Fundamentals of OCT: Physics, Light Sources, Interferometry, Scan Types and Instrumentation
The core idea
OCT is the optical analogue of ultrasound: it measures the echo delay of near-infrared light using low-coherence interferometry. Light is split into a sample arm (eye) and a reference arm (mirror or fixed path). Interference occurs only when the two pathlengths match to within the source coherence length, so the coherence gate supplies optical sectioning without any physical gate. Depth information comes from the interference pattern, not from focusing.
What determines resolution
- Axial resolution is set by the source bandwidth: dz ≈ 0.44 λ²/Δλ in air, divided by tissue refractive index. A 50 nm broadband superluminescent diode centred at 840 nm gives roughly 5–7 µm in air and ~4–5 µm in retina. Wider bandwidth = finer axial resolution.
- Lateral resolution (~15–20 µm) is set by the delivery optics, pupil size and spot size at the retina - not by bandwidth.
- Imaging depth/penetration depends on wavelength and scattering: ~840 nm gives good retinal detail; ~1050 nm (swept-source) penetrates choroid and sclera better.
Generations and trade-offs
Time-domain OCT uses a moving reference mirror, one A-scan at a time; slow, and sensitivity falls as scan speed rises. Spectral-domain OCT keeps the reference fixed and reads all depths simultaneously with a spectrometer - about 20–30 dB better SNR, 27,000–70,000 A-scans/s, but with sensitivity roll-off with depth and a mirror-image ambiguity near the zero-delay line. Swept-source OCT uses a tunable laser and photodetector, is fastest (100,000+ A-scans/s), has a longer imaging range and flatter roll-off.
Scan geometry
A-scan = depth profile; B-scan = 2D cross-section; C-scan = en-face; raster/volume; peripapillary circle; radial line.
Common trap: confusing axial with lateral resolution, or confusing displayed pixel spacing with true optical resolution - upsampling a scan does not improve resolution.
Rule of thumb: bandwidth buys depth resolution; optics buy lateral resolution; wavelength buys penetration. If an image looks blurry laterally but layers are crisp axially, the problem is focus, not the light source.