High-speed, long-range and wide-field OCT for in vivo 3D imaging of the oral cavity achieved by a 600 kHz swept source laser

We report a high-speed, long-range, and wide-field swept-source optical coherence tomography (SS-OCT) system aimed for imaging microstructures and microcirculations in the oral cavity. This system operates at a scan speed of 600 kHz, delivering a wide imaging field of view at 42 × 42 mm2 and a ranging distance of 36 mm. To simultaneously meet the requirements of high speed and long range, it is necessary for the k-clock trigger signal to be generated at its maximum speed, which may induce non-linear phase response in electronic devices due to the excessive k-clock frequency bandwidth, leading to phase errors. To address this challenge, we introduced a concept of electrical dispersion and a global k-clock compensation approach to improve overall performance of the imaging system. Additionally, image distortion in the wide-field imaging mode is also corrected using a method based on distortion vector maps. With this system, we demonstrate comprehensive structural and blood flow imaging of the anterior oral cavity in healthy individuals. The high-speed, long-range, and wide-field SS-OCT system opens new opportunities for comprehensive oral cavity examinations and holds promise as a reliable tool for assessing oral health conditions.

Recently, OCT has garnered significant attention for its application in imaging the oral cavity [28], owing to its capability of providing 3D microstructural and microcirculation information that can aid clinical assessments in dentistry.Notably, its non-radiative nature positions it as a safe imaging modality for oral cavity examinations in pregnant women and children [29].Moreover, OCT has demonstrated efficacy in detecting tooth demineralization and cavities by observing changes in backscattered signals from enamel and dentin [30].Even in the early stages of enamel demineralization, OCT excels in identifying the locations of white spot lesions [9,31].Additionally, OCT can be employed for diagnosing tooth fractures, manifesting as clear bright lines in OCT images [30].Enamel thickness, a reflection of tooth wear, can be quantified using OCT, offering a valuable tool for assessing enamel loss.Beyond hard tissues, OCT has shown applicability in examining soft tissues, addressing conditions such as gum diseases [32], oral mucosal lesions [33], and oral tumors [34].
Despite these successes, existing OCT systems in dentistry face challenges, primarily stemming from slow imaging speed and a limited imaging field of view (FoV).Additionally, the uneven topology of the oral cavity demands OCT to have a relatively long-ranging distance (>20 mm).
Currently, most existing OCT systems designed for dental applications have an A-scan rate typically ranging from 20 kHz to 200 kHz and a ranging distance of <12 mm, providing a FoV generally smaller than 10 × 10 mm 2 that covers only one to two teeth [11,[35][36][37].These limitations hinder the efficiency of data acquisition, impacting the workflow in imaging practice, especially in situations where a rapid overall visualization and examination of a patient's oral cavity is required.An ideal OCT system for dentistry should possess enhanced imaging capabilities, overcoming the current challenges.Therefore, the development of dental OCT should prioritize: 1) improved imaging speed, crucial for real-time clinical applications and patient comfort; 2) expanded field of view to capture comprehensive images of the entire oral cavity in a single scan, facilitating more efficient and thorough examinations; and 3) increased ranging distance to cope with uneven surface topology in the oral cavity.Previous studies have attempted to achieve a large FoV by montaging multiple scans [11,38].Although this approach mitigates the oral curvature, it would prolong the time required to complete an imaging session, significantly affecting patient compliance for imaging.
With the current rapid development of swept laser sources, it is theoretically straightforward to meet the requirements for an OCT system to achieve oral cavity imaging with a wide FoV and long-ranging distance.For example, a 1310 nm vertical-cavity surface-emitting laser (VCSEL) has been reported to achieve a cubic meter volume imaging for OCT applications but realized with the assistance of a high-speed oscilloscope with a 16 GHz analog bandwidth that sampled the data at 50 GS/s and underwent post-data processing [39].While exciting, achieving fast scanning speed, long imaging range, and a wide FoV simultaneously in practice pose a significant challenge for current OCT hardware, particularly when considering electrical signal collection and handling.This is simply because the signals with high-frequency and high-bandwidth must be dealt with, including both the OCT interference signals of interest and the k-clock signal used to trigger data acquisition.As a note, the bandwidth mentioned here refers to the frequency range of the generated interference signal.To avoid confusion, we refer to the wavelength range of the light emitted by the laser source as spectral bandwidth, whereas the frequency range of the interference signal as frequency bandwidth.
The signal handling unit, including photodetectors, transmission cables and acquisition cards, typically assumes a linear frequency response within a designated frequency bandwidth.If the signal frequency is close to or beyond the limit of this bandwidth, a non-linear phase response would occur, introducing a time delay.The situation worsens for OCT signals because the spectral broadband swept source generates an interference signal with a frequency bandwidth that is consequently converted into a broadband electrical signal, imposing different time delays for different wavelengths.This process is akin to the optical dispersion phenomenon observed in OCT imaging.Here, we refer to this electrical time delay as electrical dispersion when a high-frequency OCT signal with broad frequency bandwidth travels through the system hardware before being sampled by the data acquisition card.This electrical dispersion would inevitably affect the imaging performance of the OCT system.
Frequently when designing an OCT imaging system for most applications, for example retinal imaging, the frequency bandwidth of the hardware may be adequate to cope with the OCT interference signals due to the requirement of relatively short ranging distance where a ranging distance of 6 mm may be sufficient [40].To digitally sample this OCT signal, an efficient way is to sample it in a linearized k-space through a k-clock triggering acquisition, facilitated by an auxiliary Mach-Zender interferometer (MZI).This is because compared to the internal clock acquisition [41,42], this method eliminates the need for resampling interference spectrum signals of interest, reducing the burden on the acquisition card and simplifying post-data processing [43].In this case, to properly generate the k-clock signal to trigger the sampling, the frequency of the MZI signal is required to be at least twice that of the OCT signal of interest (per Nyquist theorem), determined by the optical delay set at the MZI, resulting in an expanded frequency bandwidth of the k-clock signal that likely goes beyond the designated frequency bandwidth for the hardware, leading to a noticeable phase shift for each sampling point (zero crossing point) due to the non-linear phase responses.Consequently, the k-clock triggered acquisition would inevitably deteriorate the system imaging performance, e.g., broadening the system point-spread function, especially in the regions with long-ranging distances.As a result, minimizing the influence of non-linear frequency phase response is crucial for developing high-speed, long-range, and wide FoV OCT imaging systems.
In this paper, we propose a high-speed and long-range solution applied to oral cavity examinations under the constraints of the existing data acquisition unit.This system operates at a scan speed of 600 kHz, providing a wide imaging field of 42 × 42 mm 2 and a ranging distance of 36 mm, effectively matching the physiological curvature of the oral cavity.To address the issue of k-clock non-linear phase errors, we propose a global k-clock calibration method to compensate for the electrical dispersion, verified by the analyses of system sensitivity roll-off curve and point spread function before and after compensation.Furthermore, this paper analyzes and compensates for spatial discrepancies resulting from long-range and wide-field imaging.In addition, when a detailed examination of a specific area of interest is desired, the system is designed with a function to achieve high-precision imaging by switching scanning protocols and objective lenses.

System setup and scanning protocol
Figure 1 shows a schematic diagram summarizing a representative SS-OCT system that was constructed, featuring high-speed, long-distance ranging, and wide FoV capabilities.The system incorporated a cutting-edge 600 kHz MEMS VCSELs swept laser source, centered at 1310 nm with 20 nm bandwidth.This configuration provided a theoretical axial resolution of approximately 41.5 µm in air (equivalent to ∼29.6 µm in tissue, assuming a refractive index of 1.4).Notably, the laser featured a built-in sweep trigger and linear k-clock, serving as the initiation signal for an external clock to synchronize the interference fringes and subsequent data acquisition.The laser output was directed through a 90/10 fiber coupler, splitting 90% power to the sample arm, and allocating 10% to the reference arm.In the sample arm, the laser was coupled to a fiber collimator, generating a collimated beam with a diameter of approximately 2.8 mm.This collimated beam was then guided to the sample using a pair of synchronized galvo scanners (6210 H, Cambridge Tech Inc. USA) triggered by the sweeping mechanism.The delivery of light to the sample was facilitated by an f-theta lens with a 100 mm focus length, achieving a lateral resolution of ∼60 µm.In the reference arm, an optical delay line was employed to align the optical path difference between the sample arm and reference arm.The backscattered light from the sample arm via a fiber circular (CIR-1310-50-APC, Thorlabs Inc., USA), and the matched reference light were coupled to a 50/50 fiber, subsequently directed to a balanced photodetector.To optimize the time delay between the trigger and k-clock and reduce the phase jitter arising from swept-to-swept variations [44], a function generator was employed to introduce a controlled delay to the swept trigger signal.The interference signal was sampled by a 1.8 GHz digitizer with a designed frequency bandwidth from DC to 0.8 GHz (ATS9360, AlazarTech Inc., USA).We carefully considered the electrical cable layout in the system design and construction, which makes the signal frequency bandwidth of the entire system dominated by the digitizer employed, meaning the system frequency bandwidth was also from DC to 0.8 GHz.
The designed ranging distance of the OCT system was 18 mm in the air, given the MZI with a 72 mm optical path difference (OPD) and the digitizer operating at single edge sampling mode.With this ranging distance, the laser operating at 600 kHz and 20 nm spectral bandwidth would generate a highest frequency OCT interference signal at maximum ranging distance that estimates to have an averaged frequency of ∼0.5 GHz and a frequency bandwidth of ∼0.4 GHz.While such a signal would experience electrical dispersion when it travels from the detector to the digitizer, the effect would be negligible because the signal frequency bandwidth of interest is well within the capacity of system frequency bandwidth of 0.8 GHz.This is likely the reason why prior SS-OCT system developments reported in the literature did not notice this electrical dispersion issue.However, it may not be true when handling the MZI interference signals to generate k-clock signals for triggering the data acquisition, which will be discussed in detail in next Section.In order to increase the ranging distance with the k-clock signal generated by the MZI with a fixed OPD of 72 mm, the digitizer was operated with a dual edge sampling (DES) mode in this study.This DES sampling leads to an actual imaging range of 36 mm from the originally designed 18 mm.
For in vivo demonstration of the system performance to image oral cavity, healthy volunteers were recruited.The study adheres to the principle of the Declaration of Helsinki and complies with the Health Insurance Portability and Accountability Act.Ethical approval was granted from the Institutional Review Board of the University of Washington and written consent for imaging was obtained from each participant.During the imaging session, the participants were instructed to rest their head on a chin rest for stability.To facilitate wide-field imaging, a mouth opener was employed to maximally expose the oral cavity to the OCT scans.The scanning protocol consisted of 1500 × 1500 (OCT imaging mode) and 1500 × 1500 × 4 A-lines (OCT angiography (OCTA) mode), covering a ∼42 × 42 mm 2 FoV.One single 3D scan took a duration of ∼5 sec and ∼19 sec to complete for OCT and OCTA mode, respectively.
While the wide-field imaging mode is advantageous for swift and comprehensive oral cavity examinations, the proposed system offers an additional flexibility for localized fine scanning when detailed observation of suspicious localized lesions is required.For this purpose, the system was designed to have an optional high-resolution scan mode.In this mode, we utilized a telecentric lens with a focal length of 36 mm, providing a lateral resolution of ∼22 µm.To stabilize subjects during imaging, a bite bar was utilized.The scanning protocol consisted of 1000 × 1000 (OCT imaging mode) and 1000 × 1000 × 4 A-lines (OCTA mode), covering a ∼7 × 7 mm 2 FoV.One single 3D scan took a duration of ∼2.2 sec and ∼8.7 sec to complete for OCT and OCTA mode, respectively.
An eigen-decomposition optical microangiography (OMAG) algorithm was applied to extract blood flow signals from four repeated B-scan images [45].The resulting blood flow signals at different depths were color-coded with the depth information from tissue surface and projected onto a two-dimensional plane, generating en-face OCTA images for visualization.

System k-clock calibration and practical considerations
While the analysis in this section is equally applicable to the OCT interference signals of interest, we will focus on the discussion of k-clock signal generation and the effect of its non-linear phase response on imaging performances.In conventional SS-OCT systems, a k-clock trigger signal is commonly employed to clock the acquisition of the interference signal in order to mitigate the mechanical sweep-to-sweep variations inherent in VCSEL's operation [43].The k-clock trigger signal is typically generated by an auxiliary MZI with a predetermined OPD, tailored to the specific requirements of the system design.Subsequently, a zero-crossing detector is employed to convert the MZI interference signal into the k-clock signal, aligning its frequency with that of the MZI signal.The mathematical representation of the MZI signal, characterized by an OPD of ∆Z MZI , is: where S(k) represents the laser source power spectrum, and k indicates the wavenumber.Consequently, the generated k-clock frequency (f clk ) can be expressed as: where δ k denotes the sweep rate of wavenumber during the laser source operation.Assuming an ideal linear wavenumber change in the time domain, δ k is given by: where ∆k is the variation in k within a sweep, f sweep is the sweep rate of the laser source, D is the sweep duty cycle (typically ranging from 0.3 to 0.7), ∆λ is the optical wavelength bandwidth, and λ c is the central wavelength.Thus, considering a linear sweep of wavenumber, the expression for an averaged f clk becomes: where the generated k-clock trigger would be a signal with its frequency proportional to MZI OPD (∆Z MZI ), spectral bandwidth (∆λ), and laser sweep rate (f sweep ).
In practical applications, achieving a perfectly linear sweep of wavenumber in the time domain poses challenges due to the mechanical tuning of the mirror in the MEMS VCSEL laser source.The inherent nonlinear wavenumber sweep over time is unavoidable [46].Assuming a deviation range of B k (i.e., the frequency bandwidth of δk), the frequency of the k-clock signal would also vary within a specific range according to Eq. ( 2), referred to as the k-clock frequency bandwidth, which is scaled with the MZI OPD that determines the OCT ranging distance.
For example, for the system described in this paper, if an OPD is set to 36 mm, each interference spectrum can theoretically generate a k-clock trigger signal with an average frequency of ∼0.5 GHz, under a linear wavenumber sweep rate with a setting of ∆λ = 20 nm, λc = 1310 nm, fsweep = 600 kHz and D = 0.5.However, the change in wavenumber is nonlinear in time domain, leading to an estimated k-clock frequency bandwidth of ∼0.4 GHz.Transmitting a k-clock signal through the electronic devices essential in the system introduces frequency dependent phase instability due to the nonlinear phase response of electronic devices, including the transmitting electrical cables.Essentially, the higher the k-clock frequency bandwidth produced by each MZI interference spectrum, the greater the phase error would be.It is important to note that this phase instability may not be noticeable when the ranging distance is relatively short, as in the conventional system setup where the ranging distance is typically less than 12 mm simply because they are within the capability of the system frequency bandwidth.
In our system design tailored for dental imaging applications involving uneven surfaces, maximizing the measurement range necessitates generating as many k-clock trigger signals as possible for each MZI interference spectrum.Consequently, the optical path difference in the MZI was set to 72 mm in our design, theoretically generating 920 trigger signals per interference spectrum (actually available 850).At this limit, each interference spectrum can theoretically generate an average frequency of ∼1 GHz k-clock trigger signals, with a frequency variation range from 0.5 to 1.3 GHz (beyond the system bandwidth of DC to 0.8 GHz).Under this circumstance the electrical dispersion would occur, leading to a non-linear phase error in the acquired interference signals, which would degrade OCT imaging performance.
To visually demonstrate the impact of k-clock phase errors on the system, reflection signals were collected at various depths using a mirror as the sample, with the implementation of a 52 dB attenuator to mitigate strong reflections.In Fig. 2(A), the measured point spread function (PSF) roll-off curve across the 36 mm ranging distance, employing the MZI-generated k-clock for triggering and DES sampling, is depicted, where an average SNR is approximately 97.7 dB.The PSF exhibits a relatively symmetrical profile with a narrow peak within a ranging distance of less than 10 mm, attributed to negligible phase errors from electrical dispersion for low-frequency interference signals.However, with the increase of the ranging distance, the PSF tends to broaden asymmetrically, accompanied by an elevation of side lobes on the left.This phenomenon primarily arises from the same phase drift but results in a larger phase difference in the high-frequency signal region.The axial resolution, assessed by the full width at half maximum (FWHM) of the PSF marked by black dots, experiences a gradual decline within the initial 16 mm of the ranging distance, followed by a rapid deterioration beyond this point.This decline significantly impacts the performance of the OCT system, particularly in capturing images beyond 16 mm, leading to diminished image quality as the ranging distance increases.In essence, the nonlinear phase effect in the generated k-clock signal begins to affect the acquired OCT signal around a ranging distance of approximately 16 mm (with an average frequency of ∼0.4 GHz).Interestingly, this value coincides with half of the frequency bandwidth of the system, which warrants further investigation.While the OCT interference signal is sampled in a linear k-space, there may still be residue of optical dispersion effect in the acquired interferogram.For this reason, numerical optical dispersion compensation methods [47] are often effective to further minimize this effect on the PSF performance.After employing the numerical optical dispersion compensation, the PSF performance (Fig. 2(B)) is improved, but only within a narrow imaging range of less than 11 mm.The system sensitivity -3 dB roll-off is improved from 7 mm to 11 mm.The SNR has increased from 97.7 dB to 98.1 dB.However, it does not yield a global improvement in the PSF performance, particularly at deeper-ranging distances, where the effect of electrical dispersion in the generated k-clock signal gradually becomes non-negligible in the sampled interference signal of interest.
To rectify the nonlinear phase shift in the k-clock signal and improve the PSF performance throughout the entire ranging distance, we introduced a polynomial equation to model and compensate for the nonlinear phase error.Considering an OCT signal from mirror in air at depth z as I(k)= S(k)cos(φ z ), the phase evolution of the interference signal can be expressed as here z represents the depth, k is the wavenumber, k 0 denotes the central wavenumber, ∆ k− k0 signifies the nonlinear k shift at the wavenumber of (k-k 0 ), a u represents the optical dispersion coefficients, and v is the highest dispersion order.The first term of the equation accounts for a linear k-clock induced phase component, and the second term represents a nonlinear k shift induced component, both of which are depth related.The last term accounts for phase distortion from system's optical dispersion, which is approximately depth independent.
To accurately model the nonlinear k shift, the optical dispersion-induced phase error is computed by the difference of the phase evolution from a mirror at two adjacent positions: where ∆z 1,2 = z 1 − z 2 .To determine the k shift, a polynomial equation was employed to approximate the change in k with where C p represents p-order coefficients, and m is the highest order.To optimize across the entire ranging distance from multiple measurements, the minimum phase residual: is utilized as the merit function to find a globally optimized nonlinear k shift vector, ∆ k− k0 .In this study, a six-order polynomial curve was employed to fit the electronic phase frequency response from three measurements.The impact of the nonlinear k shift was subsequently eliminated by resampling the interference signal with the subtracted nonlinear shift from the raw k vector.Upon applying the k-clock calibration method as described above, a notable enhancement in the PSF performance was achieved, as depicted in Fig. 3(A).The measured FWHM of the PSF exhibits a consistently flat profile across the entire ranging distance, resulting in a mean resolution of 42.3 µm (close to theoretical value of 41.5 µm).The average SNR increased from 97.7 dB to 101 dB.In contrast to the PSF roll-off without calibration (Fig. 2(A)), the -3 dB roll-off distance is expanded from 7 mm to 18 mm, while the total sensitivity remains consistently above 104.9dB throughout the -3 dB roll-off distance post-compensation.These improvements in PSF performance and sensitivity demonstrate the effectiveness of the k-clock compensation in enhancing the imaging capabilities of OCT system.When combined with optical dispersion compensation, the PSFs roll-off closely resembles its profile prior to dispersion compensation, with a very slight axial resolution improvements in longer ranging distances (Fig. 3(B)).The improvement in SNR is also no longer significant.This observation further attests to the fact that the phase error originating from k-clock nonlinearity, attributed to electrical frequency responses, is effectively addressed by the proposed compensation process.

Field of view calibration and lateral resolution assessment
Wide-field imaging inevitably introduces image distortion.This distortion may stem from non-uniformities in lenses including optical aberration or other optical elements in the system, leading to varying degrees of distortion across the imaging FoV during scanning.Additionally, differences in the focusing performance of the beam at the edges of the lens compared to the central portion can cause a distortion in the resulting images, especially in the periphery.To accurately estimate the FoV and calibrate this image distortion, we employed a target with a grid pattern of circular dots as the calibration reference.This target featured black dots with a radius of 50 µm, spaced 2 mm apart on a white background.Precise alignment with the OCT system was achieved using a translational stage and a kinematic mount, ensuring the target's center was perpendicular to the output probing beam.The total size of the target was ∼50 × 50 mm 2 .A 3D scan consisting of 1500 × 1500 A-lines was performed using the designed system described in the last Section to generate an enface image of the target using maximum intensity projection along the depth direction (Fig. 4(A)).The estimated FoV was approximately 42 × 42 mm 2 .The dot-grid pattern locations (red dots in Fig. 4(A)) were detected, and a distortion vector map was calculated by comparing them to the prior-known pattern.
To better visualize the FoV distortion, we calculated an absolute FoV distortion map based on the distortion vector map (Fig. 4(C)).R1 and R2 represent the circle regions with diameters of 20 mm and 30 mm, respectively, and R represents the entire FoV.The peripheral regions exhibited larger distortion, attributed to the relatively significant optical aberration of the f-theta lens at large angles.Most of the FoV regions displayed a distortion less than 200 µm, with an average distortion of 104.9 µm across the entire FoV.To correct this distortion, a post-resampling of the enface image was therefore performed using a compensation distortion map derived from even interpolation of the distortion vector map, resulting in Fig. 4(D) where most of the distortions are seen to be corrected.Table 1 summarizes the statistics of FoV distortion in different regions of interest before and after correction.The FoV distortion correction would help provide a more accurate and reliable imaging in wide-field imaging mode.The practical lateral resolution of the proposed wide field OCT system was evaluated using a knife-edge technique.The blade was fixed with its edge-oriented perpendicular to the OCT beam.Two imaging modes, namely wide-field imaging and high-resolution imaging were employed to capture images of the blade.Figures 5(A) and 5(B) illustrate the maximum intensity projection images of the blade generated by both imaging modes, respectively.The zoomed-in images on the right side of Figs.5(A 2 provides a summary of the system parameters under both the imaging modes.

Experimental in vivo OCT imaging of the oral cavity
In vivo imaging of the human oral cavity was conducted using the proposed OCT system.Figure 6(A) showcases a 3D volume rendering of the anterior oral cavity using the developed prototype in the wide field imaging mode, offering an expansive FoV of approximately 42 × 42 mm 2 .This image provides detailed morphological information of the anterior oral cavity, encompassing fourteen teeth, labial frenum, and gingiva with one single scan.Generally, OCT proves effective in distinguishing enamel caries by detecting the enhanced backscattering signal resulting from microporous structures formed by localized tooth demineralization.With light penetration within the enamel region, the depth-resolved information unveils a distinct map of enamel health.While it requires clinical expertise to confirm, the cross-sectional view of the enamel corresponding to the red arrow region of Fig. 6 highlights a suspicious enamel carious lesion, exhibiting increased backscattering in Figs.6(B) and 6(C) (red arrows).
Leveraging the extensive imaging range offered by the proposed OCT system, en-face views at depths of 1000 µm (Fig. 6 G) and 2000 µm (Fig. 6 H) present a comprehensive distribution map of dental wear, effectively revealing suspicious enamel carious lesions that were not as distinct in the volume rendering (Fig. 6(A)) or superficial slicing (Fig. 6(F)).This detailed insight may be clinically useful to aid in the early diagnosis of potential enamel caries and offer a precise assessment of cavitated enamel caries.
Furthermore, the oral epithelium, with its high scattering property, results in increased intensity within the superficial gingival region.As depth increases, the en-face view distinctly delineates boundaries among various gingival units (Figs.6(G)-(H)).Notably, the attached gingiva region With the capabilities of OCTA to capture dynamic blood flow, we also conducted in vivo OCTA imaging of the oral cavity to visualize the vascular networks.Figure 7(A) illustrates a depth-color encoded 3D projection of the anterior oral cavity, obtained through OCTA under wide-field imaging mode, covering a depth of 2 mm, which vividly depicts a dense distribution of capillaries, small connective vessels, and larger blood vessels across the wide FoV of the human gingiva.Notably, dense capillaries and small connective vessels dominate the superficial layers of the gingival tissue, while larger blood vessels are situated in deeper regions.The en-face view at a depth of 200 µm (Fig. 7(B)) accentuates the distribution of capillaries and small connective blood vessels.Despite capillary sizes exceeding the lateral resolution achievable in wide-field mode, the lower density of large blood vessels and the distinct, well-spaced arrangement of capillaries facilitate effective imaging of capillary networks within the free and attached gingiva region, as shown in the inset of Fig. 7(B).
A detailed analysis at different depths reveals notable variations in the distribution of blood vessels across various gingival regions.At a depth of 200 µm with respect to the surface, capillaries and small connective blood vessels are prominent, particularly within the attached gingiva and free gingiva regions.Remarkably, the attached gingival region displays a lower density of large blood vessels in its superficial layers (Figs.7(B)-(C)), with a slight increase in deeper layers (Fig. 7(D)).In contrast, the free gingiva region features capillaries and small connective blood vessels predominantly in the superficial layers, transitioning to larger blood vessels at deeper layers.
Conversely, the alveolar mucosa region displays a higher prevalence of connective blood vessels and larger blood vessels, with the larger vessels predominantly located in deeper layers (Figs.7(B)-(D)).Additionally, en-face views illustrate the progressive merging of capillaries and small connective blood vessels into larger vessels as one moves from superficial to deep layers, connecting with larger blood vessels from the superficial to intermediate layers.Notably, the superficial blood perfusion in the region marked by white dots (Fig. 7(B)) is subdued, attributed to the tension exerted by the frenulum when the oral cavity is exposed to the probing beam through a mouth opener during imaging, resulting in some blood to be squeezed out.
Figure 7(E) zooms into a specific region of interest from Fig. 7(A) (indicated by a rectangular box) obtained with the high-resolution imaging mode.This mode, distinguished by its heightened detail, is evident in the enface maps at depths of 200 µm, 450 µm, and 750 µm, respectively (Figs. 7(F)-(H)), offering a nuanced map of vessel network distribution.At a depth of 200 µm, dense capillaries are prevalent across the entire ROI, encompassing the free gingiva, attached gingiva, and alveolar mucosa regions.These results accentuate an increased distribution of connective vessels and large vessels at 450 µm, followed by a decline in connective vessels at 750 µm.Meanwhile, the prevalence of large vessels continues to expand, consistent with our observations in wide-field imaging.Moreover, high-resolution imaging reveals local features such as plaque and vessel loops (depicted by white arrows in Fig. 7(E)), which can be crucial for precise diagnosis and treatment of oral diseases.

Discussions and conclusion
In this study, we presented an SS-OCT system suitable for comprehensive oral examinations, featured with an imaging speed of 600kHz, a field of view of 42 mm × 42 mm and a ranging distance of 36 mm.Since the increase of the optical path difference in a MZI progressively increases the interference signal frequency bandwidth that may challenge the system bandwidth, leading to an electrical dispersion that presents in the generated k-clock signal, we proposed a method to compensate for this dispersion after the OCT interference signal of interest is digitized by the data acquisition card.This is the unique contribution of this study to the development of SS-OCT system for wide field imaging applications.We have shown that such compensation is effective and necessary when imagining a target with complex topology features, for example in the imaging of oral cavity.
While recent advancements have produced swept sources with multi-MHz line speeds (up to 40 MHz), the imaging speed of SS-OCT has not kept pace with the development of laser sources.Beyond scanning speeds, data acquisition speed is another limiting factor.If an external-triggered k-clock acquisition method is chosen, the generation speed of the k-clock signal becomes a limiting factor, creating a competitive relationship among imaging speed, imaging range, and laser source bandwidth.Enhancing either of these parameters inevitably diminishes the third.All three parameters are required to consider synergistically when designing a SS-OCT imaging system that meets the purpose for specific applications.
In this paper, we designed an optimal parameter configuration scheme for oral examinations.We chose to sacrifice a portion of axial resolution (i.e., spectral bandwidth) to maximize the requirements for imaging speed and imaging range.This is mainly because there are competing factors among various parameters of the SS-OCT system under the constraints of limited performance of the existing hardware, including digital acquisition card.Considering the necessity for rapid and comprehensive oral examinations, which demand OCT systems to possess both a high scanning speed and an extensive imaging range, the only aspect that can be adjusted is the axial resolution (spectral bandwidth).In reality, based on the information available in the literature about physiological and pathological teeth and gums, it is found that the structure of oral tissues is not as complex as that of retinal tissues.The resolution requirements are relatively modest.The resolution of commonly used X-ray imaging in dentistry is typically between 75 and 200 µm [48].In contrast, the axial resolution of commonly used SS-OCT is generally much higher than that.Therefore, there is considerable room for adjustment in the axial resolution of SS-OCT.Thus, while meeting the requirements for axial resolution in oral imaging, it is indeed possible to increase other performance parameters of SS-OCT by reducing spectral bandwidth.
To maximize imaging speed and ranging distance, the k-clock trigger signal must maintain its maximum output frequency.The proposed k-clock calibration strategy effectively mitigates the nonlinearity of electronic devices in generating high-frequency k-clock signals.It should be noted that an alternative approach would be to employ more advanced digitizers with a broader bandwidth capability to mitigate nonlinear phase errors during OCT signal sampling.However, such an approach would inevitably increase the system cost, and it also does not address electrical dispersion from the k-clock transmission process.
The choice of the laser source spectral bandwidth was deliberate; a 20 nm bandwidth can provide ∼29.6 µm axial resolution in tissue, which is significantly higher than the resolution of X-rays commonly used in dentistry.It is fully capable of handling tasks related to tooth or soft tissue detection.It's worth noting that the narrowband spectrum is less affected by k-clock nonlinearity than the broadband spectrum.This is one of the reasons why a narrowband laser source was chosen in this study.The imaging field of 42 × 42 mm 2 can cover 14 human teeth, meaning the entire oral cavity can be covered in two scans.The acquisition time for a single structural image is only 4.75 seconds, allowing the majority of patients to remain stable in such a short time.The depth-resolved en-face and cross-sectional structure images enabled the observation of general oral cavity features, including the ability to distinguish enamel health, detect carious lesions, and visualize distinct boundaries of different types of gingival.With OCTA capabilities, we successfully visualized the blood perfusion by delivery of a macroscopic distribution map of capillaries, small connective blood vessels, and large blood vessels.When precise examination of local lesions is needed, the system can provide excellent high-precision resolution by switching objective lens and scanning protocols.
The system presented in this study is not limited to dental imaging; it is equally applicable to other fields with demanding requirements for imaging speed, field of view and ranging distance, such as dermatology, orthopedic surgery, or intraoperative applications in general surgery, and so on.Although the proposed system can achieve long-range imaging, it has limitations.At the maximum imaging distance, image resolution severely decreases due to defocusing.Combining this system with a tunable lens might be a solution to maximize image quality throughout the entire oral cavity [49].Moreover, while the proposed system can quickly perform the scans, it results in larger data sizes and longer computational times for reconstructing 3D OCT images.This challenge underscores the need for efficient data processing solutions.Recent advancements in compressed sensing algorithms with optimized sampling strategies offer promising solutions.These algorithms can reconstruct 3D OCT image volumes from as little as 10% of the original data, significantly reducing both scan time and computational load [50].
In conclusion, we have demonstrated a high-speed and long-range wide-field OCT/OCTA system for oral cavity imaging based on a 600kHz MEMS-VCSEL swept laser source, enabling an expanded FOV of 42 × 42 mm 2 and an extended imaging range of 36 mm for OCT and OCTA imaging without compromising scanning time.We introduced a concept of electrical dispersion and a global k-clock compensation approach to improve overall performance of the proposed imaging system.We demonstrated that the system provides visualization of both macroscopic and microscopic oral environments, offering a potential non-invasive approach for diagnosing and monitoring oral diseases.We expect the proposed system will provide new opportunities for wide-field imaging applications in dentistry and other clinical fields.Acknowledgments.The authors thank Drs Nate Bramham, Christopher Burger and Vijaysekhar Jayaraman from Praevium Inc. for providing technical support on the swept laser source used in this study.
Disclosures.The authors declare that there are no conflicts of interest related to this article.

Fig. 2 .
Fig. 2.The system sensitivity roll-off and axial resolution measurements of the 600 kHz SSOCT system: (A) without optical dispersion compensation and (B) with optical dispersion compensation.Shown are the point spread functions (PSFs) measured at each ranging distance that are plotted with different colors, and the dash lines with circle markers show the corresponding axial resolution measured by the FWHM of the PSFs using gaussian fitting.

Fig. 3 .
Fig. 3.The system sensitivity roll-off and axial resolution measurements of the 600 kHz SSOCT system with k-clock calibration: (A) without optical dispersion compensation and (B) with optical dispersion compensation.The system point spread functions (PSFs) measured at each optical delay are plotted with different color, and the dash lines with circle markers show the corresponding axial resolution measured by the FWHM of the PSFs using gaussian fitting.

Fig. 4 .
Fig. 4. Calibration of field of view distortion.(A) En-face projection image generated from a circle-grid target with a field of view (FoV) of 42 × 42 mm 2 .Red dots indicate the detected locations of the circle-grid pattern.(B) Distortion vector map calculated from (A). (C) Absolute distortion map generated from (B), where R1 and R2 represent the circle regions with diameters of 20 mm and 30 mm, respectively, and R represents the entire FoV.(D) The absolute distortion map after correction.The scale bars for all images are 3 mm, unless otherwise indicated.
) and 5(B) provide a closer view of the image region marked by the blue box.The hollow circles in Figs.5(C) and 5(D) indicate the signal intensity at the cyan line positions corresponding to Figs. 5(A) and 5(B).The estimated focal spot size was approximately 56.7 µm.The red curves represent the intensity fitting curves obtained through a point spread function analysis.The derivatives of these curves are depicted by the yellow dashed lines in Figs.5(C) and 5(D).The lateral resolution of the system was then determined by measuring the FWHM of the derivatives.Multiple measurements yielded a spot size of 60.4 ± 5.2 µm in wide-field imaging mode and 22.3 ± 1.6 µm in local imaging mode.Table

Fig. 5 .
Fig. 5. Assessment of the lateral resolution of the proposed HLW-OCT using knife-edge.A. Maximum intensity projection en-face image in wide-field imaging mode with scale bars denoting 1 mm.B. Maximum intensity projection en-face image in local imaging mode.The zoomed-in images on the right side of Figs.5(A) and 5(B) are magnified images of the blue box positions.C and D show the intensity fitting curves and corresponding derivative curves of the blue solid line positions in A and B, respectively.

Fig. 6 .
Fig. 6.OCT structural imaging shows the oral anatomy in a wide field of view at 42 mm x 42 mm.(A) The 3D renders of anterior oral cavity; (B-E) Cross-sectional B-frames corresponding to the blue (B-C) and green (D-E) dash lines in (A); (F-H) En-face OCT slicing views at the depth of 100 µm, 1000 µm and 2000 µm with respect to the tissue surface.The scale bars for all the images (A-H) are 3 mm.DEJ: dentin enamel junction; Red arrows: suspicious enamel carious region; Blue arrows: DEJ; White arrows: gingiva sulcus.

Fig. 7 .
Fig. 7.In vivo OCTA imaging of microcirculations within the gingival tissue of the oral cavity.(A) Depth-color encoded OCTA projection image over a 2 mm depth within the oral cavity.(B-D) En-face OCTA views at various depths: 200 µm, 450 µm, and 750 µm, with respect to the tissue surface.Insets in (B) and (D) are zoomed-in views of the area marked by dashed rectangles, respectively.(E) Depth-color encoded OCTA (over 1 mm) corresponding to the white rectangular area in (A) using the high-resolution imaging mode.(F-H) En-face OCTA views at various depths corresponding to the area marked by dashed rectangle in (A) using the high-resolution imaging mode: 200 µm, 450 µm, and 750 µm, with respect to the tissue surface.The scale bars for images (A-D) are 3 mm.

Funding.
Washington Research Foundation; National Institute of Dental and Craniofacial Research (R44DE031194).