A handheld skin optical coherence tomography system and method based on continuous wide-field spatial-wavenumber encoding, field-matched reference wavefront and near 1550 nm light source

Abstract

The application provides a handheld skin optical coherence tomography system and method based on continuous wide-field spatial-wave number coding, field matching reference wave front and near 1550nm light source, which comprises a near 1550nm broadband low coherence light source, an interference module, a continuous wide-field sample arm, a field matching reference arm, a reference reflecting surface and a field matching reference wave front element, so that the difference between the optical path of the sample arm corresponding to different object space units and the optical path of the reference arm is kept within the effective coherence range of the light source, a metasurface spatial wave number coding camera, which comprises a metasurface coding layer and a surface array detector having a response to the near 1550nm wave band, is used for coding the interference light of multiple object space units according to the spatial position and wave number and outputting two-dimensional coding readings during one exposure, and a reconstruction processor is used for recovering the spatial wave number interference data of each object space unit from the two-dimensional coding readings and performing Fourier transform along the wave number direction to generate the depth structure data of the skin tissue.

Description

Technical Field

[0001] This invention relates to the field of biomedical optical imaging, and in particular to a handheld skin optical coherence tomography system and method based on continuous wide-field space-wavenumber coding, field-matched reference wavefront and near 1550nm light source. Background Technology

[0002] Traditional spectral domain or swept frequency OCT usually changes the position of the beam on the sample surface during the acquisition process through a galvanometer, MEMS scanner or other lateral movement mechanism, and then stitches the A scan or B scan in time sequence. When the detection area is in a handheld state or the subject is in a micro-movement state, the lateral scan and tissue movement will cause a spatiotemporal mismatch. On the other hand, traditional spectrometers, gratings, linear array detectors and their optical paths usually occupy a certain volume, which limits the miniaturization of the whole machine.

[0003] Meanwhile, there is another type of single-snapshot OCT or full-field OCT scheme in the existing technology, but it usually relies on sparse multi-beam sampling, spatial light modulators, line fields or dedicated optical mapping structures. For wide-field skin detection, how to maintain lateral spatial resolution, spectral resolution, interference phase stability and field matching of handheld optical path under the limited pixels of area array camera still requires specific opto-mechanical-electronic co-design. Summary of the Invention

[0004] The purpose of this invention is to provide a handheld skin optical coherence tomography system and method based on continuous wide-field space-wavenumber coding, field-matched reference wavefront and near 1550nm light source to solve the above-mentioned technical problems.

[0005] To address the aforementioned technical problems, this invention provides a handheld skin optical coherence tomography system based on continuous wide-field spatial-wavenumber coding, field-matched reference wavefront, and a near-1550nm light source, comprising: A broadband, low-coherence light source with a wavelength of nearly 1550nm; Interference module; A continuous wide-field sample arm is used to illuminate the sample light onto the detection field of view and divide the detection field of view into multiple two-dimensional adjacent object space units; The field-matching reference arm, including a reference reflecting surface and a field-matching reference wavefront element, is used to generate a reference wavefront that varies with the position of the object space unit, so that the optical path difference between the sample arm and the reference arm corresponding to different object space units is kept within the effective coherence range of the light source. A metasurface spatial wavenumber encoded camera, comprising a metasurface encoding layer and an array detector responsive to the near 1550 nm band, is used to encode the interference light of multiple object space units according to their spatial position and wavenumber during a single exposure and output a two-dimensional encoded reading. The reconstruction processor is used to recover spatial wavenumber interferometric data of each object space unit from two-dimensional coded readings based on a calibration transfer matrix including field-matched reference wavefront information, and to perform Fourier transform along the wavenumber direction to generate deep structural data of skin tissue.

[0006] Furthermore, the continuous wide-field sample arm maps the object space units within the detection field of view to multiple sub-regions of the area array detector or overlapping point diffusion responses that can be distinguished by the calibration transfer matrix, in a two-dimensional adjacency relationship. This allows each object space unit to obtain multiple independent wavenumber observations in the same exposure, which are used to recover the interference spectrum of the corresponding object space unit.

[0007] Furthermore, the total number of effective wavenumber channels is obtained by multiplying the number of spatial units in the detection field of view in the x direction, the number of spatial units in the detection field of view in the y direction, and the number of effective wavenumber channels to be recovered in each spatial unit. In the direct allocation mode, the number of effective pixels used for encoding readout by the area array detector is greater than or equal to the total number of effective wavenumber channels. In the compressed coding mode, the number of effective pixels used for encoding readout by the area array detector is less than the total number of effective wavenumber channels.

[0008] Furthermore, the reconstruction processor performs at least two of the following on the recovered spatial wavenumber interferometric data: wavenumber linearization, reference wavefront phase correction, dispersion compensation, background removal, depth roll-off correction, and window function processing, and then performs Fourier transform. Based on multiple wavenumber channels obtained from a single exposure, it reconstructs depth signals for the full band, surface-preferred subband, depth-preferred subband, polarization channel, or spectral ratio channel, respectively. After comparing the depth signal with the noise ratio, it selects the effective subband.

[0009] Furthermore, the field matching reference wavefront element is one or more combinations of curved reference mirror, reference metal lens, metasurface phase plate, liquid crystal phase modulator, and adjustable deformable mirror.

[0010] Furthermore, after compensation by the field-matched reference wavefront element, the peak-to-valley value of the optical path difference in the detected field of view is no greater than 5 μm, and the calibrated static phase residual RMS is no greater than π / 8 and the upper limit is no greater than π / 4.

[0011] Furthermore, the metasurface coding layer is one or more combinations of dielectric nanopillars, nanofins, nanopores, supergratings, and polarization-selective nanostructures, used to perform preset joint coding of the wavenumber, lateral spatial position, polarization, and phase of the input light.

[0012] Furthermore, the center wavelength of the near 1550nm broadband low coherence light source is between 1520nm and 1580nm, and the spectral range is between 1500nm and 1600nm. The light source is a superluminescent diode, an amplified spontaneous emission light source, a supercontinuum light source after bandpass filtering, or a low coherence laser.

[0013] Furthermore, several of the following components are housed within the handheld probe: a near 1550nm broadband low-coherence light source, an interference module, a continuous wide-field sample arm, a field-matching reference arm, a metasurface spatial wavenumber encoding camera, a reconstruction processor, a battery, and a communication module; the handheld probe includes at least two of the following: a transparent detection window, a distance sensor, a pressure sensor, an inertial measurement unit, a temperature sensor, a visible light-assisted camera, and an optical power monitor.

[0014] This invention also discloses a method for a handheld skin optical coherence tomography system based on continuous wide-field spatial-wavenumber coding, field-matched reference wavefront, and a near 1550 nm light source, comprising the following steps: Step 1: Form a continuous wide-field detection field on the skin surface and divide the detection field into multiple two-dimensional adjacent object space units; Step 2: Use the field-matching reference arm to form a reference wavefront related to the position of the object space unit, so that the optical path difference between the sample arm and the reference arm corresponding to different object space units is kept within the effective coherence range. Step 3: During a single area array exposure, acquire two-dimensional coded readings of multiple object space units within the detection field of view using a metasurface spatial wavenumber coded camera; Step 4: Recover the wavenumber domain interferometric data of each object space unit based on the calibration transfer matrix, and perform field-dependent reference phase correction, wavenumber linearization, and dispersion compensation. Step 5: Perform Fourier transform along the wavenumber direction to output one or more of the following: tomographic image of skin tissue, three-dimensional structural data, epidermal thickness, depth of the epidermal-dermal junction, depth of the lesion boundary, and changes before and after treatment.

[0015] The beneficial effects of this invention are as follows: 1. Continuous wide-field single-shot area acquisition: Reduces the lateral scanning time within a single field of view, and reduces spatial-temporal mismatch caused by skin micro-movements and hand tremors.

[0016] 2. Field-matched reference wavefront: This ensures that multiple adjacent spatial cells within a wide field are within an effective coherence gate and a calibrable reference phase range.

[0017] 3. Metasurface spatial wavenumber coding: Replacing large spectrometers or linear array beam-splitting architectures with thin coding layers and area array detectors, suitable for miniaturizing optical heads.

[0018] 4. High degree of freedom in lateral design: Lateral sampling is not determined by the galvanometer swing amplitude, scanning trajectory and motion mechanism bandwidth, but by the objective lens, spatial unit mapping, metasurface coding and area array pixels.

[0019] 5. Optimizable up to 1550 nm: The same frame of spectral data supports post-processing of different subbands at the surface and depth. Combined with SNR and safety feedback, it avoids simplifying wavelength selection to unverified conclusions of deeper penetration.

[0020] 6. Handheld and integrated: The light source, interferometer, reference wavefront, metasurface camera and processing module can all be built-in, making it convenient for use in outpatient, bedside, community and follow-up scenarios. Attached Figure Description

[0021] Figure 1 This is the overall system structure diagram of the present invention; Figure 2 This is a detailed optical path diagram of the near 1550 nm wide-field Linnik-type OCT of the present invention; Figure 3 This is a schematic diagram of the field matching reference wavefront and calibration compensation principle of the present invention; Figure 4 This is an example diagram showing the correspondence between the object space unit, B(i,j) pixel block, wavenumber channel and H matrix of the present invention; Figure 5 This is a single-frame data mapping diagram of the metasurface space-wavenumber encoded camera of the present invention; Figure 6 This invention relates to a single-array acquisition, OCT reconstruction, and optional inter-frame feedback diagram; Figure 7 This is a cross-sectional view and module partition diagram of the fully handheld probe of the present invention; Figure 8 This is a schematic diagram of the exposure safety gating, quality judgment and failure protection process of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the scope of protection of the present invention.

[0023] Those skilled in the art should understand that, in the disclosure of this invention, the terms "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the above terms should not be construed as limiting this invention.

[0024] It is understood that the term "a" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of an element can be one, while in another embodiment, the number of the element can be multiple, and the term "a" should not be understood as a limitation on the number.

[0025] like Figures 1-8 The present invention provides a handheld skin optical coherence tomography system based on continuous wide-field spatial-wavenumber coding, field-matched reference wavefront and near 1550nm light source, including a near 1550nm broadband low coherence light source, an interference module, a continuous wide-field sample arm, a field-matched reference arm, a metasurface spatial wavenumber coding camera and a reconstruction processor.

[0026] Among them, the near 1550nm broadband low-coherence light source is a broadband low-coherence light source with a center wavelength between 1520nm and 1580nm, preferably around 1550nm, and a spectral range of 1500nm to 1600nm. The near 1550nm broadband low-coherence light source can be a broadband superluminescent diode (SLD), an amplified spontaneous emission (ASE) light source, a supercontinuum light source after bandpass filtering, or a low-coherence laser. The system belongs to the spectral domain / Fourier domain OCT, which acquires interference information containing multiple wavenumber channels in a single camera exposure, and then performs Fourier transform on the recovered wavenumber domain signal. The selection of the near 1550nm band is based on a comprehensive trade-off of skin scattering, tissue water absorption, detector response, light source availability, system noise, and safety constraints.

[0027] It is worth mentioning that the conventional baseline system is about 1305nm, which is a mature working window for swept-frequency skin OCT; the 1500-1600nm broadband low-coherence light in this scheme is not just a shift of the center wavelength, but is designed together with InGaAs array response, communication band light source / coating, silicon or amorphous silicon metasurface material platform and space-wavenumber snapshot encoding.

[0028] The center wavelength λ0 can be 1520-1580 nm, preferably 1550 nm; the FWHM bandwidth Δλ can be 80-120 nm, preferably 100 nm; the depth SNR, surface saturation and roll-off are evaluated in 5-10 nm subbands within the 1500-1600 nm range, and then the effective subbands are selected from the same frame data.

[0029] The continuous wide-field sample arm includes a collimator, beam expander, polarizer, telecentric or near-telecentric objective lens, optional field curvature correction lens, detection window, and optional adjustable focusing element. The sample arm forms a continuous or quasi-continuous illumination area for the incident light, and the two-dimensional adjacent object space units within the detection field of view are mapped to the camera encoding area by the imaging objective lens. In this scheme, "continuous wide field" means that within a detection field of view, the object space units cover the main body of the field of view in a two-dimensional adjacent relationship, and there is no temporal gap between adjacent space units that can only be obtained by the scanning mechanism. It does not exclude the sampling interval, guard boundary, or post-processing interpolation between area array pixels, nor is it equivalent to a beam array consisting of only a few non-adjacent sparse spots.

[0030] For larger skin areas, the operator can reposition the probe and stitch together multiple independent field-of-view snapshots; the repositioning is a field-of-view manual or assisted positioning that does not change the acquisition method of single area array exposure within a single field of view and no point-by-point / line-by-line lateral scanning.

[0031] The field-matching reference arm includes a reference reflecting surface and a field-matching reference wavefront element, which is used to generate a reference wavefront related to the field of view of the sample arm, so that the difference between the sample optical path and the reference optical path at different field positions is kept within the effective coherence range of the low coherence source; the field-matching element can be one or more combinations of curved reference mirror, reference metal lens, metasurface phase plate, liquid crystal phase modulator, and adjustable deformable mirror.

[0032] The metasurface spatial wavenumber encoded camera includes a metasurface encoding layer, an optional relay / microlens layer, a near-infrared array detector, and a camera readout circuit. The metasurface encoding layer performs a preset joint encoding of the wavenumber, lateral spatial position, polarization, or phase of the input light, and the array detector acquires a two-dimensional encoded reading in a single exposure. The array detector is preferably an InGaAs array, but extended InGaAs, InGaAs / InP, HgCdTe, or other array devices that respond to near 1550 nm can also be used.

[0033] Furthermore, the object space unit S(i,j) is the lateral position of the skin's continuous field of view defined by the sampling grid; the area subarray region B(i,j) is the coded response support domain of that position on the detector; the pixel (u,v) is the actual charge readout unit; the wavenumber channel kp is the spectral basis function after resampling the 1500-1600nm spectral band at equal intervals of k=2π / λ; in direct allocation, one pixel or a group of corrected pixels corresponds to one kp; in compressed coding, there is no unique correspondence between a pixel and a wavenumber, but the response weights of the kp component of S(i,j) on multiple pixels are described by the nth column of H; the channel in this scheme refers to the space-wavenumber coding basis of the same continuous wide-field return light.

[0034] The reconstruction processor (decoding module) recovers the spatial-wavenumber interference data of each object space unit from the two-dimensional coded readings based on the calibration transfer matrix, and performs OCT reconstruction to generate the depth structure data of the skin tissue; the control module is responsible for exposure gating, quality judgment and parameter feedback.

[0035] In one embodiment of this scheme, a wide-field Linnik-type implementation is preferred. The near 1550nm low coherence light enters the unpolarized beam splitter BS1 after passing through an optical isolator, a variable optical attenuator, a polarization controller, a collimator, a beam expander, and a shimming element. The beam splitter can be a near-infrared unpolarized cubic or planar beam splitter with a 5-10mm aperture, a splitting ratio of 50:50 at the center wavelength of 1550nm, a tolerance of ±5%, and an insertion loss of no more than 1dB. To improve the utilization rate of weak sample backlight, a 70:30 or 30:70 splitting ratio can also be used, and the reference light and sample light at the array are made to meet the full-well capacity and saturation requirements through a variable attenuator on the reference arm.

[0036] The sample arm is equipped with a telecentric or near-telecentric objective lens and a transparent detection window to form a continuous wide-field illumination on the skin surface and coaxially collect the sample return light. The reference arm is equipped with a reference objective lens with the same or equivalent focal length, numerical aperture, and aberration characteristics as the sample objective lens, and a reference mirror, curved phase plate, or reference metal lens is set on the reference image plane or conjugate pupil plane to form a field-matched reference return light. The two return lights are combined at BS1 and output from the detection port. After passing through the relay imaging and polarization analysis module, the skin image plane is conjugated to the metasurface space-wavenumber encoding layer, and then the two-dimensional encoding reading is recorded by a single exposure of the InGaAs area array. The field stop of the detection port is used to suppress crosstalk between adjacent encoding regions, and the polarization analyzer is used to adjust the interference visibility of the sample return light and the reference return light. In this embodiment, no galvanometer, MEMS scanning mirror, or mechanical lateral scanning mechanism is set in a single detection field of view.

[0037] It is worth mentioning that the field-dependent optical path difference between the wide-field sample arm and the reference arm affects the reconstructability of the surface interference signal. For the object space position (x,y), the optical path of the sample arm is denoted as Ls(x,y), and the optical path of the reference arm is denoted as Lr(x,y). The field-dependent optical path difference is ΔL(x,y)=Ls(x,y)-Lr(x,y). The field-matched reference wavefront element ensures that most spatial units in the detection field of view satisfy max|ΔL(x,y)|≤Lc / 2, where Lc is the effective coherence length of the light source. The remaining ΔL(x,y) is written into the calibration transfer matrix as the reference phase term φref(i,j,k) or the depth coordinate offset Δz(i,j). Phase and depth coordinate corrections are performed during the reconstruction stage to achieve the combined effect of structural compensation and digital compensation, avoiding the edge field of view from leaving the coherence gate due to focusing only on the center field of view.

[0038] As a set of engineering parameters: when the center wavelength λ0 of the light source is 1550 nm and the FWHM bandwidth Δλ is 80–120 nm, according to the definition of the Gaussian spectrum axial envelope FWHM, the effective coherence length in air is Lc≈0.441λ0² / Δλ, approximately 8.8–13.2 μm, and Lc / 2 is approximately 4.4–6.6 μm; when Δλ=100 nm, Lc is approximately 10.6 μm, and Lc / 2 is approximately 5.3 μm; if the tissue refractive index is 1.38, The corresponding axial FWHM within the tissue is approximately 6.4-9.6 μm; the effective focal length of the near-to-far center objective is approximately 25 mm, the object-side numerical aperture is approximately 0.05, the working distance is approximately 12 mm, and the detection field of view is 6 mm × 6 mm; the uncompensated equivalent image plane field curvature radius is taken as 50-100 mm, and for a position at a distance r from the center of the field of view, the first-path field curvature arc vector is approximately s(r) = r² / (2R), and the change in the equivalent optical path length of the reflected round-trip air is approximately 2s(r), as shown in Table 1 below.

[0039] Table 1

[0040] The phase accuracy requirements for field matching are as follows: the peak-to-valley value of the optical path difference across the entire field of view after compensation by the field matching element should not exceed 5 μm; for the calibrated static phase residual, the RMS should not exceed π / 8, and the acceptable upper limit should not exceed π / 4. Based on δφ=2π·δ(OPD) / λ0, the corresponding optical path difference RMS at 1550nm is approximately 97nm and 194nm, respectively; if caused solely by the surface shape error of the two-way reflector, the corresponding surface height RMS is approximately 48nm and 97nm; absolute machining phase errors can be eliminated through calibration of the transfer matrix and reference phase term, but residual errors after repeated positioning, temperature drift, and vibration should meet the above indicators.

[0041] The field-matching reference arm can also be implemented in several other ways: the first is a quasi-common or Mirau-type structure, in which the reference reflecting surface and the sample detection surface have the same or approximately the same field of view; the second is a non-common Michelson or Mach-Zehnder structure, which uses a reference metal lens or curved mirror to match the field curvature of the sample arm; the third is an adjustable reference wavefront structure, which uses a small number of calibration exposures to estimate ΔL(x,y), and then adjusts the reference phase or compensates in the digital domain.

[0042] The metasurface coding layer is composed of one or more combinations of dielectric nanopillars, nanofins, nanopores, supergratings, and polarization-selective nanostructures. As an alternative embodiment, amorphous silicon cylindrical nanopillars on a fused silica substrate are used. The refractive index of the amorphous silicon is around 1550 nm, which can be used as an initial simulation input of about 3.4. The height of the nanopillars is 900-1000 nm, preferably 940 nm. The period of the hexagonal or square lattice is 720-850 nm, preferably 800 nm. The column diameter is within 200-650 nm, and a cell library is established with a step size of 5-10 nm. The minimum linewidth and adjacent spacing are not less than 100 nm. By changing the column diameter, a transmission phase coverage close to 2π can be obtained. The cell library is arranged into superpixels according to spatial position. Each superpixel can contain 4×4, 8×8, or other numbers of different response units.

[0043] Furthermore, the scanning range of the above-mentioned metasurface coding layer in the 1500-1600nm band and the corresponding values ​​to be derived in this scheme are shown in Table 2 below.

[0044] Table 2

[0045] The data layout is as follows: a continuous spatial unit (i,j) in the object space is mapped onto the corresponding sub-region B(i,j) on the area array via the imaging optical path. Multiple wavenumber channels k1, k2…kP are distributed within B(i,j) along one or two directions. Spatial multiplexing coding or block sparse coding can also be used to partially overlap the coding patterns of adjacent spatial units, and crosstalk is suppressed by inversion using a calibration matrix. As an example of direct allocation: a 512×512 effective pixel ROI is used, with object space units Nx=64 and Ny=64. Each B(i,j) occupies 8×8 pixels and corresponds to P=64 wavenumber channels; let the local rows and columns r,c∈{0,…,7}, then the detector coordinates u=8i+r, v=8j+c, and the channel number p=8r+c+1; B(0,0) corresponds to u,v=0…7, B(0,1) corresponds to u=0…7, v=8…15, and B(1,0) corresponds to u=8…15, v=0…7; in practice, k1-k64 can be rearranged in a serpentine or lookup table manner, but this arrangement must be written into the calibration transfer matrix.

[0046] Let Nx and Ny be the number of spatial units in the two lateral directions of the detection field of view, P be the number of effective wavenumber channels to be recovered for each spatial unit, and M be the number of effective pixels that the area array can be used for encoding readout. In the direct allocation implementation, each B(i,j) has a non-overlapping or clearly separable group of wavenumber pixels, satisfying M≥Nx·Ny·P. In the compressed coding implementation, M is allowed to be less than Nx·Ny·P, but the calibration transfer matrix must be calibrated per spatial position, per wavenumber, and representative polarization / temperature state, and pass the preset matrix conditions, cross-validation error, inversion residual, spatial crosstalk, and effective depth signal-to-noise ratio threshold. It is recommended that the normalized reading residual be no greater than 5%, the spatial crosstalk be no greater than -10dB, and the specular peak depth error be no greater than 10μm. Spatial units or entire frames that do not pass the threshold shall not be output as valid structural results. Examples of Nx, Ny, P, M and H dimension values ​​are shown in Table 3 below.

[0047] Table 3

[0048] The direct allocation examples above are used to illustrate feasible pixel budgets and do not represent the final lateral resolution. Example A corresponds to a 2.0mm × 2.0mm object-side field of view, with a lateral sampling interval of 31.25μm; when 8–15μm lateral sampling is required, the single field of view should be reduced, a larger pixel array should be used, P should be reduced, or a proven compression / multi-frame method should be adopted. Nx, Ny, and P cannot be arbitrarily increased simultaneously with a fixed M.

[0049] In one example, the detection field of view is divided into 32×32 object space units, each unit corresponding to a sub-region on the array, and 16 or 32 distinguishable wavenumber channels are set in the sub-region; the actual number of units and channels is determined based on the number of array pixels, the metasurface pattern and the target lateral resolution.

[0050] The array detector can have dual gain, dual conversion gain, variable integration time or partitioned readout to simultaneously accommodate strong surface reflections and weak deep reflections: the high-gain channel is used for weak deep reflections, and the low-gain channel is used for strong surface reflections. The two channels are registered through a unified calibration matrix to synthesize space-wavenumber data with extended dynamic range.

[0051] It is worth mentioning that the system in this scheme obtains the encoding transfer matrix H during the factory calibration or pre-use calibration stage; the calibration inputs include multiple known wavelengths, incident angles, polarization states, lateral spatial positions, reference optical path differences, and temperature states; the calibration process includes: acquiring dark field and uniform field readings; measuring the response of each encoded sub-region at multiple known wavelengths or wavenumber points; the calibration results include at least: dark field, flat field, pixel gain and bias, spectral response, spatial distortion, metasurface dispersion, area sub-region boundaries, crosstalk between adjacent spatial cells, reference arm field-dependent optical path difference, and temperature drift.

[0052] The measurement of spatial mapping and crosstalk is performed at multiple object space cell locations; the optical path of the reference arm or the state of the reference wavefront is changed to estimate the field-dependent optical path difference; and the above measurements are repeated at at least two temperature or gain states; the calibration target may include a uniform diffuse reflective target, a reference reflective plate with a known optical path difference, a narrow-band near-infrared light source, a polarizer, and a two-dimensional spatial marker plate.

[0053] The array readings after background subtraction are expanded by pixel coordinates as y∈R^M, and the spectral interferometric samples to be recovered are expanded by spatial units and wavenumber channel indices as x∈R^N, where N=Nx·Ny·P. Then the forward model is y=Hx+b+n, H∈R^{M×N}, and the element H[m,n] represents the calibration response of the (i,j,p)th spatial-wavenumber basis to the (u,v)th detector pixel. The array readings obtained from a single exposure are D(u,v)=ΣH(u,v;i,j,k)I(i,j,k)+n(u,v), where I(i,j,k) is the interferometric spectrum of the (i,j)th object space unit at wavenumber k, H represents the calibration transfer matrix including the sample arm, reference arm, metasurface coding layer and detector, and n is noise. The field-dependent phase of the reference arm can be incorporated into H, or used as an independent lookup table for φref(i,j,kp) after decoding.

[0054] The decoding module can employ block matrix inversion, regularized least squares, compressed sensing, sparse recovery, maximum likelihood estimation, or physical constraint iterative optimization. For the decoded I(i,j,k), the processor performs dark field / flat field correction, wavenumber linearization, background removal, field-matched phase correction, dispersion compensation, window function, depth roll-off correction, and Fourier transform to obtain A-scan, B-scan, C-scan, and 3D volume data.

[0055] During direct allocation, each B(i,j) does not overlap, and H can be written as diag(H00,H01,…), where each local block H(i,j)∈R^{mB×P}. When mB=P=64 and crosstalk is negligible, the local blocks are close to the diagonal or permutation matrix. The engineering implementation only stores 4,096 or 5,120 local 64×64 blocks, without needing to generate a complete dense matrix.

[0056] During compression encoding, H is a rectangular matrix, with each column vector being non-zero only within a local point of the diffusion footprint. It can be stored as a sparse matrix, a set of convolutional kernels, or a forward / adjoint operator. Recovery employs regularized least squares with non-negativity, spectral smoothing, spatial total variation, or low-rank priors, compressed sensing, sparse recovery, or physically constrained iterative optimization. The compression implementation must be cross-validated with reserved wavelengths and reserved spatial locations. It is recommended that the normalized reading residual be ≤5% and the spatial crosstalk ≤ Frames that do not meet the thresholds (10dB, mirror peak depth error ≤10μm) are marked as invalid.

[0057] For the decoded I(i,j,k), the reconstruction processor performs dark field / flat field correction, wavenumber linearization, background removal, field-matched phase correction, dispersion compensation, window function processing, depth roll-off correction, and Fourier transform to obtain A-scan, B-scan, C-scan, and 3D volume data. It then outputs tomographic images, epidermal thickness, epidermal-dermal junction depth, lesion boundary depth, or pre- and post-treatment changes. The reconstruction processor also generates quality markers for each object space cell. These quality markers are determined based on at least one of the following: coherence gate coverage, signal-to-noise ratio, array saturation, matrix inversion residual, and number of effective wavenumber channels. The quality markers are used to generate effective region masks for the depth structure data, and uncontrolled depth structures are not output. For spatial cells with excessively large condition numbers, insufficient effective wavenumber channels, pixel saturation, or low reference interferometry visibility, the system generates invalid masks or quality markers.

[0058] Furthermore, the handheld probe adopts a layout of a front optical head, a middle grip area, and a rear electronic power supply area. The front optical head houses a sample arm, a reference arm, a field matching element, a metasurface camera, and a detection window. The electronic power supply area houses a light source driver, an FPGA or embedded processor, a battery, storage, and a wireless / USB interface. The light source, interference module, sample arm, reference arm, metasurface camera, reconstruction processor, battery, and communication module can all be integrated into the same handheld probe housing. External mobile phones, tablets, or computers are only used as display, storage, remote consultation, or further analysis terminals.

[0059] The detection window can be a flat glass, a curved window, a disposable protective film, a sterilizable contact cap, or a transparent window with positioning marks. This allows for localized planarization of the skin, reduces hand-held tremors and curvature differences, and serves as a mechanical reference for the sample arm distance and optical path. Pressure sensors, distance sensors, temperature sensors, and a visible light-assisted camera, as well as an inertial measurement unit and an optical power monitor, can be installed around the window. The system triggers an exposure after meeting the stability conditions for attitude, pressure, distance, and signal. If motion, unstable contact pressure, camera saturation, abnormal temperature, or excessive optical power is detected, the exposure is canceled or the power is reduced. When there is frame loss during array readout, encoding matrix verification failure, or reference arm optical path deviation, the control module prohibits marking the frame as a valid detection result and prompts for repositioning, recalibration, or adjustment of exposure parameters. A reference reflector can also be installed inside the probe or near the detection window. This reflector is used for rapid self-checking during system startup to determine whether the coded camera response, reference arm optical path, light source power, and array readout status meet the exposure conditions.

[0060] Since multiple wavenumber channels are obtained in a single area array exposure, the system can reconstruct the full-band, surface-optimized sub-band, depth-optimized sub-band, polarization channel, or spectral ratio channel separately without additional optical exposure, and compare the signal-to-noise ratio at each depth. For example, the processor reconstructs the depth signal using the full-band, short-wavelength sub-band, and long-wavelength sub-band respectively, calculates the signal-to-noise ratio and roll-off of each sub-band at depths of 0.5mm, 1.0mm, and 1.5mm, selects the sub-band with high signal-to-noise ratio and non-saturated surface as the preferred parameters for the next frame, and saves the full-band image, preferred sub-band image, and effective imaging depth in the report.

[0061] The control module can adjust the light source power, exposure time, band subset, polarization state, or reference light intensity of the next frame based on the signal-to-noise ratio, depth roll-off, camera full-well / saturation, reference arm power, sample arm return power, probe temperature, orientation, and contact pressure. This control is an inter-frame feedback mechanism. The structural data within the current frame is still obtained from a single area array exposure. The optical power, exposure, and safety design of the near 1550nm light source are verified according to applicable laser product safety standards, medical electrical safety standards, and human body light exposure risk assessments. Monitoring, gating, and prohibition of exposure beyond limits are used as equipment control features.

[0062] To further elaborate on the application scenarios of this solution, the following scenarios are used as examples: Scenario 1: Continuous wide-field single-shot structural imaging; employing a broadband low-coherence light source with a center wavelength of approximately 1550nm and a bandwidth of approximately 100nm, using quasi-common, Mirau-type, or Linnik-type interference structures; the sample arm illuminates the skin field of view of approximately 4×4mm² or 6×6mm² through a telecentric objective lens, and the reference arm is equipped with a curved reference surface or a reference metal lens to ensure that the optical path difference within the detection field of view falls within the effective coherence range; the metasurface coding camera maps the interference spectrum of each object space unit to the surface array region or an invertible coding pattern, recovers I(i,j,k) after a single exposure, and generates three-dimensional structural maps of the epidermis, epidermal-dermal junction, and superficial dermis through Fourier transform.

[0063] Example 2: Handheld lesion boundary detection; a transparent contact window, pressure sensor, and distance sensor are installed at the probe tip. The operator places the probe against the edge of the lesion, and an exposure is triggered when the posture and pressure meet the gating conditions. The processor calculates the lateral contour, depth boundary, and structural differences from adjacent healthy skin of the lesion based on the surface layer, epidermal-dermal boundary, and low-scattering / high-scattering regions. Multiple fields of view are registered and stitched together using visible light-assisted images and contact window positioning markers.

[0064] Example 3: Same-frame depth sub-band selection; the camera records multiple wavenumber channels in the range of 1500-1600nm in a single exposure, and the processor reconstructs the depth signal using full-band, short-wavelength sub-band, and long-wavelength sub-band respectively, and calculates the SNR and roll-off of each sub-band at depths of 0.5mm, 1.0mm, and 1.5mm; the system selects the sub-band with higher SNR and no surface saturation as the preferred parameters for the next frame, and saves the full-band image, preferred sub-band image, and effective imaging depth in the report at the same time.

[0065] Example 4: Dual-gain area array and surface saturation protection; the metasurface camera employs a dual-conversion gain readout circuit or a high / low gain area array readout circuit. The high-gain channel is used for deep, weak reflections, and the low-gain channel is used for strong surface reflections; the two channels are registered using a unified calibration matrix and then synthesized into space-wavenumber data with extended dynamic range. If the surface channel still exhibits full-well saturation, the safety gating module reduces power or shortens exposure in the next frame and prompts for readjustment of the contact window.

[0066] Example 5: Optional polarization / repeated snapshots; the metasurface coding layer contains anisotropic nanostructures, which distribute orthogonal polarization components to different surface array regions, thereby generating polarization-sensitive structure maps; two or more independent snapshots can also be taken at the same location for speckle variance, motion correction or blood flow correlation contrast; repeated snapshots do not change the core architecture of the single-scan structural imaging of the present invention.

[0067] Example 6: Spatial-Wavenumber Transfer Matrix Calibration; Without placing human tissue, a uniform diffuse reflective target is placed at the detection window position, and multiple adjacent object-space spatial units within the detection field of view are sequentially covered using a two-dimensional spatial marker plate; multiple narrowband wavelength points within 1500nm to 1600nm are input into the interferometer module, while the equivalent optical path difference of the reference arm is changed, and the two-dimensional coded readings of the metasurface camera are recorded. The processor solves for the transfer matrix H based on the known input spectrum, spatial position, and reference optical path difference, and cross-validates the reconstruction error, spatial crosstalk, wavenumber positioning error, and edge field roll-off; the calibration results are saved in association with temperature, gain, and polarization state for subsequent exposure data decoding.

[0068] In one example, the detection field of view is divided into 32×32 object space units, each unit corresponding to a sub-region on the array, with 16 or 32 distinguishable wavenumber channels within the sub-region; the actual number of units and channels is determined based on the number of array pixels, the metasurface pattern, and the target's lateral resolution. This example is used to illustrate feasible mapping relationships and does not constitute a limitation on the scope of protection.

[0069] Example 7: Wide-field Linnik type detailed optical path; A broadband low-coherence light source with a center wavelength of approximately 1550nm is used. The light source output passes sequentially through an optical isolator, a variable optical attenuator, a polarization controller, a collimator, a beam expander, and a shimming element before entering a 50:50 unpolarized beam splitter BS1; BS1 divides the illumination light into a sample arm and a reference arm. The sample arm is equipped with a telecentric objective lens and a transparent contact window to form a continuous wide-field illumination on the skin surface; the backscattered light from each spatial unit of the skin returns along the original path of the sample arm; the reference arm is equipped with a reference objective lens, a reference phase element, and a reference mirror that are equivalently matched to the sample objective lens. The reference phase element applies preset phase or optical path compensation to different field-of-view positions.

[0070] The BS1 can employ a near-infrared unpolarized cubic or planar beam splitter with a 5-10mm aperture, designed with a bandwidth of 1500-1600nm; at the center wavelength of 1550nm, R:T=50, with a tolerance of ±5%, and single-path R or T maintained at 45%-55% across the entire band; insertion loss is preferably ≤1dB, polarization-dependent loss ≤0.2dB, coating residual reflection ≤0.5%, and transmission wavefront distortion ≤λ / 4PV (based on 1550nm); to improve the utilization rate of weak sample backlight, a 70:30 or 30:70 beam splitting can also be used, and the reference light and sample light at the array can meet the requirements of heterogeneous gain, full-well capacity, and saturation rate through a variable attenuator on the reference arm; the splitting ratio and polarization / temperature response are entered into the flat field and H calibration.

[0071] The sample return light and the reference return light are combined at BS1 and output from the detection port. The detection end is sequentially equipped with a polarization analyzer, a field stop, a relay imaging lens, a metasurface spatial-wavenumber coding layer, and an InGaAs area array detector. The relay imaging lens establishes a conjugate relationship between the skin detection surface and the metasurface coding surface. The field stop limits crosstalk between adjacent spatial units. The system uses axial delay of the reference arm to bring the central field of view into the coherence gate and compensates for the optical path difference of the edge fields of view through a reference phase element and a calibration matrix. In this embodiment, no galvanometer, MEMS scanning mirror, or mechanical lateral scanning mechanism is set in a single detection field of view.

[0072] This invention also discloses a method for a handheld skin optical coherence tomography system based on continuous wide-field space-wavenumber coding, field-matched reference wavefront, and a near 1550nm light source, comprising the following steps: Step 1: Form a continuous wide-field detection field on the skin surface and divide the detection field into multiple two-dimensional adjacent object space units; Step 2: Use the field-matching reference arm to form a reference wavefront related to the position of the object space unit, so that the optical path difference between the sample arm and the reference arm corresponding to different object space units is kept within the effective coherence range. Step 3: During a single area array exposure, acquire two-dimensional coded readings of multiple object space units within the detection field of view using a metasurface space-wavenumber encoded camera; Step 4: Recover the wavenumber domain interferometric data of each object space unit based on the calibration transfer matrix, and perform field-dependent reference phase correction, wavenumber linearization, and dispersion compensation. Step 5: Perform Fourier transform along the wavenumber direction to output one or more of the following: tomographic image of skin tissue, three-dimensional structural data, epidermal thickness, depth of the epidermal-dermal junction, depth of lesion boundary, or changes in indicators before and after treatment.

[0073] This invention is not limited to the preferred embodiments described above. Anyone can derive other products in various forms under the guidance of this invention. However, regardless of any changes in shape or structure, any technical solution that is the same as or similar to this application falls within the protection scope of this invention.

Claims

1. A handheld skin optical coherence tomography system based on continuous wide-field spatial-wavenumber coding, field-matched reference wavefront, and a near 1550nm light source, characterized in that, include: A broadband, low-coherence light source with a wavelength of nearly 1550nm; Interference module; A continuous wide-field sample arm is used to illuminate the sample light onto the detection field of view and divide the detection field of view into multiple two-dimensional adjacent object space units; The field-matching reference arm, including a reference reflecting surface and a field-matching reference wavefront element, is used to generate a reference wavefront that varies with the position of the object space unit, so that the optical path difference between the sample arm and the reference arm corresponding to different object space units is kept within the effective coherence range of the light source. A metasurface spatial wavenumber encoded camera, comprising a metasurface encoding layer and an array detector responsive to the near 1550 nm band, is used to encode the interference light of multiple object space units according to their spatial position and wavenumber during a single exposure and output a two-dimensional encoded reading. The reconstruction processor is used to recover spatial wavenumber interferometric data of each object space unit from two-dimensional coded readings based on a calibration transfer matrix including field-matched reference wavefront information, and to perform Fourier transform along the wavenumber direction to generate deep structural data of skin tissue.

2. The handheld skin optical coherence tomography system based on continuous wide-field spatial-wavenumber coding, field-matched reference wavefront, and near-1550nm light source as described in claim 1, characterized in that: The continuous wide-field sample arm maps object space units within the detection field of view to multiple sub-regions of the area array detector or overlapping point diffusion responses that can be distinguished by the calibration transfer matrix, according to a two-dimensional adjacency relationship. This allows each object space unit to obtain multiple independent wavenumber observations in the same exposure, which are used to recover the interference spectrum of the corresponding object space unit.

3. The handheld skin optical coherence tomography system based on continuous wide-field spatial-wavenumber coding, field-matched reference wavefront, and near-1550nm light source as described in claim 1, characterized in that: The total number of effective wavenumber channels is obtained by multiplying the number of spatial units in the detection field of view in the x direction, the number of spatial units in the detection field of view in the y direction, and the number of effective wavenumber channels to be recovered in each spatial unit. In the direct allocation mode, the number of effective pixels used for encoding readout by the area array detector is greater than or equal to the total number of effective wavenumber channels. In the compressed coding mode, the number of effective pixels used for encoding readout by the area array detector is less than the total number of effective wavenumber channels.

4. The handheld skin optical coherence tomography system based on continuous wide-field spatial-wavenumber coding, field-matched reference wavefront, and near-1550nm light source as described in claim 1, characterized in that: The reconstruction processor performs at least two of the following on the recovered spatial wavenumber interferometric data: wavenumber linearization, reference wavefront phase correction, dispersion compensation, background removal, depth roll-off correction, and window function processing. Then, it performs Fourier transform and reconstructs the depth signals of the full-band, surface-preferred sub-band, depth-preferred sub-band, polarization channel, or spectral ratio channel based on multiple wavenumber channels obtained from a single exposure. After comparing the depth signal with the noise ratio, the effective sub-band is selected.

5. The handheld skin optical coherence tomography system based on continuous wide-field spatial-wavenumber coding, field-matched reference wavefront, and near-1550nm light source as described in claim 1, characterized in that: The field matching reference wavefront element is one or more combinations of curved reference mirror, reference metal lens, metasurface phase plate, liquid crystal phase modulator, and adjustable deformable mirror.

6. The handheld skin optical coherence tomography system based on continuous wide-field spatial-wavenumber coding, field-matched reference wavefront, and near-1550nm light source as described in claim 1, characterized in that: After compensation by the field-matched reference wavefront element, the peak-to-valley value of the optical path difference in the detected field of view is no greater than 5μm, and the static phase residual RMS after calibration is no greater than π / 8 and the upper limit is no greater than π / 4.

7. The handheld skin optical coherence tomography system based on continuous wide-field space-wavenumber coding, field-matched reference wavefront, and near-1550nm light source according to claim 1, characterized in that: The metasurface coding layer is one or more combinations of dielectric nanopillars, nanofins, nanopores, supergratings, and polarization-selective nanostructures, used to perform preset joint coding of the wavenumber, lateral spatial position, polarization, and phase of the input light.

8. The handheld skin optical coherence tomography system based on continuous wide-field spatial-wavenumber coding, field-matched reference wavefront, and near-1550nm light source according to claim 1, characterized in that: The center wavelength of the near 1550nm broadband low coherence light source is between 1520nm and 1580nm, and the spectral range is between 1500nm and 1600nm. The light source is a superluminescent diode, an amplified spontaneous emission light source, a supercontinuum light source after bandpass filtering, or a low coherence laser.

9. The handheld skin optical coherence tomography system based on continuous wide-field spatial-wavenumber coding, field-matched reference wavefront, and near-1550nm light source according to claim 1, characterized in that: Several components, including a near 1550nm broadband low-coherence light source, an interference module, a continuous wide-field sample arm, a field-matching reference arm, a metasurface spatial wavenumber encoding camera, a reconstruction processor, a battery, and a communication module, are housed within the handheld probe. The handheld probe includes at least two of the following: a transparent detection window, a distance sensor, a pressure sensor, an inertial measurement unit, a temperature sensor, a visible light-assisted camera, and an optical power monitor.

10. A method for using the handheld skin optical coherence tomography system based on continuous wide-field spatial-wavenumber coding, field-matched reference wavefront, and a near-1550nm light source as described in any one of claims 1-9, characterized in that, Includes the following steps: Step 1: Form a continuous wide-field detection field on the skin surface and divide the detection field into multiple two-dimensional adjacent object space units; Step 2: Use the field-matching reference arm to form a reference wavefront related to the position of the object space unit, so that the optical path difference between the sample arm and the reference arm corresponding to different object space units is kept within the effective coherence range. Step 3: During a single area array exposure, acquire two-dimensional coded readings of multiple object space units within the detection field of view using a metasurface spatial wavenumber coded camera; Step 4: Recover the wavenumber domain interferometric data of each object space unit based on the calibration transfer matrix, and perform field-dependent reference phase correction, wavenumber linearization, and dispersion compensation. Step 5: Perform Fourier transform along the wavenumber direction to output one or more of the following: tomographic image of skin tissue, three-dimensional structural data, epidermal thickness, depth of the epidermal-dermal junction, depth of the lesion boundary, and changes before and after treatment.

Images