PS-OCT-based tooth lesion region positioning and segmentation method

Abstract

The application provides a PS-OCT-based tooth lesion region positioning and segmentation method. The segmentation method can automatically segment tooth different degree caries lesion regions and quantitatively calculate the lesion regions. The tooth sample data without noise is obtained through an intensity-based denoising registration method, the initial profile center coordinates of the segmented lesion regions are automatically determined, a relatively accurate lesion region segmentation and quantitative calculation method is realized, and the method is used for doctors to diagnose the tooth damage degree of patients and adopt a certain intervention mode, and the ability of early caries lesion diagnosis and identification is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of dental lesion segmentation detection, in particular to a dental lesion region positioning and segmentation method based on PS-OCT. BACKGROUND

[0002] In the diagnosis of dental diseases, the diagnosis and identification of early lesions of dental caries is a challenging problem. For the occlusal surface of the tooth, bacteria are prone to breed in the pits and fissures, forming dental caries. If not treated in time, the shallow lesions will worsen, forming occlusal surface decay and other symptoms. When intervention is needed, most of the dental caries and healthy tissue need to be removed, which damages the mechanical integrity of the tooth. Therefore, clinicians must detect and characterize these lesions as early as possible in order to prevent or reverse the lesions through non-surgical methods.

[0003] Currently, the tools for clinically evaluating dental caries lesions on the occlusal surface have the problem of not being able to detect early caries lesions or needing to destroy the integrity of the tooth through sectioning experiments. Therefore, to accurately determine the severity of the lesion, a new and more complex diagnostic tool is needed to more accurately characterize early caries lesions. Optical coherence tomography (OCT) is a non-invasive, non-invasive, non-ionizing, and non-radiation diagnostic method that has been widely studied in the diagnosis of hard dental tissues, especially in the detection of dental caries. Polarization sensitive optical coherence tomography (PS-OCT) is an extension of OCT. In addition to the traditional intensity signal, it can also provide a variety of polarization information with birefringence characteristics, such as phase retardation, optical axis, and degree of polarization uniformity (DOPU). Since dental caries changes the optical properties of enamel and dentin during demineralization or caries, the main component of enamel, hydroxyapatite (HAP), is a birefringent crystal, and demineralization changes the birefringence characteristics. Combining the birefringence characteristics of enamel, researchers have shown that the degree of depolarization can clearly distinguish proximal caries, i.e. DOPU can be used to characterize the severity of dental caries.

[0004] Although PS-OCT imaging has attractive advantages, there is no automatic lesion identification algorithm and quantitative evaluation method for the DOPU parameter of PS-OCT. SUMMARY

[0005] The purpose of the present application is to provide a dental lesion region positioning and segmentation method based on PS-OCT, which can accurately and automatically identify and quantitatively evaluate dental caries and demineralization regions, and provide better help for dentists in the diagnosis and treatment process.

[0006] This invention is implemented as follows:

[0007] A method for locating and segmenting dental lesion areas based on PS-OCT includes the following steps:

[0008] (a) Acquire PS-OCT data;

[0009] (b) Preprocess the PS-OCT data;

[0010] (c) Perform averaging kernel operation using a Gaussian sliding window;

[0011] (d) Calculate polarization uniformity data;

[0012] (e) Perform intensity registration-based processing on the polarization uniformity data;

[0013] (f) Extract the center coordinates of the lesion area;

[0014] (g) Perform pseudo-color image processing on the polarization uniformity data processed in step (e);

[0015] (h) The lesion region is segmented by automatically extracting the center coordinates of the lesion region and using the energy functional method, and the segmentation contour curve and the binary map of the lesion region are displayed.

[0016] (i) Calculate the area of ​​the lesion region using the binary map from step (h).

[0017] In step (a) of this invention, the acquisition of PS-OCT data specifically involves: First, setting the data acquisition size of the PS-OCT system to Z*X*Y using commercial software, with units of Pixel*Pixel*time (the scan size is set according to the lesion area of ​​the tooth, where Z represents the number of volume pixels scanned in the Z direction for each A-scan, X represents the number of A-scans continuously acquired along the X direction to obtain a B-scan along the XZ plane, and Y represents the number of times the scan is repeated at the same position), and the volume pixel size in each scanning direction of Z and X corresponds to ΔZ. Pixel and ΔX Pixel The unit is mm. Then, the polarization function of PS-OCT is used to repeatedly scan the lesion location of the tooth to obtain Y B-scan two-dimensional scan data, with the scan locations as shown... Figure 3 (a) As shown by the white line.

[0018] In step (b) of this invention, the B-scan cross-sectional data file collected in step (a) is preprocessed, that is, the polarization processing program provided by the system is used to read the data, thereby obtaining polarization intensity data I, CCD data, and original Stokes parameters Q, U, and V data.

[0019] In step (c) of the present application, the obtained Stokes parameters Q, U and V are respectively subjected to m*m size Gaussian average kernel sliding window operation, to obtain high-contrast Q1, U1 and V1 data.

[0020] In step (d) of the present application, the polarization uniformity data DOPU after average kernel processing is calculated by the formula

[0021] In step (e) of the present application, the establishment of sample information data is specifically: obtaining the position of the sample data by a registration denoising method based on intensity data. The implementation process of the method is shown in the left flowchart of Figure 3

[0022] First step: data normalization is performed on the preprocessed intensity data I to obtain intensity data I1;

[0023] Second step: speckle noise is removed from the intensity data I1 processed in the first step by using a lee filter to obtain filtered intensity data I2;

[0024] Third step: threshold segmentation is performed, a threshold range [T1, T2] is selected, background noise removal of the entire B-scan is performed, the sample is highlighted, and intensity data I3 is obtained; T1 and T2 are set parameters;

[0025] Fourth step: through iterative operation on the intensity data I3, the position of the data with non-zero air segment is screened, the data at this position is set to zero to remove the unremoved air segment noise, that is, the background noise removal is performed on the region from the starting position of each A-scan of the B-scan data to the sample, so as to remove the noise of the air segment between the tooth and the air, and intensity data I4 is obtained;

[0026] Fifth step: the over-segmented region is compensated by using linear interpolation compensation. When threshold segmentation is performed, the positions with small intensity values in the sample are set to zero. In order to ensure the integrity of the sample information, linear interpolation compensation is performed, and compensated intensity data I5 is obtained;

[0027] Sixth step: find function is used to obtain the data position with non-zero intensity data I5, and the polarization uniformity data is denoised and registered to obtain noise-free polarization uniformity data DOPU1.

[0028] ​​In step (f) of the present application, firstly, the single A-scan data of the polarization uniformity data DOPU1 without noise in step (e) is averaged, and the A-scan averaging operation is repeated along the X scanning direction to obtain a one-dimensional array in the X direction. Then, the positions of the peak and valley are extracted by peak retrieval, and the median is taken to determine the lesion segmentation region X scanning direction coordinate x0. Then, according to the value of DOPU1, the position z0 where the median in the range (0, T0) is extracted, and finally the center coordinate [z0, x0] of the lesion segmentation region is obtained. T0 is a parameter set in advance.

[0029] In step (g) of the present application, the polarization uniformity data DOPU1 without noise in step (e) is subjected to pseudo-color map setting and automatically saved.

[0030] In step (h) of the present application, the pseudo-color map saved in step (g) is read, and the lesion segmentation is performed and displayed based on the automatically extracted lesion center coordinate and the energy functional method. The specific steps are as follows:

[0031] First step: reading the saved pseudo-color map and converting it into a gray-scale map;

[0032] Second step: through the introduction of the minimum energy, and without relying on the selection of the initial curve and the boundary gradient of the image, the lesion recognition and contour extraction are performed, and then the parameters for extracting the contour are set: the iteration number, the time step, λ1 and λ2, the parameter epsilon for calculating the smooth Heaviside and dirac functions, the initial contour center coordinate point [z0, x0], and the initial contour radius R; finally, the initial contour curve u0 and the energy functional F(b1, b2, B) are determined.

[0033] In the automatic segmentation method, the introduced minimum energy functional formula is:

[0034] F(b1, b2, B) = μ·L(B) + ν·Area(in(B)) + λ1∫ in(B) |u0(z, x) - b1| 2 dz dx + λ2∫ out(B) |u0(z, x) - b2| 2 dz dx

[0035] Wherein, mu >= 0, v >= 0, lambda1 > 0, lambda2 > 0, are fixed setting parameters, the length L (B) of curve B and the area Area of internal area are constraint terms, in (B) is the internal area surrounded by curve B, out (B) is the external area surrounded by curve B, b1, b2 are the average value of u0 (z, x) in and out of in (B), when B is the boundary, the last two terms tend to be minimum, and the two terms in front are to limit the length and the area surrounded by the curve, so that the optimization of the lesion area is realized by improving the center coordinate positioning of the in (B) internal area of the curve B.

[0036] Third step: through iteration level set energy functional formula, until reaching convergence condition, thereby obtaining optimal boundary curve B.

[0037] Fourth step: set an empty matrix with the size of Z * X, first extract the position of the data less than or equal to zero in the segmentation curve B, then assign 1 to the corresponding position of the empty matrix, thereby obtaining a binary graph containing only the lesion area;

[0038] Fifth step: the position of the contour curve B is obtained, and the pseudo-color graph of the invention step (g) is used to draw the area with a colored curve function.

[0039] In step (i) of the present application, the number of non-zero data num in the binary graph is extracted using the binary graph of step (h), and the area is calculated according to the size of the pixel point, area = num * delta Z Pixel * delta X Pixel .

[0040] The present application combines intensity registration to remove the noise of polarization uniformity data, and proposes a tooth lesion area positioning segmentation method based on PS-OCT, which can accurately and automatically identify tooth caries, demineralization area, and quantitatively evaluate the lesion degree of tooth occlusal surface demineralization and caries, and can quantitatively analyze the area occupied by the tooth lesion area, to provide better help for dentists in the diagnosis and treatment process. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 It is the flow chart of the tooth lesion area positioning segmentation method based on PS-OCT of the present application.

[0042] Figure 2 It is the image data acquisition device used in the embodiment of the present application, wherein (a) is the experimental measurement process diagram of the in-vitro tooth, and (b) is the dental tray structure diagram.

[0043] Figure 3It is the pre-processing algorithm flow of intensity-based registration and OCT effect diagram in the embodiment of the application; wherein, (a) is the CCD diagram of the occlusal surface of the teeth collected; (b) is the normalized OCT original intensity diagram; (c) is the diagram processed by lee filtering; (d) is the threshold segmentation processing diagram; (e) is the air noise removal diagram; (f) is the interpolation compensation diagram; the scale is 500 μm.

[0044] Figure 4 It is the key step PS-OCT effect diagram of the tooth lesion area positioning segmentation method based on PS-OCT; wherein, (a) is the effect diagram after pre-processing based on intensity registration and denoising; (b) is the DOPU1 pseudo-color diagram obtained after registration and denoising; (c) is the binary diagram of the lesion area obtained after segmentation by the tooth lesion area positioning segmentation method based on PS-OCT proposed; (d) is the pseudo-color effect diagram of the final lesion area segmentation, the area surrounded by the middle white line is the lesion area; the scale is 500 μm.

[0045] Figure 5 It is the PS-OCT image of five tooth samples in the embodiment of the application. DETAILED DESCRIPTION

[0046] The technical solutions of the application will be described in detail below with specific embodiments. In the following description, many specific details are set forth in order to provide a thorough understanding of the application, but the application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the spirit of the application, therefore the application is not limited to the specific embodiments disclosed below. The embodiment of the application specifically describes the tooth lesion area positioning segmentation method based on PS-OCT.

[0047] When performing optical non-invasive tooth detection, due to the different structures and flatness of the occlusal surface of the teeth, it is necessary to perform scanning and collection according to the actual lesion condition to obtain the best imaging quality, thereby facilitating the later data processing.

[0048] The tooth lesion area positioning segmentation method based on PS-OCT provided by the application has the overall flow as shown in Figure 1 The specific steps include the following steps:

[0049] (a) Collecting two-dimensional B-scan cross-sectional data of the teeth based on PS-OCT.

[0050] Before the experiment, n extracted human teeth samples are prepared as experimental samples (5 in this experiment), which are: No. 1 (initial demineralization of pit and fissure caries occlusal surface point gap edge), No. 2 (slight discoloration demineralization of pit and fissure caries occlusal surface point gap), No. 3 (severe discoloration demineralization of caries occlusal surface point gap), No. 4 (shallow caries loss of pit and fissure caries occlusal surface), and No. 5 (severe caries loss of pit and fissure caries occlusal surface).

[0051] The data acquisition device adopts a polarization-sensitive optical coherence tomography system, as shown in Figure 2 , wherein Figure 2 (a) is a tooth in vitro experimental measurement process diagram, and (b) is a dental tray structure diagram. Before data acquisition, the teeth are fixed in the dental tray by modeling clay, the image size corresponding to the data acquisition is set to Z*X*Y by the PS-OCT, the unit is Pixel*Pixel*time, and the acquisition area is set through the CCD interface; the tooth sample is subjected to wet treatment through the dust-free paper and the ear bulb.

[0052] In the embodiment of the application, 5 teeth with different symptoms are used as measured objects, and a polarization-sensitive optical coherence tomography system is used to acquire two-dimensional B-scan cross-section data of different teeth in vitro, also known as PS-OCT data. Figure 3 In (a), a CCD image of the occlusal surface of the tooth collected is shown.

[0053] In (b), the PS-OCT data collected in (a) is preprocessed.

[0054] The B-scan cross-section data collected in (a) is preprocessed to obtain two-dimensional light intensity data I, CCD data, Stokes parameters Q, U, and V data, and save them in matrices A, B, C, D, and E, respectively.

[0055] In (c), the Stokes parameters Q, U, and V obtained are subjected to m*m size Gaussian average kernel sliding window operation to obtain Stokes parameters Q1, U1, and V1 with smaller contrast and noise, and save them in matrices C1, D1, and E1, respectively.

[0056] In (d), the data Q1, U1, and V1 are substituted into the formula to calculate the polarization uniformity data DOPU with obvious contrast, which is saved in the matrix Fdopu. The polarization uniformity data is processed by using a kernel with a size of [m, m].

[0057] In (e), sample data information is established by using a denoising matching method based on intensity data to obtain sample data information without background noise. The implementation process of the whole method is shown in Figure 3 . Figure 3In the middle, the left flow chart is the intensity-based registration preprocessing algorithm chart; the right is the PS-OCT effect chart of the whole process.

[0058] Step (e) specifically includes the following steps:

[0059] First step: normalize the intensity data I to [0, 1] through the normalization function mat2gray to obtain intensity data I1, and save it to matrix A1, as shown in Figure 3 (b) shows.

[0060] Second step: speckle smoothing and denoising of intensity data I1 through the lee filter to obtain intensity data I2, which is saved in matrix A2, as shown in Figure 3 (c) shows.

[0061] Third step: select a threshold range [T1, T2], and perform fixed threshold segmentation on intensity data I2 to obtain intensity data I3, which is saved in matrix A3, and remove the background noise of the entire two-dimensional B-scan cross-section data, highlighting the sample data, as shown in Figure 3 (d) shows.

[0062] Fourth step: through iterative operation on intensity data I3, filter out the positions of data that are not zero in the air segment, and set the data at these positions to zero to remove the air segment noise that has not been removed, that is, remove the background noise of the region from the starting position of each A-scan cross-section data of the two-dimensional B-scan cross-section data to the sample, thereby removing the noise of the air segment between the tooth and the air, obtaining intensity data I4, which is saved in matrix A4, as shown in Figure 3 (e) shows.

[0063] Fifth step: compensate the over-segmented region by linear interpolation compensation. When threshold segmentation is performed, the positions with smaller intensity values inside the sample will be set to zero. In order to ensure the integrity of the sample information, linear interpolation compensation is performed to obtain compensated intensity data I5, which is saved in matrix A5, as shown in Figure 3 (f) shows.

[0064] Sixth step: through the above steps, noise-free sample intensity data I5 is obtained, the find function is used to obtain the data positions of the sample intensity data I5 that are zero, the positions opposite to the positions of the sample intensity data I5 that are zero in the polarization uniformity data DOPU are found, and the data at the corresponding positions of DOPU are set to zero, to achieve denoising registration, obtaining noise-free polarization uniformity data DOPU1, which is saved in Fdopu1 matrix data, as shown in Figure 4 (a) shows.

[0065] (f) Firstly, average the single A-scan data of the polarization uniformity data DOPU1 in step (e), and repeat the average operation of the A-scan data along the X scanning direction to obtain a one-dimensional array in the X direction. Then, the positions of the peak and valley are extracted by peak retrieval, and the median is taken to determine the coordinate x0 of the lesion segmentation region in the X scanning direction. Then, according to the value of DOPU1, the position z0 where the median of the range (0, T0) is located is extracted, and finally the center coordinate [z0, x0] of the lesion segmentation region is obtained.

[0066] (g) By reading the denoised data DOPU1, the color bar of the set pseudo-color map is imported to generate a pseudo-color map with obvious contrast, and is automatically saved. Here, the pseudo-color map is saved as Figure 1, as shown in Figure 4 (b).

[0067] (h) By reading Figure 1, the lesion region is segmented and displayed based on the automatically extracted center coordinate of the lesion segmentation region and the energy functional algorithm, and the result is as shown in Figure 4 (c) and (d), (c) is a binary image, and (d) is a display effect diagram. The scale is 500 μm.

[0068] Step (h) specifically includes the following steps:

[0069] First step: convert the pseudo-color map of Figure 1 into a gray-scale image img1 by reading.

[0070] Second step: by introducing the way of minimizing energy, and not depending on the selection of initial curve and the boundary gradient of image, the lesion is identified and the contour is extracted, and then the parameters for extracting the contour are set: number = 500, timestep = 0.01, formula parameters λ1 = 1 and λ2 = 1, epsilon = 0, initial contour center coordinate point [z0, x0], initial contour radius R = 5, etc. Finally, the initial contour curve u0 and the energy functional F(b1, b2, B) are determined.

[0071] In the automatic segmentation method, the introduced minimum energy functional formula is:

[0072] F(b1, b2, B) = μ·L(B) + ν·Area(in(B)) + λ1∫ in(B) |u0(z,x)-b1| 2 dz dx + λ2∫ out(B) |u0(z,x)-b2| 2 dz dx

[0073] Where, μ≥0, v≥0, λ1>0, λ2>0, are fixed setting parameters, the length L(B) of curve B and the area Area of the internal region are constraint terms. in(B) is the internal region surrounded by curve B, out(B) is the external region surrounded by curve B, b1, b2 are the average values of u0(z, x) in and out of in(B), when B is the boundary, the last two terms in the above formula tend to be minimum. The first two terms are to limit the length and the area surrounded by the curve, so the optimization of the segmentation of the lesion area is realized by improving the way of positioning the center coordinates of the in(B) internal region of curve B.

[0074] Third step: through iterative evolution of the level set energy functional, until the convergence condition is reached, so as to obtain the optimal boundary curve B.

[0075] Fourth step: set an empty matrix Fdopu2 with a size of Z*X, then through the find function, the position of the data less than or equal to zero in the segmentation curve B is extracted, and the corresponding position of the empty matrix Fdopu2 is assigned a value of 1, so as to obtain a binary graph containing only the lesion area, as shown in Figure 4 (c).

[0076] Fifth step: through the pseudo-color map Figure 1 of step (g) of the present application, the region is drawn with a colored (white in the figure) curve function, as shown in Figure 4 (d).

[0077] (i) using the binary graph of step (h), extracting the number of non-zero data num in the binary graph, according to the pixel size ΔZ Pixel and ΔX Pixel in the scanning direction, the area is calculated as area=num*ΔZ Pixel *ΔX Pixel .

[0078] (j) through repeating steps (a)-(i), the segmentation process graph of five samples is obtained, as shown in Figure 5 . The final segmentation result is shown in Figure 5 , which shows that the method of the present application can well segment the healthy region and the lesion region of the tooth, and the segmentation effect is obvious, which also proves that the method has good robustness.

[0079] Figure 5In the image, (A) and (B) show scans from teeth numbered 1 and 2 with pits and fissures, respectively. (C) is a PS-OCT image obtained from tooth number 3, which is white and demineralized. (D) and (E) are PS-OCT images obtained from the carious infection areas of teeth numbered 4 and 5, respectively. Labels 1, 2, 3, and 4 represent the CCD image, the pre-processed and denoised intensity PS-OCT image, the pre-processed and calibrated denoised DOPU image, and the DOPU image of the lesion area obtained using the method of this invention, respectively. The scale bar is 500 μm.

[0080] This invention uses five isolated teeth with minor lesions as specific implementation samples. DOPU data from B-scan were acquired using a PS-OCT system. After employing a PS-OCT-based method for locating and segmenting the tooth lesion area, the following results were obtained: Figure 5 The segmentation effect based on polarization uniformity is shown. Figure 5 The CCD images (A1, B1, C1, D1, E1) show demineralization or caries characteristics of the occlusal surfaces of the teeth. In the five PS-OCT images (A2, B2, C2, D2, E2) labeled 2, which only contain intensity information, the strong reflections on the tooth surface are somewhat indistinguishable from lesion features. Therefore, it is difficult to determine whether the higher intensity value in this area is caused by a lesion or by strong reflection based solely on the intensity information. Figure 5 Within the same region of the five DOPU pseudocolor images (A3, B3, C3, D3, E3) labeled 3, significant differences in birefringence properties can be visualized using DOPU information. The actual region of interest is shown below. Figure 5 The five DOPU pseudo-color images (A4, B4, C4, D4, E4) labeled 4 are shown. It can be seen that the PS-OCT-based method for locating and segmenting dental lesions in this invention can effectively segment the lesion area and quantitatively calculate its area, providing doctors with a scientific diagnostic tool.

[0081] pass Figure 5 The segmentation results can be verified, showing that the segmentation algorithm proposed in this invention can segment not only connected lesions but also discontinuous lesions. Figure 5 In E4, two independent lesion areas are displayed.

[0082] The feasibility of the improved segmentation algorithm was also evaluated in this invention. Using the method described in this invention, B-scan scan data of other locations on five extracted teeth were obtained. Then, the same segmentation algorithm was used to obtain segmentation results for another five lesion areas. As can be seen from the figures, the segmentation effect is good. Therefore, the method of this invention has good feasibility.

[0083] Through the method in the application, the number of body pixels with data of 1 in the segmented binary image is calculated, and then the area of the lesion region is calculated by multiplying the number of body pixels by the actual physical size of each pixel point (the size of one body pixel in the application is set to 4.59*13 mu m 2 ), so as to finally obtain the values corresponding to the areas of the five regions, as shown in Table 1. Through the calculation of the area of the lesion region, the degree of the dental lesion of the patient can be more intuitively understood, so as to facilitate subsequent research and treatment.

[0084] Table 1: Lesion area of different occlusal surface teeth in two-dimensional cross-section

[0085] Tooth number Number 1 Number 2 Number 3 Number 4 Number 5 Area (mm 2 ) 0.1548 0.7366 0.5735 1.1762 0.6786

[0086] As can be seen from the above, the tooth lesion region positioning and segmentation method based on PS-OCT has great significance for accurately identifying and segmenting the tooth lesion region.

Claims

1. A PS-OCT-based tooth lesion area positioning segmentation method, characterized in that, It comprises the following steps: a. Collecting PS-OCT data of the tooth by a polarization sensitive optical coherence tomography system; b. Preprocessing the PS-OCT data collected in step a to obtain two-dimensional light intensity data I, CCD data, Stokes parameters Q, U, V data; c. Performing Gaussian average kernel sliding window operation on the Stokes parameters Q, U, V data obtained in step b to obtain Stokes parameters Q1, U1, V1 data; d. Calculate the polarization uniformity data DOPU by the formula DOPU = 1 - (1 - DOP) / 2 e. Obtaining the position of the sample data by a registration denoising method based on intensity data, establishing sample information data, and obtaining polarization uniformity data DOPU1 without noise; f. Averaging single A-scan data of DOPU1 data and repeatedly performing A-scan averaging along the X scanning direction to obtain a one-dimensional array in the X direction, then extracting the positions of peaks and valleys by peak retrieval, and taking the median to determine the lesion segmentation region X scanning direction coordinate x0; then extracting the position z0 where the median of DOPU1 value in the range (0, T0) is located, and finally obtaining the center coordinate [z0, x0] of the lesion segmentation region; g. Setting a false color map for the obtained polarization uniformity data DOPU1 and automatically saving it; h. Reading the false color map obtained in step g, and then performing lesion region segmentation by an energy functional method based on automatic extraction of lesion center coordinates, and displaying the segmentation contour curve and lesion region binary graph; i. Using the binary image of the lesion area from step (h), extract the number of non-zero data points num from the binary image, and then determine the pixel size ΔZ based on the scanning direction. Pixel With ΔX Pixel To calculate the area, area = num * ΔZ Pixel *ΔX Pixel .

2. The PS-OCT based tooth lesion region positioning segmentation method according to claim 1, characterized in that, Step e specifically comprises the following steps: e-1. Normalizing the intensity data I obtained in step b by a normalization function mat2gray to obtain intensity data I1; e-2. Smoothing and denoising the intensity data I1 by a lee filter to obtain filtered intensity data I2; e-3. Selecting a threshold range [T1, T2] to perform fixed threshold segmentation on the intensity data I2 to obtain intensity data I3; e-4. Iterating the intensity data I3 to filter the position of data not equal to zero in the air segment, setting the data at this position to zero to remove the air segment noise, i.e. the noise between the tooth and the air, to obtain intensity data I4; e-5. Compensating the sample information region excluded in step e-4 by linear interpolation compensation to obtain compensated intensity data I5; e-6. Using the find function to obtain the position of data not equal to zero in the intensity data I5, and denoising and registering the polarization uniformity data DOPU to obtain polarization uniformity data DOPU1 without noise.

3. The PS-OCT based dental lesion region positioning segmentation method according to claim 1, characterized in that, Step h specifically comprises the following steps: h-1. Reading the false color map in step g and converting it into a grayscale image; h-2. Identifying the lesion and extracting the contour by introducing a minimum energy; wherein the parameters for extracting the contour are set as: iteration number, time step, λ1 and λ2, setting the parameters epsilon for calculating the smooth Heaviside and dirac functions, the initial contour center coordinate point [z0, x0], and the initial contour radius R; finally determining the initial contour curve u0 and the energy functional F(b1, b2, B). The introduced minimum energy functional formula is: Wherein, μ≥0, v≥0, λ1>0, λ2>0, the length L(B) of curve B and the area Area of internal region are constraint terms; in(B) is the internal region surrounded by curve B, out(B) is the external region surrounded by curve B, b1, b2 are the average values of u0(z, x) in and out of in(B) respectively; h-3, carry out level set energy functional iterative evolution to obtain the final segmentation region curve B; h-4, set an empty matrix with a size of Z*X, extract the position of data less than or equal to zero in the segmentation curve B, then assign 1 to the corresponding position of the empty matrix, thereby obtaining a binary graph containing only the lesion region; h-5, obtain the position of the boundary curve B, and use the pseudo-color graph of step (g) to draw the lesion region with a colored curve function.

Images