Oral ablation detection method and system

Abstract

This invention discloses an oral ablation detection method and system, belonging to the field of oral detection technology, including: S1, oral tissue sampling; S2, motion compensation; S3, image reconstruction; S4, lesion analysis; S5, ablation planning; S6, ablation execution. This invention utilizes phase-amplitude joint displacement estimation, extracting depth-direction micro-displacement using phase difference and lateral displacement using amplitude centroid shift. This effectively compensates for non-rigid tissue peristalsis caused by swallowing and tongue movements without relying on external markers or high-frequency frame rates, ensuring that the subsequently reconstructed three-dimensional image and ablation target area localization are not distorted due to motion, thus improving the targeting consistency between detection and treatment. Texture analysis at different scales is applied along the oral mucosal epithelium, basement membrane zone, and lamina propria, and depth-adaptive weights are assigned based on the pathological sensitivity differences of each layer, highlighting the probabilistic response of early cancerous areas.

Description

Technical Field

[0001] This invention relates to the field of oral examination technology, specifically to an oral ablation detection method and system. Background Technology

[0002] Optical coherence tomography (OCT) has been widely used in the detection of oral mucosal diseases, and its combination with ablation therapy equipment can achieve integrated diagnosis and treatment. Currently, oral OCT detection systems typically use conventional cross-correlation registration algorithms to eliminate tissue motion artifacts and identify lesions based on global texture features; ablation planning often relies on the doctor's experience to manually set energy parameters.

[0003] However, existing technologies have the following shortcomings: oral tissues will undergo non-rigid peristalsis during the scanning process due to swallowing, tongue movement, etc., and their lateral displacement is much greater than the depth displacement. Traditional isotropic registration algorithms are difficult to accurately compensate for this movement with obvious directional differences, resulting in interlayer misalignment in the 3D image reconstruction, which in turn affects the accurate positioning of the ablation target area.

[0004] In response to this problem, this application proposes an oral ablation detection method and system to solve the above-mentioned issues. Summary of the Invention

[0005] The purpose of this invention is to provide an oral ablation detection method and system to solve the problem that in the prior art, oral tissues will produce non-rigid peristalsis during the scanning process due to swallowing, tongue movement, etc., and the lateral displacement is much greater than the depth direction displacement. Traditional isotropic registration algorithms are difficult to accurately compensate for this movement with obvious directional differences, resulting in interlayer misalignment in the three-dimensional image reconstruction, which in turn affects the accurate positioning of the ablation target area.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] Firstly, this application provides a method for oral ablation detection, including:

[0008] The raw interference spectrum signal of oral tissue was acquired to obtain the raw interference signal;

[0009] Motion compensation based on tissue peristalsis characteristics is performed on the original interference signal to obtain a compensated interference signal;

[0010] The compensated interference signal is subjected to frequency domain transformation and scattering correction to obtain a three-dimensional tissue structure image;

[0011] Depth-adaptive texture heterogeneity analysis was performed on the three-dimensional tissue structure image to obtain a probability distribution map of the ablation target area;

[0012] Based on the probability distribution map of the ablation target area, the ablation energy timing and needle path are planned to obtain the dynamic ablation parameter sequence;

[0013] The ablation energy is output according to the dynamic ablation parameter sequence, and the pyroelectric signal is collected simultaneously to obtain the real-time ablation response waveform.

[0014] Furthermore, the motion compensation step includes:

[0015] The two consecutive frames of raw interference signals are converted into complex domain signals, and the amplitude and phase components are extracted.

[0016] The displacement in the depth direction is calculated using the phase difference, and the lateral displacement is calculated using the amplitude centroid offset, thus obtaining the displacement vector of each pixel.

[0017] The displacement vector is then interpolated using thin-plate splines to generate a dense deformation field. Finally, the dense deformation field is used to reverse-sample the next frame signal to obtain a compensated interference signal.

[0018] Furthermore, the depth displacement and lateral displacement are calculated using the following formulas:

[0019]

[0020]

[0021] Where ϕn,ϕn+1 are the phases of the complex signals of two adjacent frames, |An|, |An+1| are the corresponding amplitudes, λ is the center wavelength of the light source, n is the average refractive index of the oral tissue, (x0,z0) is the current pixel coordinates, and the horizontal integration range covers a window centered at x0 with a width of 2Wx, where Wx is greater than the half-width Wz of the depth direction window.

[0022] Furthermore, the depth-adaptive texture heterogeneity analysis includes: dividing the three-dimensional tissue structure image into three regions along the depth direction: epithelial layer, basement membrane zone, and lamina propria; calculating local binary pattern texture features for each region; and then weighting and fusing them according to the region depth index, wherein the weight factor of the basement membrane zone region is set to twice that of the epithelial layer, to obtain the probability distribution map of the ablation target area.

[0023] Furthermore, in the weighted fusion of the regional depth index, the weight of each region is determined according to the formula w(d)=w0⋅e -βd The weight of superficial abnormal proliferation decreases exponentially with depth d, where w0 is the initial weight of the epithelial surface and β is the decay coefficient. This makes the weight of superficial abnormal proliferation higher than that of deep normal tissue, thus highlighting the probability response of early cancerous areas.

[0024] Furthermore, the step of planning the ablation energy timing and needle path further includes thermal damage avoidance assessment: based on the density of healthy tissue within 2 mm around each candidate point in the probability distribution map of the ablation target area, the virtual thermal diffusion priority is calculated, the needle path sequence that makes the thermal accumulation temperature of the surrounding healthy tissue below 42°C is selected, and the interval time between adjacent pulses is adjusted to match the local blood perfusion time constant to obtain a dynamic ablation parameter sequence.

[0025] Furthermore, the virtual thermal diffusion priority is determined by comparing the thermal influence radius of each candidate point with the distance to adjacent important structures, including nerve bundles or glands. If the distance is less than 0.5 mm, the ablation power at that point is automatically reduced to 60% of the rated value, while the pulse interval is extended to 1.5 times the original interval.

[0026] Furthermore, in the step of outputting ablation energy and simultaneously acquiring pyroelectric signals, the peak amplitude and time constant of the pyroelectric signal are recorded with a period of 10 milliseconds, and the peak amplitude is compared with a preset ablation depth-amplitude calibration curve. If the measured peak amplitude is lower than the lower limit of the curve, the energy density of subsequent pulses is automatically increased by 5%; if it is higher than the upper limit of the curve, the energy density is reduced by 3%, thus obtaining the closed-loop corrected real-time ablation response waveform.

[0027] Furthermore, the method also includes the following steps: displaying the three-dimensional tissue structure image, the probability distribution map of the ablation target area, and the real-time ablation response waveform on the same three-dimensional oral cavity model in a semi-transparent overlay manner, wherein the ablated area is marked with a red gradient and the area to be ablated is marked with a blue gradient, and the ablation progress percentage is updated in real time.

[0028] Secondly, this application provides an oral ablation detection system, comprising:

[0029] The oral tissue sampling module is used to acquire the raw interference spectrum signal of oral tissue to obtain the raw interference signal.

[0030] The motion compensation module is used to perform motion compensation on the original interference signal based on tissue peristalsis characteristics to obtain a compensated interference signal;

[0031] The image reconstruction module is used to perform frequency domain transformation and scattering correction on the compensated interference signal to obtain a three-dimensional tissue structure image;

[0032] The lesion analysis module is used to perform depth-adaptive texture heterogeneity analysis on the three-dimensional tissue structure image to obtain a probability distribution map of the ablation target area;

[0033] The ablation planning module is used to plan the ablation energy timing and needle path based on the probability distribution map of the ablation target area, and obtain a dynamic ablation parameter sequence.

[0034] The ablation execution module is used to output ablation energy according to the dynamic ablation parameter sequence and simultaneously acquire pyroelectric signals to obtain the real-time ablation response waveform.

[0035] Compared with existing technologies, the oral ablation detection method and system provided by this invention utilizes phase-amplitude joint displacement estimation. It extracts depth-direction micro-displacement using phase difference and lateral displacement using amplitude centroid shift, effectively compensating for non-rigid tissue peristalsis caused by swallowing and tongue movements without relying on external markers or high-frequency frame rates. This ensures that the subsequently reconstructed 3D image and ablation target localization are not distorted due to movement, improving the targeting consistency between detection and treatment. Texture analysis at different scales is applied along the oral mucosal epithelium, basement membrane, and lamina propria, and depth-adaptive weights are assigned based on the pathological sensitivity differences of each layer, highlighting the probabilistic response of early-stage cancerous areas. Simultaneously, virtual thermal diffusion priority is introduced into the ablation planning, automatically adjusting energy parameters and pulse intervals based on the density of healthy tissue surrounding the target point and the distance to important anatomical structures (such as nerve bundles and glands). This achieves selective ablation of lesions while minimizing thermal damage to surrounding healthy tissues. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0037] Figure 1 This is a flowchart of an oral ablation detection method provided in an embodiment of the present invention;

[0038] Figure 2 This is a block diagram of an oral ablation detection system provided in an embodiment of the present invention. Detailed Implementation

[0039] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0040] As attached Figure 1 As shown:

[0041] Example 1:

[0042] A method for detecting oral ablation includes:

[0043] S1. Acquire the raw interference spectrum signal of oral tissue to obtain the raw interference signal;

[0044] In step S1, a dual-band light source is used for alternating illumination. First, the deep connective tissue signal is obtained with the 1300nm near-infrared band, and then the surface epithelial signal is obtained with the 800nm ​​visible light band. The two band signals are then linearly registered according to spatial coordinates and fused to obtain the multi-spectral original interference signal.

[0045] Furthermore, step S1 is implemented using optical coherence tomography (OCT) technology. The system employs a swept-frequency light source or a broadband light source, with the center wavelength selectable to alternate between 1300 nm (near-infrared) and 800 nm (visible light) dual-band illumination.

[0046] In practice, the fiber optic probe is attached to the surface of the oral mucosa (such as the back of the tongue, buccal mucosa, or soft palate), and the interference spectral signals in the horizontal (x direction) and depth (z direction) are obtained by scanning with a two-dimensional galvanometer.

[0047] The optical path difference between the reference arm and the sample arm generates interference fringes, which are recorded by a balanced detector to obtain the original interference signal (time domain or frequency domain signal).

[0048] To obtain more complete tissue information, the system first scans at 1300 nm to acquire the scattering signals of deep connective tissue (such as lamina propria and collagen fiber bundles); then it quickly switches to 800 nm to acquire the fine structure signals of the surface epithelium (approximately 200–300 μm thick).

[0049] The two-band signals are kept consistent in spatial coordinates by using a common optical path, and then they are fused into a multi-band original interference signal by pixel-level linear registration (e.g., using feature points of the non-moving region of the hard palate for alignment).

[0050] The signal is stored in the form of three-dimensional data blocks. The first dimension is the horizontal scan points (e.g., 500 points), the second dimension is the depth sampling points (e.g., 1024 points), and the third dimension is the frame sequence (e.g., 300 consecutive frames).

[0051] To ensure the temporal relevance of subsequent motion compensation, the frame rate is set to 30–50 Hz, and the line density per frame can be adjusted according to the size of the lesion (usually 256 lines / frame).

[0052] S2. Perform motion compensation based on tissue peristalsis characteristics on the original interference signal to obtain a compensated interference signal;

[0053] In step S2, the motion compensation step includes:

[0054] The two consecutive frames of raw interference signals are converted into complex domain signals, and the amplitude and phase components are extracted.

[0055] The displacement in the depth direction is calculated using the phase difference, and the lateral displacement is calculated using the amplitude centroid offset, thus obtaining the displacement vector of each pixel.

[0056] The displacement vector is then interpolated using thin-plate spline to generate a dense deformation field. Finally, the dense deformation field is used to reverse-sample the next frame signal to obtain a compensated interference signal.

[0057] The depth displacement and lateral displacement are calculated using the following formulas:

[0058]

[0059]

[0060] Where ϕn,ϕn+1 are the phases of the complex signals of two adjacent frames, |An|, |An+1| are the corresponding amplitudes, λ is the center wavelength of the light source, n is the average refractive index of the oral tissue, (x0,z0) is the current pixel coordinates, and the horizontal integration range covers a window centered at x0 with a width of 2Wx, and Wx is greater than the half-width Wz of the depth direction window;

[0061] Furthermore, the oral mucosa undergoes peristaltic deformation under natural conditions due to swallowing, tongue movement, and respiration. This deformation is characterized by large lateral displacement (up to 0.5 mm / frame along the tangent to the mucosal surface) and extremely small displacement in the depth direction (perpendicular to the mucosal surface) (< 30 μm / frame), and the deformation is non-rigid and locally non-uniform. Traditional rigid body registration or isotropic optical flow methods are ineffective. This invention employs a phase-amplitude joint displacement estimation method, specifically implemented as follows:

[0062] First, the original interference signal is converted into a complex domain signal (the amplitude A and phase φ are extracted by Hilbert transform).

[0063] The displacement in the depth direction is calculated using the phase difference: Δz = (λ / (4πn))·Δφ, where λ is the center wavelength of the light source (take the weighted average, for example, 1050 nm for dual-band), and n is the average refractive index of the oral mucosa (measured as 1.38–1.42).

[0064] Because the phase difference is extremely sensitive to minute displacements (nanometer level) and is unaffected by amplitude fluctuations, it can accurately capture the elastic compression / stretching in the depth direction caused by creep.

[0065] Lateral displacement is achieved using the amplitude centroid tracking method: at each depth layer z0, the centroid of the amplitude distribution within the lateral window (window width Wx = 11 pixels, approximately 110 μm) is calculated, and the centroid offset between frames is the lateral displacement Δx. This method avoids the high complexity of cross-correlation calculations and is also robust to low-texture regions (such as uniform mucus).

[0066] After obtaining the sparse displacement field (one control point per 5×5 pixels), thin plate spline (TPS) interpolation is used to generate the dense deformation field of the whole image.

[0067] The TPS kernel function is r 2 ln(r) can simulate the nonlinear creep of soft tissue while maintaining the continuity of the first derivative. Finally, the signal of the next frame is subjected to reverse deformation resampling (bicubic interpolation) to map all pixels back to the coordinate system of the reference frame (usually the first frame or the moving average frame), thereby eliminating motion artifacts and outputting a compensated interference signal.

[0068] For the formulas of depth displacement and lateral displacement, λ and n are known constants of the system (λ is calibrated by a spectrometer, and n can be determined by a refractive index meter on excised oral tissue samples, with a statistical value of 1.40).

[0069] Specifically, λ was obtained by spectrometer calibration (e.g., 1325 nm ± 10 nm), and n was measured using an Abbe refractometer on ex vivo oral mucosal samples (at least 10 samples, mean 1.40 ± 0.02). The window width Wx was determined by statistically analyzing the 95th percentile of peristaltic displacement in living oral cavity samples.

[0070] The method for determining the window width Wx is as follows: OCT sequences (30 Hz frame rate) of the tongue dorsum region of 10 volunteers during natural swallowing were acquired. The absolute value of the centroid offset between horizontally adjacent frames was calculated, and the 95th percentile was set to 8 pixels (approximately 80 μm). Therefore, Wx = 11 pixels was set to cover 110% of the maximum displacement. No additional window parameters are required in the phase difference calculation; each pixel is calculated independently. However, to prevent phase wrapping (2π jump), a phase unwrapping algorithm (such as quality-guided path integral) is used.

[0071] S3. Perform frequency domain transformation and scattering correction on the compensated interference signal to obtain a three-dimensional tissue structure image;

[0072] In step S3, the compensated interference signal remains in the original interference spectrum form (wavenumber domain). First, a Fast Fourier Transform (FFT) is performed on each A-scan (depth direction) to convert the wavenumber domain signal to the depth domain (z-space), obtaining the reflectance distribution. Since the spectral shape of the light source is not ideally rectangular, spectral shaping is required (e.g., adding a Hanning window to suppress sidelobes).

[0073] Meanwhile, the light intensity attenuation caused by tissue scattering decreases exponentially with depth, thus requiring scattering correction. Typically, the attenuation coefficient μ is assumed to be constant, and logarithmic gain compensation is used, i.e., multiplying each depth point by exp(2μz). The value of μ is obtained by estimating the attenuation rate of the epithelial surface (approximately 3–5 dB / mm). Furthermore, dispersion mismatch leads to a decrease in resolution, which can be corrected using a numerical dispersion compensation algorithm (multiplying the phase by a quadratic term before the Fourier transform).

[0074] Furthermore, after the above processing, the gray value of each pixel represents the backscattering intensity at that location. By arranging all A-scans in the scanning order, a three-dimensional tissue structure image can be obtained.

[0075] This image is typically presented in cross-sectional (B-scan) or 3D volumetric rendering format, clearly showing the layered structure of the oral mucosa:

[0076] The outermost layer is keratinized / non-keratinized epithelium (high scattering, thin layer), the middle layer is the basement membrane zone (slightly low scattering), and the lower layer is the lamina propria (fibrous, moderate scattering).

[0077] To achieve depth-adaptive analysis, the system automatically calculates the epithelial layer thickness (by detecting the axial distance between the surface peak and the basement membrane valley) and divides the image into three sub-regions along the depth direction:

[0078] The epithelial layer (surface to the upper edge of the basement membrane band), the basement membrane band (a high-contrast thin layer, approximately 50–80 μm thick), and the lamina propria (down to a depth of 500 μm). This layering information serves as prior input for subsequent texture analysis.

[0079] S4. Perform depth-adaptive texture heterogeneity analysis on the three-dimensional tissue structure image to obtain a probability distribution map of the ablation target area;

[0080] In step S4, the depth-adaptive texture heterogeneity analysis includes: dividing the three-dimensional tissue structure image into three regions along the depth direction: epithelial layer, basement membrane zone, and lamina propria; calculating local binary pattern texture features for each region; and then weighting and fusing them according to the region depth index, wherein the weight factor of the basement membrane zone region is set to twice that of the epithelial layer, to obtain the probability distribution map of the ablation target area.

[0081] In the weighted fusion of the regional depth index, the weight of each region is calculated according to the formula w(d)=w0⋅e -βd The weight of superficial abnormal proliferation decreases exponentially with depth d, where w0 is the initial weight of the epithelial surface and β is the decay coefficient, which makes the weight of superficial abnormal proliferation higher than that of deep normal tissue, thus highlighting the probability response of early cancerous areas.

[0082] Furthermore, abnormal proliferation (such as oral precancerous lesions and early squamous cell carcinoma) often first manifests as increased nuclear atypia in the epithelial layer, rupture or disorder of the basement membrane band, and changes in the arrangement of collagen fibers.

[0083] Implementation details: For each voxel in the 3D image, a local sub-block (size adjusted according to depth) is extracted centered on that voxel. Local Binary Pattern (LBP) operator is used to extract texture features: for each pixel within the sub-block, its grayscale is compared to the grayscale of the center pixel; higher grayscale is recorded as 1, otherwise 0, forming a binary code, and a histogram is calculated as the feature vector. However, simple LBP cannot adapt to different histological scales at different depths. Therefore, a depth-adaptive strategy is adopted: in the epithelial layer (depth 0–200 μm), a small neighborhood radius (r=1 pixel, corresponding to approximately 10 μm) is used to capture subtle morphological changes in the cell nucleus; in the basement membrane zone (depth 200–280 μm), a medium radius (r=2 pixels) is used to identify basal cell polarity disorder; in the lamina propria (depth >280 μm), a large radius (r=3 pixels) is used to assess the degree of orientation disorder of collagen bundles.

[0084] The responses of each depth region are fused using an exponentially decaying weighted fusion: w(d) = w0·exp(-βd), where d is the depth (μm), w0 = 1.0 (epithelial surface), and β = 0.003 μm. -1 (This makes the basement membrane band weight approximately 0.55 times that of the epithelial layer;)

[0085] Therefore, a dual-weighting approach was adopted: first, the original feature response value Rraw(d) was calculated, and then multiplied by a depth sensitivity factor k(d) = exp(-βd) for attenuation; finally, the response value of the basement membrane zone was additionally multiplied by an enhancement factor 2, i.e., the final probability P(d) = k(d)·(1+δ{BM})·Rraw(d), where δ{BM} is 1 within the basement membrane zone. This approach reflects both the natural attenuation of deep signals and highlights the pathological changes in key layers.

[0086] After thresholding (e.g., >0.6) and connected component filtering (to remove isolated noise) of the probability values ​​of all voxels, an ablation target probability distribution map is generated. The grayscale value of 0 to 1 indicates the probability that the location needs to be ablated.

[0087] β was obtained by fitting the depth attenuation curve of OCT images of normal oral tissue. Tongue dorsum images from 20 normal volunteers were used, and the average signal intensity of each depth layer (every 10 μm layer) was calculated. An exponential attenuation model was used for fitting, resulting in a mean value of 0.0028 μm for β. -1 The standard deviation is 0.0003, and 0.003 is taken as a fixed value. To improve adaptability, β for the current patient can also be automatically calculated during system initialization (by measuring the signal attenuation from the epithelial surface to the lamina propria in lesion-free areas).

[0088] The radius r is chosen to match the OCT lateral resolution (10 μm / pixel), ensuring that r=1 corresponds to 10 μm (comparable to the diameter of epithelial cells), r=2 corresponds to 20 μm (basal cell clusters), and r=3 corresponds to 30 μm (collagen fiber bundle cycle).

[0089] These parameters can be determined by comparing with pathological images: digitize tissue sections from the same location, calculate the correlation between their LBP characteristics and pathological grading, and select the optimal radius combination.

[0090] S5. Based on the probability distribution map of the ablation target area, plan the ablation energy timing and needle path to obtain the dynamic ablation parameter sequence;

[0091] In step S5, the planning of ablation energy timing and needle path further includes: thermal damage avoidance assessment: based on the density of healthy tissue within 2 mm around each candidate point in the probability distribution map of the ablation target area, the virtual thermal diffusion priority is calculated, the needle path sequence that makes the thermal accumulation temperature of the surrounding healthy tissue below 42℃ is selected, and the interval time of adjacent pulses is adjusted to match the local blood perfusion time constant to obtain the dynamic ablation parameter sequence.

[0092] The virtual thermal diffusion priority is determined by comparing the thermal influence radius of each candidate point with the distance to adjacent important structures, including nerve bundles or glands. If the distance is less than 0.5 mm, the ablation power of that point is automatically reduced to 60% of the rated value, and the pulse interval is extended to 1.5 times the original interval.

[0093] Further, step S5 transforms the probability map into executable ablation commands. The ablation actuator is a miniature radio frequency or laser probe that can be inserted submucosally or onto a contact surface. The goal of dynamic programming is to minimize thermal damage to surrounding healthy tissue while completely ablating the target area.

[0094] First, a 3D mesh of the probability map is generated (voxel size consistent with OCT resolution). Each voxel is assigned an ablation necessity score (probability value).

[0095] Then, a greedy path planning method is adopted: the voxel with the highest probability value that has not yet been ablated is selected as the first ablation point, and its "thermal damage avoidance priority" is calculated - that is, the volume ratio of healthy tissue (probability <0.2) within a radius of 2 mm centered on the point is evaluated.

[0096] If the probability exceeds 30%, the priority of that point is reduced, and the next highest probability point is selected instead. This process is repeated until all high-probability regions (>0.5) are covered by at least one ablation point.

[0097] For each selected ablation point, the needle tract angle is determined according to its depth (surface contact ablation is used for superficial lesions, while needle tract ablation is required for deep lesions).

[0098] The ablation energy timing parameters include: pulse power (W), pulse duration (ms), and pulse interval (ms). During implementation, if the ablation point is less than 0.5 mm from important structures (nerve bundles, glands) (by comparison with a pre-imported oral anatomy atlas), the power is reduced to 60% of the rated value (e.g., reduced from rated 10W to 6W), and the interval is extended by 1.5 times (e.g., from 100ms to 150ms).

[0099] Simultaneously, a local blood flow perfusion time constant τ is introduced (estimated by measuring the microvessel density near the ablation point using OCT angiography, typically 0.5–2 s). The interval between adjacent pulses is set to 0.2 times τ to ensure that heat has sufficient time to be carried away by blood flow and does not accumulate.

[0100] The final generated dynamic ablation parameter sequence is a list, with each row containing: spatial coordinates (x, y, z), power, pulse width, pulse interval, and total number of pulses. This sequence can be dynamically adjusted based on real-time thermal feedback during the procedure.

[0101] S6. Output the ablation energy according to the dynamic ablation parameter sequence and simultaneously collect the pyroelectric signal to obtain the real-time ablation response waveform;

[0102] In step S6, the step of outputting ablation energy and simultaneously acquiring pyroelectric signals involves recording the peak amplitude and time constant of the pyroelectric signal at a period of 10 milliseconds, and comparing the peak amplitude with a preset ablation depth-amplitude calibration curve. If the measured peak amplitude is lower than the lower limit of the curve, the energy density of subsequent pulses is automatically increased by 5%; if it is higher than the upper limit of the curve, the energy density is reduced by 3%, thus obtaining the closed-loop corrected ablation real-time response waveform.

[0103] Furthermore, during ablation, the system sends the parameter sequence to the energy generator in real time. Simultaneously, a miniature thermocouple (diameter <0.3mm) is integrated at the tip of the ablation probe to acquire pyroelectric signals (temperature change rate) with a period of 10 milliseconds.

[0104] After amplification and filtering, the pyroelectric signal yields the real-time ablation response waveform, which has the following typical characteristics: the rising slope is the rate of temperature increase over time (°C / ms), the peak amplitude is the maximum temperature rise (°C), and the time constant is the time required for the temperature to drop from the peak to 37°C (ms).

[0105] The core of step S6 is closed-loop correction control. The system has pre-calibrated an "ablation depth-peak amplitude" curve through in vitro experiments: at different ablation energies (2–15 J), the necrosis depth of isolated pork tissue after ablation is measured and the corresponding pyroelectric peak amplitude is recorded.

[0106] Store the curve in a lookup table. During treatment, immediately read the measured peak amplitude after each pulse is output.

[0107] If the measured amplitude is lower than the lower limit of the curve (for example, the amplitude corresponding to the expected ablation depth of 1 mm is 0.8℃ / ms, but the measured amplitude is 0.6℃ / ms), the energy density of the subsequent pulse will be automatically increased by 5% (by increasing the power or extending the pulse width).

[0108] If the measured amplitude exceeds the upper limit (e.g., 1.1℃ / ms), it is reduced by 3%. This adaptive control can cope with the differences in tissue water content and blood flow among different individuals, ensuring the consistency and safety of ablation depth.

[0109] Simultaneously, the time constant is used to determine whether carbonization has occurred: if the time constant is abnormally short (<20ms), it indicates that the tissue has dried and carbonized, and ablation at that point should be stopped immediately and an alarm should be triggered. All real-time response waveforms, along with the parameter sequence, are recorded to form a treatment log.

[0110] Specifically, the method further includes the following steps: displaying the three-dimensional tissue structure image, the probability distribution map of the ablation target area, and the real-time ablation response waveform on the same three-dimensional oral cavity model in a semi-transparent overlay manner, wherein the ablated area is marked with a red gradient and the area to be ablated is marked with a blue gradient, and the ablation progress percentage is updated in real time;

[0111] Furthermore, to facilitate doctor monitoring, the system renders a 3D oral cavity model on the touchscreen, overlaying the probability distribution map obtained in step four as a semi-transparent color cloud map. Ablated areas (points confirmed by pyroelectric feedback to have completed pulse output and reached the target temperature) are marked with a red gradient, while areas to be ablated are marked with a blue gradient. Simultaneously, the system displays the current ablation progress percentage and pyroelectric waveform curve in real time, supporting intraoperative manual intervention (pausing or modifying parameters).

[0112] As can be seen from the above, the present invention uses phase-amplitude joint displacement estimation to extract depth direction micro-displacement by phase difference and amplitude centroid offset to extract lateral displacement. It can effectively compensate for non-rigid tissue peristalsis caused by swallowing, tongue movement, etc. without relying on external markers or high-frequency frame rates. This ensures that the reconstructed three-dimensional image and the ablation target area are not distorted due to movement, thus improving the targeting consistency of detection and treatment.

[0113] Texture analysis at different scales was performed along the oral mucosal epithelium, basement membrane zone, and lamina propria, and depth adaptive weights were assigned based on the differences in pathological sensitivity of each layer to highlight the probabilistic response of early cancerous areas. At the same time, virtual thermal diffusion priority was introduced into the ablation planning, and energy parameters and pulse intervals were automatically adjusted based on the density of healthy tissue around the target and the distance from important anatomical structures (such as nerve bundles and glands), which enabled selective ablation of lesions and minimized thermal damage to surrounding healthy tissues.

[0114] like Figure 2 As shown, in one embodiment, the present invention also provides an oral ablation detection system, comprising:

[0115] The oral tissue sampling module is used to acquire the raw interference spectrum signal of oral tissue to obtain the raw interference signal.

[0116] The motion compensation module is used to perform motion compensation on the original interference signal based on tissue peristalsis characteristics to obtain a compensated interference signal;

[0117] The image reconstruction module is used to perform frequency domain transformation and scattering correction on the compensated interference signal to obtain a three-dimensional tissue structure image;

[0118] The lesion analysis module is used to perform depth-adaptive texture heterogeneity analysis on the three-dimensional tissue structure image to obtain a probability distribution map of the ablation target area;

[0119] The ablation planning module is used to plan the ablation energy timing and needle path based on the probability distribution map of the ablation target area, and obtain a dynamic ablation parameter sequence.

[0120] The ablation execution module is used to output ablation energy according to the dynamic ablation parameter sequence and simultaneously acquire pyroelectric signals to obtain the real-time ablation response waveform.

[0121] The beneficial effects and technical effects of the same oral ablation detection method embodiment will not be repeated here.

[0122] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for detecting oral cavity ablation, characterized in that, include: The raw interference spectrum signal of oral tissue was acquired to obtain the raw interference signal; Motion compensation based on tissue peristalsis characteristics is performed on the original interference signal to obtain a compensated interference signal; The compensated interference signal is subjected to frequency domain transformation and scattering correction to obtain a three-dimensional tissue structure image; Depth-adaptive texture heterogeneity analysis was performed on the three-dimensional tissue structure image to obtain a probability distribution map of the ablation target area; Based on the probability distribution map of the ablation target area, the ablation energy timing and needle path are planned to obtain the dynamic ablation parameter sequence; The ablation energy is output according to the dynamic ablation parameter sequence, and the pyroelectric signal is collected simultaneously to obtain the real-time ablation response waveform.

2. The oral ablation detection method according to claim 1, characterized in that, The motion compensation steps include: The two consecutive frames of raw interference signals are converted into complex domain signals, and the amplitude and phase components are extracted. The displacement in the depth direction is calculated using the phase difference, and the lateral displacement is calculated using the amplitude centroid offset, thus obtaining the displacement vector of each pixel. The displacement vector is then interpolated using thin-plate splines to generate a dense deformation field. Finally, the dense deformation field is used to reverse-sample the next frame signal to obtain a compensated interference signal.

3. The oral ablation detection method according to claim 2, characterized in that, The depth displacement and lateral displacement are calculated using the following formulas: Where ϕn,ϕn+1 are the phases of the complex signals of two adjacent frames, |An|, |An+1| are the corresponding amplitudes, λ is the center wavelength of the light source, n is the average refractive index of the oral tissue, (x0,z0) is the current pixel coordinates, and the horizontal integration range covers a window centered at x0 with a width of 2Wx, where Wx is greater than the half-width Wz of the depth direction window.

4. The oral ablation detection method according to claim 1, characterized in that, The depth-adaptive texture heterogeneity analysis includes: dividing the three-dimensional tissue structure image into three regions along the depth direction: epithelial layer, basement membrane zone, and lamina propria; calculating local binary pattern texture features for each region; and then weighting and fusing them according to the region depth index, wherein the weight factor of the basement membrane zone region is set to twice that of the epithelial layer, to obtain the probability distribution map of the ablation target area.

5. The oral ablation detection method according to claim 4, characterized in that, In the weighted fusion of the regional depth index, the weight of each region is calculated according to the formula w(d)=w0⋅e -βd The weight of superficial abnormal proliferation decreases exponentially with depth d, where w0 is the initial weight of the epithelial surface and β is the decay coefficient. This makes the weight of superficial abnormal proliferation higher than that of deep normal tissue, thus highlighting the probability response of early cancerous areas.

6. The oral ablation detection method according to claim 1, characterized in that, The planned ablation energy timing and needle path further include: thermal damage avoidance assessment: based on the density of healthy tissue within 2 mm around each candidate point in the probability distribution map of the ablation target area, the virtual thermal diffusion priority is calculated, the needle path sequence that makes the thermal accumulation temperature of the surrounding healthy tissue below 42℃ is selected, and the interval time between adjacent pulses is adjusted to match the local blood perfusion time constant to obtain the dynamic ablation parameter sequence.

7. The oral ablation detection method according to claim 6, characterized in that, The virtual thermal diffusion priority is determined by comparing the thermal influence radius of each candidate point with the distance to adjacent important structures, including nerve bundles or glands. If the distance is less than 0.5 mm, the ablation power at that point is automatically reduced to 60% of the rated value, and the pulse interval is extended to 1.5 times the original interval.

8. The oral ablation detection method according to claim 1, characterized in that, In the step of outputting ablation energy and simultaneously acquiring pyroelectric signals, the peak amplitude and time constant of the pyroelectric signal are recorded with a period of 10 milliseconds. The peak amplitude is compared with the preset ablation depth-amplitude calibration curve. If the measured peak amplitude is lower than the lower limit of the curve, the energy density of the subsequent pulse is automatically increased by 5%. If it is higher than the upper limit of the curve, the energy density is reduced by 3%, thus obtaining the ablation real-time response waveform after closed-loop correction.

9. The oral cavity ablation detection method according to claim 1, characterized in that, The method further includes the following steps: displaying the three-dimensional tissue structure image, the probability distribution map of the ablation target area, and the real-time ablation response waveform on the same three-dimensional oral cavity model in a semi-transparent overlay manner, wherein the ablated area is marked with a red gradient and the area to be ablated is marked with a blue gradient, and the ablation progress percentage is updated in real time.

10. An oral ablation detection system, characterized in that, include: The oral tissue sampling module is used to acquire the raw interference spectrum signal of oral tissue to obtain the raw interference signal. The motion compensation module is used to perform motion compensation on the original interference signal based on tissue peristalsis characteristics to obtain a compensated interference signal; The image reconstruction module is used to perform frequency domain transformation and scattering correction on the compensated interference signal to obtain a three-dimensional tissue structure image; The lesion analysis module is used to perform depth-adaptive texture heterogeneity analysis on the three-dimensional tissue structure image to obtain a probability distribution map of the ablation target area; The ablation planning module is used to plan the ablation energy timing and needle path based on the probability distribution map of the ablation target area, and obtain a dynamic ablation parameter sequence. The ablation execution module is used to output ablation energy according to the dynamic ablation parameter sequence and simultaneously acquire pyroelectric signals to obtain the real-time ablation response waveform.

Images