Optical Coherence Tomography

Overview

• Optical coherence tomography consists of measuring the optical reflections of biological structures using low-coherence light. ^optical

  • Ultrashort pulsed lasers and supercontinuum lasers are used, but less frequently than superluminescent diodes.

• waves are described as coherent if they have the same waveform and are perfectly in phase.

  • In this context, incoherent means that broadband light is used, modelled as a Gaussian source.

• Optical image tomography is analogous to optical ultrasound. Multiple longitudinal scans are performed laterally to produce a two-dimension map of reflection sites in the scanned area. ^optical

• Typically uses near-infrared light.

• A light source produces a beam of incoherent light that is split by a Michelson interferometer. The interference pattern produced by the interference between the light returning from the sample and the light returning from a reference mirror is then analyzed ^optical.

• To produce a B-scan image, the sample is scanned laterally forming a cross-section of A-scan images.

• Each measurement produces an A-scan image, analogous to that of an ultrasound machine. To produce a 2D or 3D image, the image must be scanned in one or two lateral dimensions respectively.

Time domain OCT (TD-OCT)

^aumann.

• Goal is to identify the position of the scattering object.

• In time-domain OCT, the position of a movable mirror is adjusted, resulting in a burst pattern when the mirror and the object are equidistant from the detector.

  • A broadband source is used because maximal constructive interference is only produced when the two are equidistant.
  • If monochromatic or 2-wavelength sources are used, the interference signal is periodic as a function of τ and the location of the scattering object cannot be determined.

Fourier domain OCT (FD-OCT)

^aumann

• The rate at which the measured reflectance changes as a function of depth varies with wavelength.

  • Red light (greater wavelength) modulates more slowly than blue light (shorter wavelength).

• When reflectance is then plotted as a function of wavenumber ṽ, the values of reflectance for different colours of light form sine waves.

  • The period of these sine waves encodes the depth of the scattering object.

• In spectral domain OCT, the position of the reference mirror is not changed.

• Instead, a spectrometer at the detector measures the amplitude of the interference pattern as a function of the wavenumber ṽ.

• A Fourier transform is then performed on the measured reflectance vs. wavenumber plot to determine the reflectance vs. depth.

• Source: ^twenty

  1. A Gaussian source as for time-domain OCT with a spectrometer. This is Spectral Domain OCT.
  2. A swept-source with a simple detector. A rapidly tunable laser emits a single intense wavelength at a time. This is Swept Source OCT.

• FD-OCT is almost always favoured over TD-OCT because of improved sensitivity as a result of reduced noise. ^twenty

Advantages

^aumann

• Both the axial and lateral resolution of OCT are significantly greater than ultrasound, CT, MRI, and other commonly used imaging technologies.
  • These resolution are typically 5–20 µm

Applications

• In 2014, 5.13 million OCT scans were performed in the US alone, making it one of the most common medical imaging procedures ^Fauw.

• Optical coherence tomography is used most often in ophthalmology, but can be used wherever translucent tissues are found.

  • Effective in early detection of macular degeneration, diabetic retinopathy, glaucoma, and other retinal diseases ^aumann.

  • The ability of OCT to measure the thickness of retinal layers makes it uniquely suitable for glaucoma diagnosis.

  • First introduced in 1996, an opthalmic OCT device was first introduced to the market, but was unpopular because of the low resolution and slow speed of TD-OCT ^aumann.

  • In 2006, SD-OCT devices were first introduced, and have become indispensable to many diagnosticians ^aumann.

• Optical coherence tomography angiography is an angiography technique which uses OCT scans to image the vasculature of the eye in higher-definition than dye angiography and without the risk of dye injection ^{tan2018overview}.

Improvements

• In ophthalmology, the bottleneck lies not in performing OCT scans, but in diagnosing eye conditions and making appropriate referrals from these scans ^Fauw.

• In 2018, researchers at Google’s DeepMind proposed a neural network that could perform segmentation, diagnosis, and make referral recommendations based on 3D OCT scans with accuracy equal to or exceeding that of ophthalmologists ^Fauw.

  • Their model was trained on only 14,884 scans and works on a wide-range of commercially available OCT scanners.

• Their model is composed of two neural networks, a supervised segmentation network which identifies tissues boundaries surround the macula—the central region of the retina—and a supervised classification network which makes diagnosis and referral decisions based on 14,884 scan-diagnosis training pairs.

• While their model shows promise, a randomized controlled trial will need to be performed to determine its effectiveness.

Bibliography

[optical] Huang, Swanson, Lin, Schuman, Stinson, Chang, Hee, Flotte, Gregory, Puliafito & others, Optical coherence tomography, science, 254(5035), 1178-1181 (1991). ↩

[aumann] @incollectionaumann, title=Optical Coherence Tomography (OCT): Principle and Technical Realization, author=Aumann, Silke and Donner, Sabine and Fischer, J"org and M"uller, Frank, booktitle=High Resolution Imaging in Microscopy and Ophthalmology, pages=59-85, year=2019, publisher=Springer ↩

[twenty] De Boer, Leitgeb & Wojtkowski, Twenty-five years of optical coherence tomography: the paradigm shift in sensitivity and speed provided by Fourier domain OCT, Biomedical optics express, 8(7), 3248-3280 (2017). ↩

[Fauw] De Fauw, Ledsam, Romera-Paredes, Nikolov, Tomasev, Blackwell, Askham, Glorot, O’Donoghue, Visentin & others, Clinically applicable deep learning for diagnosis and referral in retinal disease, Nature medicine, 24(9), 1342-1350 (2018). ↩

[tan2018overview] Tan, Tan, Denniston, Keane, Ang, Milea, Chakravarthy & Cheung, An overview of the clinical applications of optical coherence tomography angiography, Eye, 32(2), 262-286 (2018). ↩