OCT-C common mistakes
OCT-C · Understand the traps and how to avoid them.
When asked to compare two OCT light sources, the candidate plugs the spectral bandwidth straight into a memorised resolution formula and concludes that the wider-bandwidth 1050 nm source must have the same or better axial resolution than an 840 nm source, ignoring the centre wavelength entirely.
The axial resolution relationship is memorised as a single string of symbols ('bandwidth divided by something') rather than understood as an inverse-square dependence on centre wavelength, so the candidate never realises that moving the centre wavelength to the right degrades axial resolution for the same bandwidth.
What to do instead
Write out the axial resolution expression in full, with the centre wavelength squared in the numerator and bandwidth in the denominator, and then run two side-by-side calculations: change only the centre wavelength in one, only the bandwidth in the other, and state aloud which parameter moved the number and in which direction.
The candidate describes spectral-domain OCT as depth-resolved by a moving reference mirror and describes swept-source OCT as using a spectrometer, or states that the detector type sets the axial resolution.
Textbook diagrams of the Michelson interferometer dominate study time, so the time-domain architecture becomes the default mental model and the detection arm is treated as an interchangeable detail rather than the defining feature of each generation.
What to do instead
Build a three-column table (time domain, spectral domain, swept source) with rows for what is moved or tuned, what the detector records, and how depth information is recovered (moving mirror; spectrometer plus Fourier transform of the interference spectrum; swept source plus photodetector plus Fourier transform). Note explicitly that axial resolution is set by source bandwidth and centre wavelength, not by detector choice.
On a normal foveal B-scan the candidate labels the outermost hyperreflective line as Bruch's membrane, assigns the RPE label to the band above it, and refers to the ellipsoid zone only as the 'IS/OS junction', missing the interdigitation zone altogether.
The outer bands are learned as an undifferentiated list of acronyms without any histologic anchor, so adjacent bands collapse into each other and older terminology is used without knowing which structure it now names.
What to do instead
Draw the outer retina from inner to outer as four discrete hyperreflective lines with the structure and current name written beside each (external limiting membrane; ellipsoid zone, formerly IS/OS; interdigitation zone, formerly cone outer segment tips; RPE-Bruch's complex), and identify all four on the same normal foveal scan before attempting any pathology scan.
On peripapillary and optic nerve head scans the candidate applies macular layer labels wholesale: the thick innermost band is called the ganglion cell layer, and an RPE band is drawn straight across the cup, with cup and rim judged by eye rather than against Bruch's membrane opening.
All layer practice is done on macular scans, and the candidate does not know that the retinal nerve fibre layer dominates the peripapillary profile and that Bruch's membrane opening defines the neural canal and the reference for disc margins.
What to do instead
Study a normal peripapillary circle scan and a radial disc scan separately, labelling the RNFL as the innermost thick band and noting where the RPE-Bruch's band terminates at Bruch's membrane opening; then state the rule that cup and rim are defined relative to Bruch's membrane opening and the internal limiting membrane, not relative to a visible colour change.
The candidate captures a macular volume without checking the infrared or fundus preview, accepts whatever centration and focus autofocus produced, and files the scan even when the foveal depression sits at the edge of the display or the peripapillary circle is off the disc.
Automated placement and autofocus are trusted as correct by default, and centration is judged from the B-scan only, where a decentred fovea can still produce a plausible-looking image with no depression visible.
What to do instead
Before every capture, verify three things in sequence on the preview image: the target structure sits in the centre of the field, the focus and Z are optimised for sharp layer boundaries, and the signal-strength indicator meets the platform's recommended minimum. Recapture rather than accept any scan that fails one of the three.
The candidate scans a patient with the system laterality still set from the previous patient, or applies the macular cube protocol to an optic nerve head study, and does not notice the mismatch between the saved label and the anatomy shown.
Laterality and protocol are treated as data-entry formalities completed before the patient is positioned, so they are never re-verified against the image actually being acquired.
What to do instead
Adopt a fixed pre-capture checklist that includes confirming patient identity, eye (OD/OS), and protocol against the preview anatomy, and a post-capture check that the saved laterality matches the eye scanned and the protocol matches the structure imaged; correct the record before the patient leaves the chair.
The candidate reports a full-thickness defect or a hyporeflective lesion in the outer retina on a scan where a vertical shadow column runs through all layers and the underlying choroidal signal is extinguished, without identifying the shadow as an artefact.
Only the shape at the level of the outer retina is examined; the candidate has not learned to trace a vertical column of signal loss upward to its cause (floater, dense haemorrhage, ptosis, lens edge, or tear film) and downward to the choroidal dropout it produces.
What to do instead
For every suspect lesion, follow the column: check whether signal is lost above and below it, whether the structure is present in adjacent B-scans of the same volume, and whether the infrared image shows a corresponding opacity. Classify as artefact whenever the abnormality is confined to the shadow column and absent on neighbouring slices.
The candidate accepts automated segmentation lines that jump onto the RPE or onto the inner limiting membrane and then reports the resulting thickness-map values as retinal thickening, without inspecting the lines on the B-scans.
Thickness maps and sector values are treated as direct measurements rather than as the output of an automated algorithm, and the candidate does not know that segmentation failure is common at the edges of pathology and at low signal strength.
What to do instead
Override the map: step through the B-scans of the volume with the segmentation lines displayed, check the inner and outer boundaries at the fovea and at each quadrant, and correct or exclude any slice where the lines depart from the true boundaries before quoting a thickness value.
The candidate labels every hyporeflective space beneath the neurosensory retina as subretinal fluid, and does not distinguish a space bounded below by an intact RPE-Bruch's band (subretinal fluid), a space within the retina with vertical bridging strands (intraretinal fluid), and a dome-shaped elevation of the RPE band with hyporeflective content beneath it (pigment epithelial detachment).
The three appearances are learned from single textbook images rather than from the defining boundary of each compartment, so any dark space under the retina defaults to the most familiar term.
What to do instead
For each hyporeflective space, answer two questions in writing before naming it: which structure forms the floor of the space, and is the RPE band continuous beneath it? Floor of neurosensory retina with intact RPE below indicates subretinal fluid; a space within retinal layers indicates intraretinal fluid; an elevated, continuous RPE band with content beneath indicates pigment epithelial detachment. Confirm by scrolling the adjacent slices to see the three-dimensional shape.
The candidate images the cornea with the standard retinal acquisition protocol, identifies the first hyperreflective surface band as the endothelium, and quotes corneal thickness values without noting that the axial scale differs from the retinal setting.
Posterior-segment habits are transferred unchanged to the anterior segment: the corneal layers are assumed to mirror the retinal order, and machine calibration is assumed to be valid for any tissue in front of the lens.
What to do instead
Use the anterior segment module and its corneal calibration, then label the cornea from front to back on a normal scan (epithelium including Bowman's layer as the first bright band, stroma, Descemet's membrane with endothelium as the posterior bright line), and state explicitly that corneal thickness measurement depends on the correct refractive-index scaling for cornea rather than retina.
When asked about between-patient practice, the candidate discusses cleaning the objective lens and mentions that the instrument does not contact the eye, but omits hand hygiene, the chin-rest and forehead-rest surfaces, and the need to clean any surface the patient's face or hands touched.
OCT is mentally filed as a non-invasive, non-contact test, so the candidate concludes that infection control requirements are minimal and rehearses only the one component that involves the instrument optics.
What to do instead
Memorise the contact points rather than the device category: hands before and after every patient, chin rest and forehead band wiped per practice policy between patients, any surface touched by the patient cleaned, and the fact that non-contact Class 1 low-coherence near-infrared imaging removes ocular surface contact risk but not surface contamination risk.