Functional Optical Coherence Tomography

• Functional optical coherence tomography (fOCT) is a suite of imaging techniques that extend traditional OCT by capturing dynamic physiological, mechanical, and molecular contrasts.
• fOCT measures time-varying changes in amplitude, phase, polarization, and dispersion to reveal flow, cellular activity, and molecular interactions with high resolution.
• These methods enable diverse applications including retinal optoretinography, dynamic cellular imaging, and bond-selective photothermal OCT for advanced tissue analysis.

Functional optical coherence tomography (fOCT) denotes OCT methods that extract information beyond pure morphology from the same depth-resolved interferometric framework that underlies structural OCT. In this broader usage, OCT is no longer restricted to a map of backscattered intensity or interface position; it is used to measure physiological, mechanical, spectroscopic, metabolic, or chemically selective observables through time-varying changes in amplitude, phase, polarization, wavelength dependence, thermal response, or molecularly mediated contrast. The resulting family spans retinal intrinsic signal optoretinography, angiographic and Doppler flow imaging, dynamic cellular imaging, thermometry, dispersion mapping, and bond-selective photothermal OCT, with implementations ranging from point-scanning Fourier-domain systems to high-NA full-field interferometric microscopes (Kim et al., 2022, Jianlong et al., 2023, Zong et al., 2023).

1. Definition and scope

Structural OCT is a low-coherence interferometric modality that reconstructs depth-resolved reflectivity and thereby maps tissue microstructure, but its standard output remains primarily morphological. fOCT extends this framework by exploiting additional observables: flow or motion, polarization, elastic or mechanical response, intrinsic optical properties, thermal or photothermal changes, molecular state, and spectrally dependent propagation effects. In that sense, fOCT is not a single technique but a class of OCT methods in which physiology or material state modulates the OCT signal in a measurable, depth-resolved way (Zong et al., 2023, Jianlong et al., 2023).

Within this taxonomy, optoretinography is the retinal specialization of fOCT: intrinsic optical signals are measured as stimulus-evoked changes in OCT intensity or phase, enabling layer-specific functional imaging of photoreceptors, subretinal space, retinal vasculature, and inner retinal circuitry. Dynamic full-field implementations form another major branch, using temporal fluctuations of interferometric intensity to reveal cellular and subcellular activity in living tissues. Other branches derive functional contrast from dispersion, viscoelastic fluctuations, or mid-infrared photothermal modulation that converts vibrational absorption into an OCT-detectable signal (Kim et al., 2022, Apelian et al., 2016, Kolenderska et al., 2017).

A recurrent misconception is that fOCT is synonymous with OCT angiography or Doppler OCT. The published literature instead treats angiography, dynamic cellular OCT, optoretinography, thermometry, dispersion mapping, and bond-selective photothermal OCT as distinct but related functional extensions of OCT, each tied to a different physical observable (Jianlong et al., 2023, Zong et al., 2023).

2. Signal channels and governing observables

The most direct distinction among fOCT methods is the OCT signal component they interrogate. Phase-sensitive retinal fOCT uses the fact that an OCT phase difference corresponds to an optical path length change according to

$$\Delta \mathrm{OPL} = \frac{\lambda_0}{4\pi}\Delta\phi ,$$

so nanometric stimulus-evoked deformations in photoreceptor outer segments or inter-layer spacing can be inferred from repeated complex OCT measurements. Differential Phase Mapping extends this logic by differentiating unwrapped phase along depth, producing a depth-resolved map of relative scatterer spacing and its temporal evolution (Kim et al., 2022).

Amplitude-based fOCT relies on temporal changes in the OCT magnitude or intensity. In the dynamic-amplitude formalism, the first-order field autocorrelation and second-order intensity autocorrelation are

$$g_1(\tau)=\frac{\langle E(t)E^*(t+\tau)\rangle}{\langle |E(t)|^2\rangle}, \qquad g_2(\tau)=\frac{\langle I(t)I(t+\tau)\rangle}{\langle I(t)\rangle^2},$$

with the Siegert relation

$g_2(\tau)=1+\beta |g_1(\tau)|^2 .$

These quantities encode motion-induced decorrelation from blood flow, diffusion, intracellular transport, or thermally induced displacement. Spectrum-based dynamic OCT instead analyzes the temporal power spectral density of OCT amplitude fluctuations and uses mean frequency, frequency variance, or band-integrated power as functional descriptors (Jianlong et al., 2023).

Dynamic full-field OCT makes this particularly explicit. At a fixed coherence-gated plane, the interferometric signal fluctuates because intracellular scatterers undergo Brownian and active motion, cytoskeletal rearrangement, membrane fluctuations, and organelle transport. The temporal standard deviation, autocorrelation, or power spectrum of each pixel then becomes a functional readout of subcellular motility and metabolic state rather than a nuisance speckle term (Apelian et al., 2016, Scholler et al., 2019).

Photothermal and chemically selective fOCT use a different chain of observables. In bond-selective full-field OCT, a tunable mid-infrared pump excites specific vibrational modes, producing a temperature rise

$$\Delta T(\mathbf r)\propto \frac{\mu_a(\tilde\nu)\,F_{\mathrm{MIR}}(\mathbf r)}{\rho c_p},$$

which in turn changes optical path length through thermal expansion and refractive-index modulation. The reconstructed OCT field magnitude in “hot” and “cold” states is differenced, giving a photothermal signal

$$\Delta E_{\mathrm{sample}}(\tilde\nu)\propto \mu_a(\tilde\nu)\,F_{\mathrm{MIR}} .$$

Bond selectivity is therefore inherited from the mid-IR absorption coefficient, while depth sectioning is inherited from OCT coherence gating (Zong et al., 2023).

A further functional observable is chromatic dispersion. Dispersion mapping estimates the group velocity dispersion from the difference in optical thickness measured in two numerically synthesized spectral sub-bands: $$\beta_2 = - \frac{1}{c\,\Delta \omega_{ab}} \frac{\Delta z_{ab}}{l_s}.$$ Here the functional parameter is not reflectivity but the sample’s group velocity dispersion, extracted from ordinary Fourier-domain OCT data by postprocessing alone (Kolenderska et al., 2017).

3. Major methodological families

The diversity of fOCT is most transparent when organized by contrast mechanism.

Family	Primary contrast source	Representative implementations
Flow and hemodynamics	Amplitude decorrelation, variance, dynamic light scattering, phase shifts	SSADA, svOCT, cmOCT, DLS-OCT (Jianlong et al., 2023)
Retinal intrinsic signals	Stimulus-evoked intensity and phase changes, OPL shifts	Optoretinography, Differential Phase Mapping (Kim et al., 2022)
Dynamic cellular OCT	Temporal speckle fluctuations from intracellular motion	dmOCT, D-FFOCT, dynamic OCT (Kohlfaerber et al., 2022, Scholler et al., 2019)
Interface-stabilized full-field dynamic OCT	Same dynamic contrast with self-referenced interferometry near coverslips	iSR D-FFOCT (Monfort et al., 2023)
Chemical and bond-selective OCT	Mid-IR photothermal modulation from vibrational absorption	BS-FF-OCT (Zong et al., 2023)
Dispersion- and fluctuation-based material OCT	GVD walk-off or thermal fluctuation spectra	Dispersion mapping, dynamically enhanced OCT (Kolenderska et al., 2017, Mitsui et al., 2017)

Point-scanning Fourier-domain systems dominate vascular and retinal fOCT because they provide flexible sampling, robust confocal rejection, and established phase-sensitive processing pipelines. Full-field systems occupy a different regime: they acquire an entire en face plane simultaneously, favor high lateral resolution with high-NA objectives, and are naturally suited to widefield pump-probe or dynamic speckle analyses. Comparative work on full-field OCT and optical transmission tomography further emphasizes that the same full-field platform can operate in static morphology and dynamic metabolic-contrast regimes, with reflection-mode FFOCT providing stronger axial sectioning and transmission-mode OTT providing complementary forward-scattering sensitivity (Alhaddad et al., 2023).

This multiplicity matters conceptually. fOCT does not denote one canonical biomarker; it denotes a family of OCT-derived observables whose interpretability depends on acquisition geometry, scattering regime, temporal sampling, and the specific physical process that perturbs the complex OCT field.

4. Representative implementations and demonstrated applications

In airway imaging, dynamic microscopic OCT uses temporal tissue fluctuation as contrast to reveal ciliary motion, mucus secretion and transport, immune cell trafficking, and metabolically active epithelium in vivo or ex vivo. In that implementation, 150 B-scans are acquired at the same position with an effective B-scan rate of 111 Hz, and the temporal spectrum is integrated in three bands: $<0.4$ Hz, $0.4\text{–}4.9$ Hz, and $4.9\text{–}25$ Hz. Cilia appear as bright yellow structures in the false-color representation, and ciliary beat frequency is extracted voxel-wise as the dominant frequency above 3 Hz. After ATP-$\gamma$-S stimulation, epithelium cilia increased from $$g_1(\tau)=\frac{\langle E(t)E^*(t+\tau)\rangle}{\langle |E(t)|^2\rangle}, \qquad g_2(\tau)=\frac{\langle I(t)I(t+\tau)\rangle}{\langle I(t)\rangle^2},$$0 Hz to 25 Hz and gland duct cilia from $$g_1(\tau)=\frac{\langle E(t)E^*(t+\tau)\rangle}{\langle |E(t)|^2\rangle}, \qquad g_2(\tau)=\frac{\langle I(t)I(t+\tau)\rangle}{\langle I(t)\rangle^2},$$1 Hz to 23 Hz, while previously inactive ductal cilia began beating at 14–20 Hz (Kohlfaerber et al., 2022).

Retinal fOCT for intrinsic signal optoretinography targets a different functional regime. Phase-sensitive OCT has shown a rapid outer-segment optical path length decrease, consistent with fast photoreceptor outer-segment shrinkage within a few ms after light onset, followed by a slower outer-segment OPL increase over $$g_1(\tau)=\frac{\langle E(t)E^*(t+\tau)\rangle}{\langle |E(t)|^2\rangle}, \qquad g_2(\tau)=\frac{\langle I(t)I(t+\tau)\rangle}{\langle I(t)\rangle^2},$$2 s. Inner-segment ellipsoid reflectivity changes appear with a delay of $$g_1(\tau)=\frac{\langle E(t)E^*(t+\tau)\rangle}{\langle |E(t)|^2\rangle}, \qquad g_2(\tau)=\frac{\langle I(t)I(t+\tau)\rangle}{\langle I(t)\rangle^2},$$3 ms relative to fast outer-segment signals, subretinal space thickness changes track light- and dark-dependent fluid transport, and hemodynamic IOS in the superficial, intermediate, and deep vascular plexuses follow neural IOS with significant delay. The review also reports phase-resolved full-field OCT detection of an $$g_1(\tau)=\frac{\langle E(t)E^*(t+\tau)\rangle}{\langle |E(t)|^2\rangle}, \qquad g_2(\tau)=\frac{\langle I(t)I(t+\tau)\rangle}{\langle I(t)\rangle^2},$$4 nm OPL increase between ganglion cell layer and inner plexiform layer over several seconds (Kim et al., 2022).

Dynamic full-field OCT has established a cellular and subcellular branch of fOCT. In fresh ex vivo tissues, temporal standard deviation of the interferometric signal suppresses static fibrous scatterers and reveals metabolically active cells; glycolysis inhibition with 2-deoxy-D-glucose caused the global D-FFOCT contrast in rat liver to drop sharply from $$g_1(\tau)=\frac{\langle E(t)E^*(t+\tau)\rangle}{\langle |E(t)|^2\rangle}, \qquad g_2(\tau)=\frac{\langle I(t)I(t+\tau)\rangle}{\langle I(t)\rangle^2},$$5 to $$g_1(\tau)=\frac{\langle E(t)E^*(t+\tau)\rangle}{\langle |E(t)|^2\rangle}, \qquad g_2(\tau)=\frac{\langle I(t)I(t+\tau)\rangle}{\langle I(t)\rangle^2},$$6, whereas rotenone had no significant effect under the same ex vivo conditions (Apelian et al., 2016). In retinal organoids, D-FFOCT generated colored images with an endogenous contrast linked to organelle motility, with sub-micrometer spatial resolution and millisecond temporal resolution, and distinguished retinal progenitors, differentiating retinal pigment epithelium, and photoreceptor precursors by dynamic signature alone (Scholler et al., 2019). A later microscope-integrated D-FFOCT module extended this to longitudinal in vitro imaging over periods of minutes to weeks on the same sample, demonstrated daily volumetric imaging from d27 to d43, an optimal workflow with a time gain factor of 10, and deep imaging of d266 retinal organoids at 190–230 $$g_1(\tau)=\frac{\langle E(t)E^*(t+\tau)\rangle}{\langle |E(t)|^2\rangle}, \qquad g_2(\tau)=\frac{\langle I(t)I(t+\tau)\rangle}{\langle I(t)\rangle^2},$$7m depth using 810 nm illumination and temporal binning (Monfort et al., 2023).

Interface self-referenced D-FFOCT addressed a practical limit of conventional Linnik D-FFOCT near reflective coverslips by using the coverslip itself as a defocused reference. In fibroblast cultures, this removed the $$g_1(\tau)=\frac{\langle E(t)E^*(t+\tau)\rangle}{\langle |E(t)|^2\rangle}, \qquad g_2(\tau)=\frac{\langle I(t)I(t+\tau)\rangle}{\langle I(t)\rangle^2},$$8m dead zone near the interface, enabled artefact-free imaging at $$g_1(\tau)=\frac{\langle E(t)E^*(t+\tau)\rangle}{\langle |E(t)|^2\rangle}, \qquad g_2(\tau)=\frac{\langle I(t)I(t+\tau)\rangle}{\langle I(t)\rangle^2},$$9m from the coverslip, and improved mean-frequency stability by 42.08 dB relative to standard D-FFOCT. The reported dynamic point-spread function at 810 nm was 378.4 nm transversely and 415 nm axially, with maximal sensitivity of 43.52 dB in fibroblasts (Monfort et al., 2023).

Bond-selective full-field OCT extends fOCT into molecularly selective vibrational imaging. The system used a 545 nm, 100 nm FWHM broadband LED probe with 50× air objectives, a tunable mid-IR OPO spanning 1320–1775 cm$g_2(\tau)=1+\beta |g_1(\tau)|^2 .$0, and a four-phase phase-shifting FF-OCT reconstruction. It demonstrated $g_2(\tau)=1+\beta |g_1(\tau)|^2 .$1m PMMA beads in agarose with on-resonance contrast at 1730 cm$g_2(\tau)=1+\beta |g_1(\tau)|^2 .$2 and disappearance off resonance at 1770 cm$g_2(\tau)=1+\beta |g_1(\tau)|^2 .$3, 3D hyperspectral imaging of polypropylene fibers from a surgical mask with peaks at $g_2(\tau)=1+\beta |g_1(\tau)|^2 .$4 and $g_2(\tau)=1+\beta |g_1(\tau)|^2 .$5 cm$g_2(\tau)=1+\beta |g_1(\tau)|^2 .$6, protein-rich contrast in formalin-fixed T24 bladder cancer spheroids at 1650 cm$g_2(\tau)=1+\beta |g_1(\tau)|^2 .$7, protein-rich internal structures in fixed adult C. elegans, and protein- and lipid-rich myelinated axon bundles across a 150 $g_2(\tau)=1+\beta |g_1(\tau)|^2 .$8m mouse brain slice with strong contrast at 1650 and 1740 cm$g_2(\tau)=1+\beta |g_1(\tau)|^2 .$9 and very weak contrast at 1775 cm$$\Delta T(\mathbf r)\propto \frac{\mu_a(\tilde\nu)\,F_{\mathrm{MIR}}(\mathbf r)}{\rho c_p},$$0 (Zong et al., 2023).

A separate mechanically oriented branch uses Fourier-domain OCT plus temporal fluctuation spectroscopy to measure viscoelastic characteristics of each cross-section. In that work, thermal fluctuation spectra of liquid interfaces and layered samples were measured depth-resolved, and noise was reduced below shot-noise levels using averaged correlations, allowing extraction of spectral information otherwise unobtainable. Demonstrations included silicone oil, water-plus-glass-plus-oil stacks, a human finger, and a sweetfish eye (Mitsui et al., 2017).

5. Acquisition, reconstruction, and quantification

The data path from structural OCT to fOCT is usually a repetition strategy plus a functional estimator. In amplitude-based vascular fOCT, repeated B-scans or volumes at the same location are processed into decorrelation or variance maps. SSADA computes split-spectrum amplitude decorrelation between repeated B-scans, speckle-variance OCT computes the temporal variance of repeated intensity frames, and correlation-mapping OCT computes local correlation or decorrelation within a spatial window. DLS-OCT extends this by fitting voxel-wise temporal autocorrelation functions to recover flow fraction, axial and transverse velocity components, and a diffusion-related parameter (Jianlong et al., 2023).

At cellular scale, the same principle becomes fluctuation spectroscopy. dmOCT acquires a time series at a fixed cross-section and Fourier transforms the absolute OCT signal in time; band-integrated amplitudes then populate RGB channels, while voxel-wise dominant-frequency analysis above 3 Hz yields ciliary beat maps (Kohlfaerber et al., 2022). D-FFOCT and iSR D-FFOCT adopt an HSB mapping: brightness from running temporal standard deviation over a 50-frame window, hue from mean frequency of the power spectral density, and saturation from the standard deviation of frequency. In the microscope-integrated D-FFOCT workflow, acquisition, GPU processing, and saving were overlapped so that the total time per D-FFOCT image approached the acquisition-limited 5.12 s for a 512-frame sequence (Monfort et al., 2023, Monfort et al., 2023).

Bond-selective OCT adds pump-probe timing and interferometric demodulation. For each depth, the reference mirror is stepped through $$\Delta T(\mathbf r)\propto \frac{\mu_a(\tilde\nu)\,F_{\mathrm{MIR}}(\mathbf r)}{\rho c_p},$$1; multiple camera frames acquired during mid-IR “on” and “off” periods are averaged; hot and cold FF-OCT images are reconstructed independently; and the photothermal image is obtained by subtraction. In the reported implementation, MIR and probe pulses ran at 20 kHz, the chopper frequency was 50 Hz, and the camera frame rate was 100 Hz, so alternate frames corresponded to hot and cold states. For thick brain tissue, an alternative timing with longer delays was required to match the measured cooling time constant $$\Delta T(\mathbf r)\propto \frac{\mu_a(\tilde\nu)\,F_{\mathrm{MIR}}(\mathbf r)}{\rho c_p},$$2 ms (Zong et al., 2023).

Some fOCT variants are almost purely computational add-ons. Dispersion mapping requires no system modification: the original broadband interferogram is resampled to $$\Delta T(\mathbf r)\propto \frac{\mu_a(\tilde\nu)\,F_{\mathrm{MIR}}(\mathbf r)}{\rho c_p},$$3, numerically dispersion-compensated, split into two Gaussian sub-spectra, Fourier transformed separately, and analyzed for optical-thickness walk-off. Because the method depends on axial peak localization, its precision is governed by bandwidth, sub-band separation, and zero-padding. The paper explicitly quantified the effect of FFT size, showing discretization of measured $$\Delta T(\mathbf r)\propto \frac{\mu_a(\tilde\nu)\,F_{\mathrm{MIR}}(\mathbf r)}{\rho c_p},$$4 values with insufficient zero-padding and improved resolution with larger transforms (Kolenderska et al., 2017).

6. Interpretation, limitations, and future directions

Several interpretive issues recur across the field. First, fOCT signals are rarely generated by a single microscopic process. In retinal ORG, IOS are a superposition of outer-segment mechanics, osmotic and volume changes, inner-segment metabolic responses, subretinal fluid transport, and hemodynamic coupling; the review explicitly cautions that interpretation of IOS as a single mechanism such as “pure OS swelling” is risky without additional evidence (Kim et al., 2022). Dynamic cellular OCT has a related ambiguity: temporal fluctuations are linked to metabolism and organelle motion, but the exact cellular or subcellular processes generating the signals are not fully known, and label-free dynamics do not yet distinguish specific immune-cell subtypes in airway tissue (Kohlfaerber et al., 2022, Monfort et al., 2023).

Second, robustness differs strongly across fOCT branches. Phase-based ORG offers nanometric sensitivity but is highly vulnerable to eye motion and phase noise. Amplitude-based dynamic methods are more robust and less complex, which is one reason OCT angiography achieved significant clinical success, but they often provide qualitative or semi-quantitative rather than uniquely parameterized physiological observables unless a DLS-style model is imposed (Jianlong et al., 2023). Full-field dynamic systems can deliver subcellular resolution, but conventional two-arm D-FFOCT is sensitive to vibration; iSR D-FFOCT mitigates this near interfaces yet remains intrinsically limited to a few micrometers above the coverslip by the coherence length (Monfort et al., 2023).

Third, some of the most distinctive fOCT contrasts remain difficult to translate in vivo. BS-FF-OCT requires a tunable mid-IR laser, optical chopping, synchronization, and careful control of water absorption; biological samples in the reported work were often in D$$\Delta T(\mathbf r)\propto \frac{\mu_a(\tilde\nu)\,F_{\mathrm{MIR}}(\mathbf r)}{\rho c_p},$$5O-based buffers and sandwiched between CaF$$\Delta T(\mathbf r)\propto \frac{\mu_a(\tilde\nu)\,F_{\mathrm{MIR}}(\mathbf r)}{\rho c_p},$$6 windows, and the authors explicitly identify thermal safety, imaging speed, and full-field motion sensitivity as major obstacles to in vivo deployment (Zong et al., 2023). Dispersion mapping is simpler computationally but assumes layered geometry, sufficient source bandwidth, accurate system-dispersion compensation, and reliable layer-thickness estimates; curved or strongly scattering tissues make these assumptions less secure (Kolenderska et al., 2017).

A further limitation is standardization. The ORG review identifies the lack of standardized imaging protocols, stimulus paradigms, normalization schemes, and layer assignments as a central deployment challenge for clinical retinal fOCT (Kim et al., 2022). The dynamic-amplitude review makes a parallel point for OCTA, DLS-OCT, dynamic OCT, and thermometry: repeated sampling increases sensitivity to motion artefacts, while quantitative interpretation of decorrelation or spectral features remains system- and protocol-dependent (Jianlong et al., 2023).

The current trajectory nevertheless points toward consolidation rather than fragmentation. Faster cameras, MHz OCT engines, voxel-wise 3D registration, dynamic FF-OCT methods, sparse spectral sampling for bond-selective imaging, and machine-learning analysis of fluctuation signatures all appear in the recent literature as concrete development paths. This suggests a mature view of fOCT: not a single modality competing with structural OCT, but a multi-contrast framework in which structure, flow, phase, intracellular dynamics, dispersion, temperature, and molecularly selective photothermal response can be co-registered within one interferometric imaging paradigm (Monfort et al., 2023, Zong et al., 2023, Jianlong et al., 2023).