at 1040 nm in healthy human subjects were presented in 2005.⁵⁶ With improved InGaAs line cameras, 1060-nm spectral domain OCT was demonstrated in patients, also demonstrating superior performance in cataract patients.⁵⁷ Further developments included increased scanning speed enabling wide-field 1060-nm OCT,⁵⁸ automated choroidal thickness segmentation⁵⁹ for 3D thickness mapping (similar to retinal thickness mapping),⁶⁰ automated choroidal vessel segmentation⁶¹ for differentiating between Haller's and Sattler's layer of the choroid, as well as simultaneous dual-wavelength eye-tracked UHR retinal and choroidal OCT⁶² (cf. Fig. 11). In numerous clinical 3D 1060-nm OCT studies, the impairment of the choroid in various retinal pathologies could successful be demonstrated.^63,64 Based on these results and those of other groups, there is an increasing trend of developing commercial OCT systems at 1060 nm. 

Figure 11

 

Choroidal OCT. Historic development of choroidal OCT, that is, OCT in the 1040- to 1060-nm wavelength region since 2004.

Fourier-Domain OCT

The introduction of FD OCT marked a major technology advance in OCT, establishing the basis for all modern OCT systems today, as well as for important functional extensions such as OCT angiography. 

The technology of recording an interference pattern as function of optical frequency was already known as white light interferometry.^65,66 Central to the technique is the recording of the spectral interference pattern as a function of optical frequency with a spectrometer. First application to biological tissue (corneal thickness measurement) was presented by Fercher et al. in 1995,⁶⁷ calling it spectral interferometry (Fig. 12). In 1998, Häusler and Lindner⁶⁸ presented first tomograms of skin using the name “spectral radar.” Alternatively, the pattern can be recorded as a function of time using wavelength tuning sources. This method was known earlier as optical frequency domain reflectometry (OFDR).⁶⁹ First imaging results were shown by Chinn et al. in 1997.⁷⁰ In the same year, Lexer et al.⁷¹ demonstrated first in vivo ocular biometry results (Fig. 13). 

Figure 12

 

Measurement of corneal thickness by FD PCI. The main peak at z = 0.77 mm indicates the optical corneal thickness originating from the interference between corneal front and back-surface reflex. Reprinted with permission from Fercher AF, Hitzenberger CK, Kamp G, El-Zaiat SY. Measurement of intraocular distances by backscattering spectral interferometry. Opt Commun. 1995;117:43–48. © 1995 Published by Elsevier B.V.⁶⁷

Figure 13

 

Wavelength tuning interferometry scan (or SS PCI) of a human eye in vivo obtained by tuning the wavelength of an external cavity laser diode over 0.15 nm. ACP, peak indicating anterior chamber depth; AEL, axial eye length; ASL, anterior segment length; VCD, vitreous chamber depth. Reprinted with permission from Lexer F, Hitzenberger CK, Fercher AF, Kulhavy M. Wavelength-tuning interferometry of intraocular distances. Appl Optics. 1997;36:6548–6553. © 1997 Optical Society of America.⁷¹

Interestingly, before retinal tomograms were recorded by FD OCT, already functional FD OCT extensions were demonstrated, such as spectroscopic FD OCT on technical samples,⁷² polarization-sensitive FD OCT on skin,⁷³ and Doppler FD OCT already at 15 kHz A-scan rate.⁷⁴ Finally in 2002, as a result of collaboration between Nicolaus Copernicus University, Torun, and of the University of Vienna, the first in vivo retinal images obtained with FD OCT were presented⁷⁵ (cf. Fig. 14). Those results were surprising: to avoid signal fringe washout by eye motions, very short integration times of 1 ms or less were used, a speed that had been thought to yield unacceptably low sensitivity, based on TD OCT experience and theory. These results led to a careful analysis of signal, noise, and sensitivity in FD OCT. In 2003, Leitgeb et al.⁷⁶ and de Boer et al.⁷⁷ demonstrated that FD OCT has a huge sensitivity advantage over TD OCT. These authors gave a full theoretical description of this advantage with experimental proof. Choma et al.⁷⁸ in the same year demonstrated that the sensitivity advantage holds also for swept source (SS)-based FD OCT. This advantage in sensitivity translated immediately to an improvement in imaging speed without compromising image quality. In 2003, Wojtkowski et al.⁷⁹ showed for the first time video rate retinal tomograms at a 15-kHz A-scan rate and in 2004 already 30 kHz had been achieved by Nassif et al.⁸⁰ for retinal imaging with a system that came closest to modern retinal FD OCT scanners. A year later, the first clinical images were presented with ultrahigh-resolution FD OCT operating at 30 kHz⁸¹ (Fig. 15). Complementary metal oxide semiconductor (CMOS) sensor technology ultimately enabled trespassing of the 100-kHz A-scan rate boundary enabling for imaging rates of several volumes per second.^82,83 

Figure 14

 

First FD OCT scan of the optic nerve head region of a human volunteer in vivo. Reprinted with permission from Wojtkowski M, Leitgeb R, Kowalczyk A, Bajraszewski T, Fercher AF. In vivo human retinal imaging by Fourier domain optical coherence tomography. J Biomed Opt. 2002;7:457–463. © 2002 Society of Photo-Optical Instrumentation Engineers.⁷⁵

Figure 15

 

Example of first demonstration of clinical FD OCT application: RPE layer showing bull's eye dystrophy: OCT fundus view, generated from the UHR OCT A-scans (A) and a 3D UHR OCT overview (B; 3 × 5 × 1 mm) indicate the exact corresponding location of the tomograms depicted in (E–I). Different views of a 3D UHR OCT representation are shown in (B–D); single UHR OCT scans (E–I) taken through the fovea imply thickening of the inner retinal layers, appearing as an inverse traction to the RPE-choroid. Also shown are intraretinal changes such as thinning of the ONL and an absence of the inner segment photoreceptor (ISPR) and outer segment photoreceptor (OSPR) bands in areas of RPE loss (cf. F: area of intraretinal layer impairment is indicated by the red rectangle as well as arrows in I). The two hyperreflective elevations on the RPE band become confluent centrally and intrachoroidal hyperreflective bands (e.g., I, *) merged, creating a prominent choroidal ring pattern within the inferior curvature of the ring structure seen in the rotated specimen (G–I and also C, D). (E, H, rectangles) zoomed areas shown in (F) and (I). Reprinted with permission from Schmidt-Erfurth U, Leitgeb RA, Michels S, et al. Three-dimensional ultrahigh-resolution optical coherence tomography of macular diseases. Invest Ophthalmol Vis Sci. 2005;46:3393–3402. © 2005 The Association for Research in Vision and Ophthalmology, Inc.⁸¹

In FD OCT, the depth structure is encoded by the spectral interference pattern. Fourier transform of the spectrum translates the fringe frequency into a peak position along the delay/depth coordinate being the A-scan. The delay might, however, be positive or negative, depending on the relative position between sample interface and fixed reference arm mirror. This ambiguity leads to mirror image terms that can be suppressed by halving the usable depth range in FD OCT. A first approach to overcome this limitation was demonstrated by Fercher et al.⁸⁴ It is based on phase-stepping interferometry, allowing recording the complex valued interference pattern. Fourier transform of the complex data suppresses the mirror terms and exploits the full nominal depth range of FD OCT, coining the term full-range FD OCT.^85,86 Different phase shifting and modulation techniques have been used with different unique advantages.^87–90 A simple and elegant phase-modulation technique is based on offsetting the sample beam on the galvo-mirror.^91,92 Alternatively, one can exploit the signal dispersion properties that are differently affecting structure and mirror terms.⁹³ 

Current A-scan speed records are set by the alternative FD OCT method of using wavelength tuning. The sources typically operate above 1 μm center wavelength, with 1050-nm apt for human retinal imaging (see the section Deep-Penetration OCT). Several high-speed SS technologies have been introduced so far, capable of reaching 100-kHz A-scan rate and more.⁹⁴ Impressive multi-MHz A-scan rate for retinal imaging has been achieved with a source based on the principle of FD mode locking.⁹⁵ In particular, functional OCT extensions such as OCT angiography benefit from the high-speed capabilities.⁹⁶ 

Recent Developments and Outlook

In this final section, we describe recent developments in the field of OCT by our group that are not yet industry standard but have the potential to set standards for the future. It should be mentioned and acknowledged again that several other groups are working on similar and related developments that are not covered here because of space restrictions. 

Adaptive-Optics OCT

Although UHR OCT improved the depth resolution down to 2 to 3 μm, the transverse resolution of retinal OCT systems was still limited to approximately 20 μm by the aberrations of ocular media. To overcome this limit, adaptive optics (AO), a technique first used in astronomy, was combined with retinal imaging, first for 2D imaging in an AO flood illumination fundus camera,⁹⁷ and later for 3D imaging as AO OCT.^98,99 With this technology, the deformed wave front emerging from the eye is measured by a wave front sensor, analyzed, and corrected by a deformable mirror. The correction of individual aberrations of the investigated eye, including chromatic aberration (pancorrection),¹⁰⁰ finally enabled in vivo cellular resolution of cone photoreceptors and RPE cells in the human retina.¹⁰¹ An alternative version of the technology using a transverse scanning scheme¹⁰² in combination with a lens-based AO OCT scan engine provided 3D visualization of the most densely packed cone photoreceptors in the fovea centralis and of the very narrow rod photoreceptors.¹⁰³ It should be mentioned that these AO OCT systems are still rather bulky and complicated to use. Therefore, their use is presently restricted to a few research laboratories and selected compliant patients. 

Further Improvement of Imaging Speed: Parallel OCT

Ultrahigh-speed SS OCT enables wide-field ophthalmic imaging with multi-MHz A-scans per second.¹⁰⁴ Cost-effective data acquisition and SS technology are still a challenge for clinical use, though. 

Parallel OCT, on the other hand, illuminates the tissue with either a line or over a full area, where each lateral pixel records the depth structure. This intrinsically yields high imaging speed at reduced complexity. Line-field (LF) FD OCT, first demonstrated by Zuluaga and Richards-Kortum,¹⁰⁵ images a line along the sample onto a 2D sensor after being dispersed into its wavelength components. A full tomogram can therefore be recorded with a single camera shot as shown by Grajciar et al.¹⁰⁶ and Nakamura et al.¹⁰⁷ for in vivo ocular imaging. Spectrometer-based FD OCT, however, suffers from sensitivity roll-off and pixel cross-talk. Swept-source OCT, on the other hand, samples the spectral data separated in time, significantly reducing those artifacts. Bonin et al.^108,109 demonstrated multi-MHz A-scan rate for full-field SS OCT with an expensive high-speed 2D camera. Fechtig et al.^110,111 showed retinal LF SS OCT with off-shelve camera technology at up to 1-MHz A-scan rate. A particular advantage of parallel OCT is the exceptional lateral phase stability, enabling digital refocusing and aberration correction.^112,113 

Doppler OCT and OCT Angiography

The main shortcoming of OCT based on backscattering alone is its missing specificity to biological structures of interest. Functional extensions allow partially mitigating this deficiency by providing label-free intrinsic tissue contrast. Doppler OCT (DOCT), as one of the most promising functional OCT candidates,¹¹⁴ should provide depth-resolved quantitative information on blood flow and thereby generate angiographic maps. First implementations of DOCT were based on TD OCT and showed potential to contrast and quantify blood flow within selected tissue vessels.^115–117 However, TD OCT was too slow to assess retinal vasculature over a large range of vessels and was limited in field of view.¹¹⁸ This changed with the introduction of FD OCT that brought a strong impetus for DOCT.^83,119–123 Dynamic DOCT could resolve pulsatile flow in a reliable quantitative manner and without prior knowledge on vessel geometry, based on multibeam^124–128 or volume-acquisition schemes.^129–131 The high acquisition speed allowed furthermore for novel flexible scanning schemes to obtain 3D angiographic maps without compromising overall imaging time.^132,133 Modern OCT angiography displays signal differences between successive tomograms, being thereby sensitive to motion of red blood cells, and achieves high microvascular contrast down to the level of smallest capillaries.^96,132–141 

Polarization-Sensitive OCT

Conventional OCT measures the intensity of backscattered light; however, some tissues show only poor intensity contrast. By measuring the light's polarization state, additional intrinsic, tissue-specific contrast can be generated. Polarization sensitive (PS) OCT is based on PS PCI¹⁴² and was first reported by de Boer et al.¹⁴³ for skin imaging. Quantities measured by early time-domain PS OCT were retardation¹⁴³ and axis orientation¹⁴⁴ of birefringent skin and muscle, first applications in the human eye measured retardation of the retinal nerve fiber layer¹⁴⁵ (useful for glaucoma diagnostics) and of cornea¹⁴⁶ (e.g., for keratoconus diagnostics). The first high-speed FD PS OCT retinal scanner was developed by Götzinger et al. in 2005.¹⁴⁷ Other groups developed similar systems based on somewhat different technology (cf., e.g., Refs. 148, 149). One of the most important developments in this field was the discovery of light depolarization by the RPE,¹⁵⁰ which can be used to segment the RPE.¹⁵¹ This effect was used to segment lesions of the RPE,¹⁵² which is of considerable interest for quantifying drusen¹⁵³ and geographic atrophy¹⁵⁴ in patients with AMD. A comprehensive overview of PS OCT in the eye can be found in a recent review.¹⁵⁵ 

Acknowledgments

The authors thank the following colleagues for their contributions: A. Baumgartner, H. Sattmann, L. Schachinger, B. Hermann, B. Povazay, A. Unterhuber, K. Bizheva, M. Esmaeelpour, B. Hofer, E.J. Fernandez, B. Grajciar, C. Blatter, D. Fechtig, A. Singh, L. Ginner, A. Kumar, T. Schmoll, M. Sticker, M. Pircher, E. Götzinger, B. Baumann, S. Zotter, T. Torzicky, F. Felberer, W. Trasischker, M. Sugita, R. Haindl, and W. Wartak, from the Center of Medical Physics and Biomedical Engineering of the Medical University of Vienna, Austria, as well as B. Bobr/Kiss, M. Wirtitsch, C. Leydolt/Köppl, K. Kriechbaum, G. Rainer, V. Petternel, M. Bolz, A. Prinz, W. Buehl, M. Stur, E. Ergun, C. Scholda, S. Michels, C. Ahlers, F. Schlanitz, C. Schütze, P. Roberts, M. Ritter, R. Sayegh, W. Geitzenauer, J. Lammer, C. Skorpik, C. Vass, and U. Schmidt-Erfurth, from the Department of Ophthalmology and Optometry of the Medical University of Vienna, Austria; J.E. Morgan, V. Kajic, C. Torti, A. Tumlinson from the School of Optometry, Cardiff University, UK; J.G. Fujimoto, U. Morgner, F.X. Kärtner, R.K. Ghanta, E.P. Ippen, C. Pitris, and T.H. Ko from Massachusetts Institute of Technology; P.K. Ahnelt, M. Glössmann, C. Schubert, E.M. Anger from the Department of Physiology, Medical University Vienna; S. Binder and C. Glittenberg from the Rudolfstiftung, Vienna; P. Artal from Laboratorio de Optica, Departamento deFísica, Universidad de Murcia (Spain); H.D. Gnad and M. Juchem from the Department of Ophthalmology, Lainz Hospital, Vienna; D. Stifter and M. Wurm, Recendt GmbH, Linz, T. Lasser, A. Bachmann, M. Villiger, Michaely, Ecole Polytechnique Federale, Lausanne (Switzerland), M. Wojtkowski, and A. Kowalczyk, Nicolaus Copernicus University, Torun (Poland); and K. Mengedoht and E. Roth, University of Essen (Germany). 

Supported by Medical University Vienna, Austrian Science Fund, Cardiff University (Wales, UK), Carl Zeiss Meditec (US and Germany), Heidelberg Engineering, Femtolasers GmbH (Austria), Imagine Eyes (France), Hamamatsu (Japan), EXALOS (Switzerland), Superlum (Russia), Canon (Japan), Christian Doppler Gesellschaft, FWF-NFN PAI Network Macular Vision Research Foundation (US), Action Medical Research (AP1110), DTI (1544C), BBSRC, MRC, EC FUN OCT (FP7 HEALTH, 201880), FAMOS (FP7 ICT, 317744), BiopsyPen (FP7 ICT, 611132). 

Disclosure: C.K. Hitzenberger, None; W. Drexler, None; R.A. Leitgeb, None; O. Findl, None; A.F. Fercher, None