Intraoperative assessment of tumor margin status

Abstract

The application relates to the medical technical field, in particular to a method for evaluating the benignity and malignancy of a tumor incision edge in operation, which comprises the following steps: obtaining optical coherence tomography data of historical incision edge tissue; extracting an A-line reflectivity curve based on the optical coherence tomography data of the historical incision edge tissue; calculating and extracting target features based on the A-line reflectivity curve, wherein the target features comprise depth features, slope features, distance features between local maximum values, intensity rapid fluctuation features, frequency features and optical attenuation coefficient features; training a preset classifier based on the target features and benignity and malignancy labels of the A-line reflectivity curve corresponding to the target features, so as to obtain a trained classifier; and inputting target features extracted based on optical coherence tomography data of to-be-detected incision edge tissue into the trained classifier to obtain an evaluation result. In this way, the accuracy and real-time performance of the evaluation of the benignity and malignancy of the tumor incision edge can be greatly improved.

Description

Technical Field

[0001] This invention relates to the field of medical technology, specifically to a method for assessing the benign or malignant nature of tumor margins during surgery. Background Technology

[0002] With the ever-expanding patient population for cancer, surgery has become a primary treatment method. In this process, the accuracy of intraoperative margin assessment is crucial for avoiding secondary surgeries, especially as it is a key factor in ensuring surgical success and patient recovery. Achieving negative intraoperative margins not only ensures complete tumor removal but also effectively reduces damage to healthy tissues, significantly improving postoperative quality of life and recovery speed. Therefore, enhancing the accuracy of intraoperative margin assessment is of paramount importance for the treatment of cancer patients.

[0003] Optical coherence tomography (OCT) is a cutting-edge non-invasive imaging technique that utilizes subtle changes in the light scattering properties of different tissues to reveal the three-dimensional microstructural details of tissues in a label-free, real-time manner. OCT possesses unique penetrating power, reaching approximately 2 mm below the tissue surface while maintaining a spatial resolution of 10-20 micrometers. This allows it to generate image quality similar to that of histopathology (histopathology resolution reaches 0.25 micrometers (40X), 0.5 micrometers (20X), while OCT resolution is 10-20 micrometers; currently, full-field OCT based on OCT technology achieves a resolution of 1 micrometer). Given that a 2 mm surgical margin is often used clinically as an important criterion for assessing negative tumor margins, OCT, with its high resolution and real-time imaging advantages, has become the optimal choice for intraoperative assessment of negative tumor margins.

[0004] OCT-based tumor assessment technology has been extensively studied and has shown a high correlation with histopathological examination. Currently, this technology is mainly used to train surgeons from different specialties to identify suspicious and non-suspicious areas at the edges of tumor specimens during surgery. However, due to the massive amount of OCT scan data acquired during surgery and the need for real-time determination of suspicious areas, relying solely on visual assessment by clinicians is both time-consuming and labor-intensive. Therefore, developing an algorithm that can automatically identify and label these suspicious areas can not only effectively improve assessment efficiency but also ensure accuracy. The application of such an algorithm will play a crucial role in the real-time guidance of tumor surgery. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a method for assessing the benign or malignant nature of tumor margins during surgery, so as to overcome the problems of poor accuracy and real-time performance in the current assessment of the benign or malignant nature of tumor margins during surgery.

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

[0007] This application provides a method for intraoperative assessment of the benign or malignant nature of tumor margins, including:

[0008] Acquire optical coherence tomography data of historical cutting edge tissue;

[0009] Based on the optical coherence tomography data of the historical cutting edge tissue, the A-line reflectance curve is extracted. The A-line reflectance curve is the curve of image intensity changing with the depth of the A-line.

[0010] Target features are calculated and extracted based on the A-line reflectance curve. These target features include depth features, slope features, distance features between local maxima, rapid intensity fluctuation features, frequency features, and optical attenuation coefficient features.

[0011] Based on the target features and the benign or malignant labels of the corresponding A-line reflectance curves, a preset classifier is trained to obtain the trained classifier.

[0012] The target features extracted from the optical coherence tomography data of the tissue under test are input into the trained classifier to obtain the evaluation results.

[0013] Furthermore, in some embodiments of this application, the depth feature includes the penetration depth of the effective signal range, and the process of extracting the penetration depth of the effective signal range includes:

[0014] The current A-line reflectance curve is smoothed to obtain the current A-line reflectance smoothed curve;

[0015] Determine the effective signal range of the current A-line reflectance smoothing curve;

[0016] The difference in depth between the start and end points of the effective signal interval is taken as the penetration depth of the effective signal interval of the current A-line reflectivity curve.

[0017] Furthermore, in some embodiments of this application, determining the effective signal range of the current A-line reflectance smoothing curve includes:

[0018] The depth position corresponding to the first time the image intensity reaches the target value in the current A-line reflectance smoothing curve is taken as the starting point of the effective signal interval;

[0019] The effective signal interval of the current A-line reflectivity smoothing curve is selected by using a preset sliding window. When the linear fitting slope of the signal within the window approaches 0 or the signal within the window is completely 0, the starting depth position within the current window is taken as the end point of the effective signal interval.

[0020] Furthermore, in some embodiments of this application, the slope feature includes: the average slope of the effective signal interval, the overall slope of the effective signal interval, and the overall intercept of the effective signal interval. The process of extracting the slope feature includes:

[0021] Multiple data subsets are obtained by sliding a preset sliding window to select the effective signal range of the current A-line reflectivity smoothing curve;

[0022] Calculate the linear fitting slope for each data subset separately, and take the average of the linear fitting slopes of all data subsets as the average slope of the effective signal range of the current A-line reflectance curve.

[0023] Calculate the linear fit slope and intercept of the effective signal interval of the current A-line reflectance smoothing curve. Use the slope as the overall slope of the effective signal interval of the current A-line reflectance curve and the intercept as the overall intercept of the effective signal interval of the current A-line reflectance curve.

[0024] Furthermore, in some embodiments of this application, the distance features between local maxima include: the mean distance and the standard deviation of the distance, and the process of extracting the distance features between local maxima includes:

[0025] Determine the local maximum value of image intensity within the effective signal range of the current A-line reflectance curve;

[0026] Calculate the depth difference corresponding to each adjacent local maximum value;

[0027] The mean of all the differences is used as the mean distance of the current A-line reflectance curve, and the standard deviation of all the differences is used as the standard deviation distance of the current A-line reflectance curve.

[0028] Furthermore, in some embodiments of this application, the rapid intensity fluctuation feature includes: the standard deviation of the rapid intensity fluctuation and the peak value of the rapid intensity fluctuation, and the process of extracting the rapid intensity fluctuation feature includes:

[0029] The current A-line reflectance curve is smoothed to obtain the current A-line reflectance smoothed curve;

[0030] Determine the difference in intensity rapid fluctuations between the current A-line reflectance curve and the current A-line reflectance smooth curve;

[0031] Calculate the standard deviation of the difference, which is used as the standard deviation of the rapid fluctuation of the intensity of the current A-line reflectance curve;

[0032] A difference curve is generated based on the difference, and the local maximum value with the largest absolute value in the difference curve is taken as the peak value of the rapid fluctuation of the intensity of the current A-line reflectance curve.

[0033] Furthermore, in some embodiments of this application, the frequency-type features include: a weighted average of the window frequency spectrum, the area of ​​the main peak of the window frequency spectrum, the number of main peaks in the window frequency spectrum, and the standard deviation of the window frequency spectrum. The process of extracting the frequency-type features includes:

[0034] By sliding through a preset sliding window, the current A-line reflectance curve can be selected to obtain multiple sets of A-line reflectance curve information;

[0035] For each set of A-line reflectivity curve information, a Fast Fourier Transform is used to transform it from the time domain to the frequency domain to determine the frequency spectrum of each frequency point in the set. The weighted average of the frequency spectrum of the current set of A-line reflectivity curve information is then calculated using the following formula:

[0036]

[0037] Where k is the label of each frequency point; N is the total number of frequency components; f k The frequency of point k; F(f) k ) is f k The frequency spectrum corresponding to the frequency;

[0038] The average of the frequency spectrum weighted average of all A-line reflectance curve information is used as the window frequency spectrum weighted average of the current A-line reflectance curve.

[0039] For each group of A-line reflectivity curve information, the target frequency points in the group whose frequency spectrum is greater than the first preset threshold are determined, and the frequency spectra corresponding to all target frequency points in the group are accumulated to obtain the main peak area of ​​the frequency spectrum of the current group of A-line reflectivity curve information; the average value of the main peak area of ​​the frequency spectrum of all groups of A-line reflectivity curve information is taken as the window frequency spectrum main peak area of ​​the current A-line reflectivity curve.

[0040] For each group of A-line reflectivity curve information, determine the number of frequency points in the frequency spectrum that are greater than the second preset threshold in that group, and obtain the number of main peaks in the frequency spectrum of the current group of A-line reflectivity curve information; take the average of the number of main peaks in the frequency spectrum of all groups of A-line reflectivity curve information as the number of main peaks in the window frequency spectrum of the current A-line reflectivity curve.

[0041] For each group of A-line reflectivity curve information, calculate the standard deviation of the frequency spectrum in that group to obtain the standard deviation of the frequency spectrum of the current group of A-line reflectivity curve information; use the average of the standard deviations of the frequency spectrum of all groups of A-line reflectivity curve information as the window standard deviation of the current A-line reflectivity curve.

[0042] Furthermore, in some embodiments of this application, the frequency-type features further include the weighted average of the overall frequency spectrum, the area of ​​the main peak of the overall frequency spectrum, the number of main peaks of the overall frequency spectrum, and the standard deviation of the overall frequency spectrum. The process of extracting the frequency-type features includes:

[0043] The current A-line reflectance curve is transformed from the time domain to the frequency domain using a Fast Fourier Transform to determine the frequency spectrum at each frequency point. The overall frequency spectrum weighted average of the current A-line reflectance curve is then calculated using the following formula:

[0044]

[0045] Where k is the label of each frequency point; N is the total number of frequency components; f k The frequency of point k; F(f) k ) is f k The frequency spectrum corresponding to the frequency;

[0046] Identify the target frequency points whose frequency spectrum is greater than the first preset threshold among all information of the current A-line reflectivity curve, and accumulate the frequency spectra corresponding to the target frequency points to obtain the overall frequency spectrum main peak area of ​​the current A-line reflectivity curve;

[0047] Determine the number of frequency points in the frequency spectrum of the current A-line reflectance curve that are greater than the second preset threshold, and obtain the total number of main peaks in the frequency spectrum of the current A-line reflectance curve.

[0048] Calculate the standard deviation of the frequency spectrum of all information in the current A-line reflectance curve to obtain the overall standard deviation of the frequency spectrum of the current A-line reflectance curve.

[0049] Furthermore, in some embodiments of this application, the optical attenuation coefficient type features include: attenuation coefficient and attenuation coefficient amplitude, and the process of extracting the optical attenuation coefficient type features includes:

[0050] By incorporating the axial confocal diffusion function and additive substrate noise of the OCT system into the optical attenuation model based on Beer's law, a tissue optical attenuation model based on the OCT system is obtained.

[0051] The optical coherence tomography (OCT) data of the historical cut edge tissue were nonlinearly fitted using the least squares method to determine the target parameters in the tissue optical attenuation model based on the OCT system.

[0052] The target parameters include the attenuation coefficient and the amplitude of the attenuation coefficient.

[0053] Furthermore, in some embodiments of this application, the tissue optical attenuation model based on the OCT system is as follows:

[0054]

[0055] Where I(z) is the light intensity value at depth z; z cf The focal plane is at a depth relative to the tissue boundary; z R A1 is the Rayleigh length; A1 is the OCT signal amplitude; µ t A1 is the attenuation coefficient; A2 is the additive base noise.

[0056] This application relates to the field of medical technology, specifically to a method for intraoperative assessment of the benign or malignant nature of tumor resection margins. The method includes: acquiring optical coherence tomography (OCT) data of historical resection margin tissues; extracting A-line reflectance curves based on the OCT data of historical resection margin tissues, where the A-line reflectance curve is a curve showing the change in image intensity with the depth of the A-line; calculating and extracting target features based on the A-line reflectance curves, including depth features, slope features, distance features between local maxima, rapid intensity fluctuation features, frequency-related features, and optical attenuation coefficient-related features; training a pre-defined classifier based on the target features and the benign or malignant labels of the corresponding A-line reflectance curves, resulting in a trained classifier; and inputting the target features extracted from the OCT data of the resection margin to be tested into the trained classifier to obtain the assessment result. This significantly improves the accuracy and real-time performance of tumor resection margin benign or malignant assessment. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 This is a schematic flowchart of the intraoperative tumor margin benignity / malignancy assessment method provided in an embodiment of the present invention;

[0059] Figure 2 These are OCT images from the intraoperative tumor margin benignity / malignancy assessment method provided in this embodiment of the invention;

[0060] Figure 3 This is a schematic diagram illustrating the principle of the intraoperative tumor margin benignity / malignancy assessment method provided in this embodiment of the invention;

[0061] Figure 4 This is a schematic diagram illustrating the extraction of depth features in the intraoperative tumor margin benignity / malignancy assessment method provided in this embodiment of the invention.

[0062] Figure 5 This is a schematic diagram illustrating the extraction of slope features in the intraoperative tumor margin benignity / malignancy assessment method provided in this embodiment of the invention.

[0063] Figure 6 This is a schematic diagram illustrating the extraction of distance features between local maximum values ​​in the intraoperative tumor margin benignity / malignancy assessment method provided in this embodiment of the invention.

[0064] Figure 7 This is a schematic diagram illustrating the extraction of rapid intensity fluctuation features in the intraoperative tumor margin benignity / malignancy assessment method provided in this embodiment of the invention.

[0065] Figure 8 This is a schematic diagram illustrating the extraction of frequency-based features in the intraoperative tumor margin benignity / malignancy assessment method provided in this embodiment of the invention.

[0066] Figure 9 These are typical representative images of six categories of features for benign and malignant samples in the intraoperative tumor margin benignity and malignancy assessment method provided in this embodiment of the invention;

[0067] Figure 10 This image shows the validation effect of evaluating a classifier trained using the intraoperative tumor margin benignity / malignancy assessment method provided in this application. Detailed Implementation

[0068] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0069] Figure 1 This is a flowchart illustrating the intraoperative tumor margin benignity / malignancy assessment method provided in this embodiment of the invention. Please refer to [link / reference]. Figure 1 This embodiment may include the following steps:

[0070] S101. Obtain optical coherence tomography data of historical cutting edge tissue.

[0071] S102. Based on the optical coherence tomography data of historical cutting edge tissue, extract the A-line reflectance curve.

[0072] It should be noted that the A-line reflectance curve mentioned in this application is consistent with the A-line reflectance curve in the prior art. It is a curve in which the image intensity changes with the depth of the A-line. That is, in the OCT imaging modality, multiple A-line reflectance curves generally form a reflectance profile of a B-scan, and the numerous B-scans within the scanning range form a three-dimensional reflectance volume with a cross-section of the en face.

[0073] S103. Calculate and extract target features based on A-line reflectance curve.

[0074] Specifically, in this application, the target features extracted based on the A-line reflectivity curve include depth features, slope features, distance features between local maxima, rapid intensity fluctuation features, frequency features, and optical attenuation coefficient features.

[0075] S104. Based on the target features and the benign or malignant labels of the corresponding A-line reflectance curves, train the preset classifier to obtain the trained classifier.

[0076] Specifically, in this application, the benign or malignant tumor can be labeled on historical cutting margin tissue or A-line reflectance curves based on existing information data. After the labeling is completed, the A-line reflectance curves extracted from the optical coherence tomography data of historical cutting margin tissues, and the target features extracted from the A-line reflectance curves, will have corresponding benign or malignant labels. Thus, the preset classifier is trained by calculating the determined target features and corresponding labels to obtain the trained classifier.

[0077] S105. Input the target features extracted from the optical coherence tomography data of the tissue to be tested into the trained classifier to obtain the evaluation results.

[0078] Using the same target feature extraction principle, features extracted from optical coherence tomography data of the tissue margin to be tested are input into the trained classifier to obtain the evaluation results, thereby achieving the assessment of the benign or malignant nature of the tumor margin during surgery.

[0079] The intraoperative tumor margin benignity / malignancy assessment method provided in this application acquires optical coherence tomography (OCT) data of historical surgical margin tissues. Based on this data, an A-line reflectance curve is extracted, which represents the change in image intensity with the depth of the A-line. Target features are calculated and extracted based on the A-line reflectance curve, including depth features, slope features, distance between local maxima, rapid intensity fluctuation features, frequency-related features, and optical attenuation coefficient-related features. A pre-defined classifier is trained based on the target features and their corresponding benign / malignancy labels on the A-line reflectance curves. The target features extracted from the OCT data of the surgical margin to be assessed are then input into the trained classifier to obtain the assessment result. This significantly improves the accuracy and real-time performance of intraoperative tumor margin benignity / malignancy assessment.

[0080] It should be noted that the number of A-line reflectance curves extracted from optical coherence tomography data based on historical cutting edge structures is generally multiple. Therefore, in this application, the set consisting of a large number of A-line reflectance curves can be referred to as the A-line reflectance curve set.

[0081] In some embodiments of this application, the A-line reflectance curve is extracted from a specific optical coherence tomography (OCT) scan based on historical edge tissue. This can be achieved by performing DC removal, windowing, dispersion correction, and Fourier transform on the original signal acquired by the frequency sweep OCT, thereby obtaining the A-line reflectance curve of the OCT.

[0082] In some embodiments of this application, the process of extracting and labeling A-line reflectance curves may include: firstly, extracting A-line reflectance curve sets from OCT images of normal tissue and cancerous tissue, respectively. The process for determining the benign or malignant label of the A-line curves may involve: firstly, an OCT specialist combining the OCT images (e.g., ...) Figure 2 As shown, the overall outline, gross structural regions, and microscopic regional features are matched one-to-one with the pathological H&E images; then, the pathologist labels the pathological H&E images; next, the OCT specialist labels the OCT images according to the labeling of the pathological H&E images; finally, the OCT specialist and the pathologist confirm the benign or malignant label of the A-line curve based on the labeling results of both parties (e.g., Figure 3 (As shown).

[0083] Based on this, the process of extracting each target feature in this application will be described in detail below.

[0084] First, the depth feature can specifically include the penetration depth of the effective signal interval. The process of extracting the penetration depth of the effective signal interval includes: smoothing the current A-line reflectivity curve to obtain a smoothed A-line reflectivity curve; determining the effective signal interval of the smoothed A-line reflectivity curve; and using the difference in depth between the start and end points of the effective signal interval as the penetration depth of the effective signal interval of the current A-line reflectivity curve (e.g., ...). Figure 4 (As shown).

[0085] And determine the effective signal range of the current A-line reflectance smoothing curve, which may include: taking the depth position corresponding to the first time the image intensity reaches the target value in the current A-line reflectance smoothing curve as the starting point of the effective signal range; selecting the part after the starting point of the effective signal range of the current A-line reflectance smoothing curve through a preset sliding window, and when the linear fitting slope of the signal in the window tends to 0 or the signal in the window is completely 0, taking the starting depth position of the current window as the ending point of the effective signal range.

[0086] In practical applications, the following operations are performed for each A-line reflectance curve:

[0087] First, the A-line reflectance curve is smoothed by using a Savitzky-Golay filter through polynomial fitting to process noise while preserving the peak and valley characteristics of the data, thus obtaining the corresponding smoothed A-line reflectance curve.

[0088] Then, the starting position of the effective signal, i.e., the starting point of the effective signal interval, is determined on the A-line reflectance smoothing curve. This includes: first reaching the target value, such as the maximum image intensity value of all signals in the current A-line reflectance smoothing curve. The location at 10% of the depth is used as the starting point of the effective signal range. Furthermore, the termination position of the effective signal, i.e., the end point of the effective signal interval, can be determined using a sliding, fixed-size window. In practical applications, the fixed length L of the sliding window can be set to 20% of the entire A-line reflectivity curve length, with a 50% overlap between the two coverage areas before and after the window. When the slope k of the linear fit of the signal within the window approaches zero, or when the signal across the entire window is completely zero, the starting position of the current window is determined as the end point of the effective signal interval. By determining the start and end points, the penetration depth of the effective signal interval is obtained. Formula 1 describes the specific calculation method for the penetration depth of the effective signal interval:

[0089] (1)

[0090] In the formula, D represents the penetration depth of the effective signal range, and d end The depth d represents the endpoint of the effective signal interval. start It indicates the depth of the starting point of the effective signal range.

[0091] Furthermore, the slope features include the average slope of the effective signal interval, the overall slope of the effective signal interval, and the overall intercept of the effective signal interval. For each A-line reflectance smoothing curve, the process of extracting slope features includes: selecting the effective signal interval of the current A-line reflectance smoothing curve through a preset sliding window to obtain multiple data subsets; calculating the linear fitting slope corresponding to each data subset, and using the average of the linear fitting slopes of all data subsets as the average slope of the effective signal interval of the current A-line reflectance curve; calculating the linear fitting slope and intercept of the effective signal interval of the current A-line reflectance smoothing curve, using the slope as the overall slope of the effective signal interval of the current A-line reflectance curve, and using the intercept as the overall intercept of the effective signal interval of the current A-line reflectance curve (e.g., ...). Figure 5 (As shown).

[0092] Specifically, for each A-line reflectance smoothing curve, the following operations are performed:

[0093] First, using the same principles as in the deep feature extraction described above, the A-line reflectance curve is smoothed to determine the effective signal range. Then, a sliding window is used to select a subset of data from the smoothed curve within each window. Based on this, the signal strength data of the smoothed curve within each window is analyzed. The slope is calculated using a linear fitting equation (i.e., Equation 2).

[0094] For a window of length L Equation 2 is obtained by performing a linear fit on a subset of data within the range, where k1-k n c1-c represents the slope of the linear fit for different window data (i.e., different subsets of data). n For the intercepts of different windows, d1-d n The data is for fitting, where d1=d stat The specific linear fit is shown in Equation 2:

[0095] (2)

[0096] After sliding through all valid signals in the window, the average value is calculated using Formula 3 to obtain the average slope of the valid signal range of the A-line reflectivity curve. The specific formula is shown below:

[0097] (3)

[0098] in, This represents the total number of times the window has been swiped.

[0099] Furthermore, for each A-line reflectance curve, all data of the overall effective signal interval of the A-line reflectance smoothing curve are directly fitted using Formula 4 to obtain the overall slope and overall intercept of the effective signal interval of the A-line reflectance curve.

[0100] (4)

[0101] Furthermore, the distance features between local maxima include the mean distance and the standard deviation of the distance. The process of extracting the distance features between local maxima includes: determining the local maxima of image intensity within the effective signal range of the current A-line reflectance curve; calculating the depth difference corresponding to each adjacent local maxima; using the mean of all differences as the mean distance of the current A-line reflectance curve, and using the standard deviation of all differences as the standard deviation of the distance of the current A-line reflectance curve (e.g., ...). Figure 6 (As shown).

[0102] Specifically, the following operations can be performed on each A-line reflectance curve:

[0103] Calculate the distance between adjacent local maxima of image intensity in the A-line reflectance curve, and obtain their mean and standard deviation.

[0104] In practical applications, the argrelextrema function from the SciPy library can be used to identify the depth positions d1-d of all local maxima of image intensity within the effective signal range of the A-line reflectance curve. m , where m is the number of local maxima. Then, the mean distance between the depth locations corresponding to the intensities of these adjacent images is calculated. , as expressed by Formula 5, where d i The depth location of the local maximum intensity of the i-th image, at a distance from the standard deviation. As shown in Formula 6:

[0105] (5)

[0106] (6)

[0107] Furthermore, the rapid intensity fluctuation characteristics include the standard deviation and peak value of the rapid intensity fluctuations. The process of extracting the rapid intensity fluctuation characteristics includes: smoothing the current A-line reflectance curve to obtain a smoothed current A-line reflectance curve; determining the difference in rapid intensity fluctuations between the current A-line reflectance curve and the smoothed current A-line reflectance curve; calculating the standard deviation of the difference as the standard deviation of the rapid intensity fluctuations of the current A-line reflectance curve; generating a difference curve based on the difference, and taking the local maximum value with the largest absolute value in the difference curve as the peak value of the rapid intensity fluctuations of the current A-line reflectance curve (e.g., peak value). Figure 7 (As shown).

[0108] Specifically, the following operations can be performed on each A-line reflectance curve:

[0109] First, the image intensity in the A-line reflectance curve is calculated. Image intensity in the corresponding A-line reflectance smoothing curve The difference in intensity between the two values ​​is used to eliminate the slow intensity changes caused by normal light attenuation, as shown in Equation 7:

[0110] (7)

[0111] Where N represents the length of the A-line reflectance curve or the A-line reflectance smooth curve (both are the same), which is the number of pixels in the image depth.

[0112] Then, calculate these differences according to Formula 8. Standard deviation :

[0113] (8)

[0114] in The difference The average value.

[0115] Use Formula 9 to find the difference curve (based on the difference). And corresponding depth generation, i.e. Figure 7 The local maximum value with the largest absolute value in the rapid intensity fluctuation curve shown in the figure is taken as the peak value of the rapid intensity fluctuation, as shown in Formula 9 below:

[0116] (9)

[0117] Furthermore, frequency-related features include the weighted average of the window frequency spectrum, the area of ​​the main peak in the window frequency spectrum, the number of main peaks in the window frequency spectrum, and the standard deviation of the window frequency spectrum. The process of extracting frequency-related features includes:

[0118] For the window frequency spectrum weighted average: the current A-line reflectivity curve is selected through a preset sliding window to obtain multiple sets of A-line reflectivity curve information; for each set of A-line reflectivity curve information, it is transformed from the time domain to the frequency domain through a fast Fourier transform to determine the frequency spectrum of each frequency point in the set, and the frequency spectrum weighted average of the current set of A-line reflectivity curve information is calculated; the average of the frequency spectrum weighted averages of all sets of A-line reflectivity curve information is used as the window frequency spectrum weighted average of the current A-line reflectivity curve.

[0119] For the main peak area of ​​the window frequency spectrum: For each group of A-line reflectance curve information, determine the target frequency points in the group whose frequency spectrum is greater than the first preset threshold, and accumulate the frequency spectra corresponding to all target frequency points in the group to obtain the main peak area of ​​the frequency spectrum of the current group of A-line reflectance curve information; the average value of the main peak areas of the frequency spectrum of all groups of A-line reflectance curve information is used as the main peak area of ​​the window frequency spectrum of the current A-line reflectance curve.

[0120] Regarding the number of main peaks in the frequency spectrum of the window: For each group of A-line reflectance curve information, determine the number of frequency points in the frequency spectrum that are greater than the second preset threshold in the group, and obtain the number of main peaks in the frequency spectrum of the current group of A-line reflectance curve information; take the average of the number of main peaks in the frequency spectrum of all groups of A-line reflectance curve information as the number of main peaks in the window frequency spectrum of the current A-line reflectance curve.

[0121] For the standard deviation of the frequency spectrum within the window: For each group of A-line reflectance curves, calculate the standard deviation of the frequency spectrum within that group to obtain the standard deviation of the frequency spectrum for the current group of A-line reflectance curves; use the average of the standard deviations of the frequency spectrum across all groups of A-line reflectance curves as the standard deviation of the window frequency spectrum for the current A-line reflectance curve (e.g., ...). Figure 8 (As shown).

[0122] Specifically, the following operations are performed for each A-line reflectance curve:

[0123] First, the signal of each window (i.e., by sliding the A-line reflectivity curve through a sliding window to obtain information data of multiple windows, where the information data selected by each window is a set of A-line reflectivity curve information, that is, the signal of that window) is transformed from the time domain to the frequency domain by using Fast Fourier Transform, the amplitude of each frequency point is calculated to obtain the frequency spectrum, and the frequency spectrum is normalized.

[0124] Based on this, for the weighted average of the frequency spectrum of the window: the weighted average of the signal of each window (i.e., the information of each group of A-line reflectivity curves) is obtained. The frequency in the normalized result is used as the weight to calculate the weighted average of the frequency spectrum, and the weighted average of the signal of each window (i.e., the information of each group of A-line reflectivity curves) is obtained. The mean value is taken to obtain the weighted average of the frequency spectrum of the window corresponding to the A-line reflectivity curve, as shown in Formula 10:

[0125] (10)

[0126] Where k is the label of each frequency point; N is the total number of frequency components; f k The frequency of point k; F(f) k ) is f k The frequency spectrum corresponding to the frequency.

[0127] Regarding the area of ​​the main peak in the frequency spectrum of each window: For the signal in each window (i.e., the A-line reflectance curve information of each group), analyze the frequency spectrum to find the frequency points whose frequency spectrum exceeds a set threshold T. Accumulate the frequency spectra at these frequency points to obtain the total frequency of the peak region composed of the main frequencies of the signal in each window (i.e., the A-line reflectance curve information of each group). This gives the area of ​​the main peak in the frequency spectrum of the signal in each window (i.e., the A-line reflectance curve information of each group). F As shown in Formula 11. Then, the average of the main peak area of ​​the frequency spectrum of all windows (i.e., all groups of A-line reflectance curve information) is taken as the main peak area of ​​the window frequency spectrum of the A-line reflectance curve.

[0128] (11)

[0129] Regarding the number of main peaks in the frequency spectrum of each window: For the signal in each window (i.e., the A-line reflectance curve information of each group), the number of frequency points exceeding the set threshold T in the frequency spectral density is calculated as the number of main peaks in the frequency spectrum, as shown in Formula 12, to obtain the number of main peaks in the frequency spectrum of the signal in each window (i.e., the A-line reflectance curve information of each group). F The number of main peaks in the window frequency spectrum of the A-line reflectivity curve is obtained by taking the average value.

[0130] (12)

[0131] For the standard deviation of the frequency spectrum of the window: For the signal in each window (i.e., the A-line reflectance curve information of each group), calculate the standard deviation of the frequency spectrum to obtain the characteristic of the standard deviation of the frequency spectrum of the signal in each window (i.e., the A-line reflectance curve information of each group). The average value is taken to obtain the standard deviation of the window frequency spectrum of the A-line reflectivity curve, as shown in Formula 13.

[0132] (13)

[0133] Furthermore, frequency-related features also include the weighted average of the overall frequency spectrum, the area of ​​the main peak in the overall frequency spectrum, the number of main peaks in the overall frequency spectrum, and the standard deviation of the overall frequency spectrum. The process of extracting frequency-related features includes: transforming all information of the current A-line reflectivity curve from the time domain to the frequency domain using a Fast Fourier Transform to determine the frequency spectrum of each frequency point, and calculating the weighted average of the overall frequency spectrum of the current A-line reflectivity curve; identifying target frequency points whose frequency spectra are greater than a first preset threshold among all information of the current A-line reflectivity curve, and accumulating the frequency spectra corresponding to the target frequency points to obtain the area of ​​the main peak in the overall frequency spectrum of the current A-line reflectivity curve; and determining the number of frequency points whose frequency spectra are greater than a second preset threshold among all information of the current A-line reflectivity curve, thus obtaining the number of main peaks in the overall frequency spectrum of the current A-line reflectivity curve.

[0134] Calculate the standard deviation of the frequency spectrum of all information in the current A-line reflectance curve to obtain the overall standard deviation of the frequency spectrum of the current A-line reflectance curve.

[0135] It should be noted that the calculation method for frequency characteristics of the entire effective signal range is the same in principle as the calculation method for frequency characteristics of a window. The only difference is the object of operation (the object of operation for frequency characteristics of the entire effective signal range is the data information of the entire effective signal range of the A-line reflectivity curve, while the object of operation for frequency characteristics of a window is the data of each window of the A-line reflectivity curve and then the average value is taken). You can refer to the calculation method for frequency characteristics of a window for understanding, and it will not be elaborated here.

[0136] Furthermore, the optical attenuation coefficient features include the attenuation coefficient and the attenuation coefficient amplitude. The process of extracting the optical attenuation coefficient features includes: introducing the axial confocal diffusion function and additive floor noise of the OCT system into the optical attenuation model based on Beer's law to obtain the tissue optical attenuation model based on the OCT system; performing nonlinear fitting on the optical coherence tomography data of historical cutting edge tissues using the least squares method to determine the target parameters in the tissue optical attenuation model based on the OCT system; wherein, the target parameters include the attenuation coefficient and the attenuation coefficient amplitude.

[0137] Specifically, in the embodiments of this application, the axial confocal diffusion function (PSF) and additive floor noise of the OCT system are introduced into the light attenuation model based on Beer's law to obtain the tissue optical attenuation model based on the OCT system. The distribution of intensity with depth, i.e., the model, can be described by Equation 14.

[0138] In practical applications, the least squares method can be used to perform nonlinear fitting on the optical coherence tomography (OCT) data of historical cut edge tissues to obtain three unknown parameters, namely the target parameters: the OCT signal amplitude A1 at the tissue boundary, the optical attenuation coefficient, and the optical coherence tomography (OCT) parameters. , and additive basis noise A2.

[0139] In some implementations of this application, the Levenberg-Marquardt numerical optimization iterative algorithm can be used to find the optimal solution for these parameters.

[0140] (14)

[0141] Where I(z) is the light intensity value at depth z, z cf z represents the depth position of the focal plane from the tissue boundary. R The Rayleigh length can be determined from the optical coherence tomography data of historical cutting edge tissues, where the attenuation coefficient amplitude can be obtained based on I(z).

[0142] Based on the above embodiments, 18 target features (i.e., depth features, slope features, distance features between local maxima, rapid intensity fluctuation features, frequency features, and optical attenuation coefficient features) determined from optical coherence tomography data of historical cutting edge tissue, along with their corresponding benign or malignant labels, are input into a preset classifier. For example, classifiers such as random forest, support vector machine, Naive Bayes classifier, and neural network can be used for training and validation. The trained classifier outputs the probability value of benign or malignant, and has the ability to evaluate the benign or malignant nature of A-line.

[0143] The trained classifier described above is used for benign and malignant assessment of OCT A-line. During training and validation, the classifier's input consists of 18 target features of OCT A-line and corresponding benign / malignant labels (in actual assessment, the input is the corresponding features extracted from optical coherence tomography data based on the tissue under test using the same principle). In some embodiments of this application, typical representative images of the six major feature categories for benign and malignant samples are shown below. Figure 9 As shown.

[0144] Figure 10The validation effect diagram for evaluating the classifier trained using the intraoperative tumor margin benignity / malignancy assessment method provided in this application is shown (an instance of RandomForestClassifier is created using the scikit-learn library, with the number of decision trees built in the forest set to n_estimators=50, and the standard criterion='entropy' used when splitting the trees). Figure 10 It is evident that the classifier trained using the intraoperative tumor margin benignity / malignancy assessment method provided in this application has good assessment performance.

[0145] It is understood that the same or similar parts in the above embodiments can be referred to each other, and the contents not described in detail in some embodiments can be referred to the same or similar contents in other embodiments.

[0146] It should be noted that in the description of this invention, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Furthermore, in the description of this invention, unless otherwise stated, "a plurality of" means at least two.

[0147] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0148] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0149] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it includes one or a combination of the steps of the method embodiments.

[0150] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0151] The storage media mentioned above can be read-only memory, disk, or optical disk, etc.

[0152] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0153] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for assessing the benign or malignant nature of tumor margins during surgery, characterized in that, include: Acquire optical coherence tomography data of historical cutting edge tissue; Based on the optical coherence tomography data of the historical cutting edge tissue, the A-line reflectance curve is extracted. The A-line reflectance curve is the curve of image intensity changing with the depth of the A-line. Target features are calculated and extracted based on the A-line reflectivity curve. These target features include depth features, slope features, distance features between local maxima, rapid intensity fluctuation features, frequency features, and optical attenuation coefficient features. The depth features include the penetration depth of the effective signal range. Based on the target features and the benign or malignant labels of the corresponding A-line reflectance curves, a preset classifier is trained to obtain the trained classifier. The target features extracted from the optical coherence tomography data of the tissue under test are input into the trained classifier to obtain the evaluation results; The optical attenuation coefficient features include attenuation coefficient and attenuation coefficient amplitude. The process of extracting the optical attenuation coefficient features includes: introducing the axial confocal diffusion function and additive floor noise of the OCT system into the optical attenuation model based on Beer's law to obtain the tissue optical attenuation model based on the OCT system; performing nonlinear fitting on the optical coherence tomography data of the historical cutting edge tissue using the least squares method to determine the target parameters in the tissue optical attenuation model based on the OCT system; wherein, the target parameters include the attenuation coefficient and the attenuation coefficient amplitude. The tissue optical attenuation model based on the OCT system is as follows: Where I(z) is the light intensity value at depth z; z cf The focal plane is at a depth relative to the tissue boundary; z R A1 is the Rayleigh length; A1 is the OCT signal amplitude; µ t A1 is the attenuation coefficient; A2 is the additive basis noise. The rapid intensity fluctuation feature includes: the standard deviation of the rapid intensity fluctuation and the peak value of the rapid intensity fluctuation. The process of extracting the rapid intensity fluctuation feature includes: The current A-line reflectance curve is smoothed to obtain the current A-line reflectance smoothed curve; Determine the difference in intensity rapid fluctuations between the current A-line reflectance curve and the current A-line reflectance smooth curve; Calculate the standard deviation of the difference, which is used as the standard deviation of the rapid fluctuation of the intensity of the current A-line reflectance curve; A difference curve is generated based on the difference, and the local maximum value with the largest absolute value in the difference curve is taken as the peak value of the rapid fluctuation of the intensity of the current A-line reflectance curve.

2. The method for assessing the benign or malignant nature of tumor margins during surgery according to claim 1, characterized in that, The process of extracting the penetration depth of the effective signal range includes: The current A-line reflectance curve is smoothed to obtain the current A-line reflectance smoothed curve; Determine the effective signal range of the current A-line reflectance smoothing curve; The difference in depth between the start and end points of the effective signal interval is taken as the penetration depth of the effective signal interval of the current A-line reflectivity curve.

3. The method for assessing the benign or malignant nature of tumor margins during surgery according to claim 2, characterized in that, Determining the effective signal range of the current A-line reflectivity smoothing curve includes: The depth position corresponding to the first time the image intensity reaches the target value in the current A-line reflectance smoothing curve is taken as the starting point of the effective signal interval; The effective signal interval of the current A-line reflectivity smoothing curve is selected by using a preset sliding window. When the linear fitting slope of the signal within the window approaches 0 or the signal within the window is completely 0, the starting depth position within the current window is taken as the end point of the effective signal interval.

4. The method for assessing the benign or malignant nature of tumor margins during surgery according to claim 1, characterized in that, The slope features include: the average slope of the effective signal interval, the overall slope of the effective signal interval, and the overall intercept of the effective signal interval. The process of extracting the slope features includes: Multiple data subsets are obtained by sliding a preset sliding window to select the effective signal range of the current A-line reflectivity smoothing curve; Calculate the linear fitting slope for each data subset separately, and take the average of the linear fitting slopes of all data subsets as the average slope of the effective signal range of the current A-line reflectance curve. Calculate the linear fit slope and intercept of the effective signal interval of the current A-line reflectance smoothing curve. Use the slope as the overall slope of the effective signal interval of the current A-line reflectance curve and the intercept as the overall intercept of the effective signal interval of the current A-line reflectance curve.

5. The method for assessing the benign or malignant nature of tumor margins during surgery according to claim 1, characterized in that, The distance features between local maxima include: the mean distance and the standard deviation of the distance. The process of extracting the distance features between local maxima includes: Determine the local maximum value of image intensity within the effective signal range of the current A-line reflectance curve; Calculate the depth difference corresponding to each adjacent local maximum value; The mean of all the differences is used as the mean distance of the current A-line reflectance curve, and the standard deviation of all the differences is used as the standard deviation distance of the current A-line reflectance curve.

6. The method for assessing the benign or malignant nature of tumor margins during surgery according to claim 1, characterized in that, The frequency-related features include: weighted average of the window frequency spectrum, area of ​​the main peak in the window frequency spectrum, number of main peaks in the window frequency spectrum, and standard deviation of the window frequency spectrum. The process of extracting the frequency-related features includes: By sliding through a preset sliding window, the current A-line reflectance curve can be selected to obtain multiple sets of A-line reflectance curve information; For each set of A-line reflectivity curve information, a Fast Fourier Transform is used to transform it from the time domain to the frequency domain to determine the frequency spectrum of each frequency point in the set. The weighted average of the frequency spectrum of the current set of A-line reflectivity curve information is then calculated using the following formula: Where k is the label of each frequency point; N is the total number of frequency components; f k The frequency of point k; F(f) k ) is f k The frequency spectrum corresponding to the frequency; The average of the frequency spectrum weighted average of all A-line reflectance curve information is used as the window frequency spectrum weighted average of the current A-line reflectance curve. For each group of A-line reflectivity curve information, the target frequency points in the group whose frequency spectrum is greater than the first preset threshold are determined, and the frequency spectra corresponding to all target frequency points in the group are accumulated to obtain the main peak area of ​​the frequency spectrum of the current group of A-line reflectivity curve information; the average value of the main peak area of ​​the frequency spectrum of all groups of A-line reflectivity curve information is taken as the window frequency spectrum main peak area of ​​the current A-line reflectivity curve. For each group of A-line reflectivity curve information, determine the number of frequency points in the frequency spectrum that are greater than the second preset threshold in that group, and obtain the number of main peaks in the frequency spectrum of the current group of A-line reflectivity curve information; take the average of the number of main peaks in the frequency spectrum of all groups of A-line reflectivity curve information as the number of main peaks in the window frequency spectrum of the current A-line reflectivity curve. For each group of A-line reflectivity curve information, calculate the standard deviation of the frequency spectrum in that group to obtain the standard deviation of the frequency spectrum of the current group of A-line reflectivity curve information; use the average of the standard deviations of the frequency spectrum of all groups of A-line reflectivity curve information as the window standard deviation of the current A-line reflectivity curve.

7. The method for assessing the benign or malignant nature of tumor margins during surgery according to claim 1, characterized in that, The frequency-related features also include the weighted average of the overall frequency spectrum, the area of ​​the main peak in the overall frequency spectrum, the number of main peaks in the overall frequency spectrum, and the standard deviation of the overall frequency spectrum. The process of extracting the frequency-related features includes: The current A-line reflectance curve is transformed from the time domain to the frequency domain using a Fast Fourier Transform to determine the frequency spectrum at each frequency point. The overall frequency spectrum weighted average of the current A-line reflectance curve is then calculated using the following formula: Where k is the label of each frequency point; N is the total number of frequency components; f k The frequency of point k; F(f) k ) is f k The frequency spectrum corresponding to the frequency; Identify the target frequency points whose frequency spectrum is greater than the first preset threshold among all information of the current A-line reflectivity curve, and accumulate the frequency spectra corresponding to the target frequency points to obtain the overall frequency spectrum main peak area of ​​the current A-line reflectivity curve; Determine the number of frequency points in the frequency spectrum of the current A-line reflectance curve that are greater than the second preset threshold, and obtain the total number of main peaks in the frequency spectrum of the current A-line reflectance curve. Calculate the standard deviation of the frequency spectrum of all information in the current A-line reflectance curve to obtain the overall standard deviation of the frequency spectrum of the current A-line reflectance curve.

Images