Oral cancer edge cutting specimen positioning fluorescence detection method

Abstract

The invention relates to the technical field of fluorescence detection, in particular to a positioning fluorescence detection method for an oral cancer incision edge specimen. According to the invention, the nano graphene oxide and the fluorescent targeting peptide are mixed and purified to obtain a fluorescent targeting probe solution with high specificity and stability, and the probe is injected into a specimen tissue by using a microneedle system, so that the consistency of the injection depth and the diffusion range is ensured, and the accuracy and reliability of marking are improved; spatial segmentation and mutation point analysis are carried out through optical signal intensity distribution characteristics, low-intensity and abnormal fluctuation signals are eliminated, a signal region boundary is optimized, multi-site extraction is carried out on cutting edge signals through a convolutional neural network, signal intensity center calculation and spatial mapping are analyzed, and a two-dimensional quantizable point set is generated. According to the method, the signal distribution trend is marked, the consistency of boundaries is calibrated, an incisal edge positioning map with higher space consistency and systematicness is generated, the reliability of incisal edge judgment is improved, the omission risk is reduced, and the precision of operation auxiliary diagnosis is improved.

Description

Technical Field

[0001] The invention relates to the technical field of fluorescence detection, and in particular to a method for positioning fluorescence detection of oral cancer cutting edge specimens. Background Art

[0002] The field of fluorescence detection technology aims to use the optical properties of fluorescent substances to detect, quantify or locate specific molecules, compounds and biological components. By stimulating fluorescent substances to emit light signals of specific wavelengths, the presence, concentration or spatial distribution of target substances can be quickly and accurately analyzed with the help of fluorescence spectrometers, fluorescence microscopes and imaging equipment.

[0003] The purpose of the fluorescence detection method for positioning oral cancer resection margin specimens is to quickly and accurately identify the tumor tissue area of ​​the pathological resection margin during oral cancer resection surgery through fluorescent labeling and detection, to clarify whether there is residual cancer tissue or a positive resection margin, to improve the accuracy of surgical margin judgment, reduce the risk of postoperative recurrence, and assist in pathological diagnosis.

[0004] In the process of fluorescent labeling, the existing technology makes it difficult to accurately control the injection depth of the specimen tissue and the diffusion range of the probe, which can easily cause insufficient labeling depth and uneven distribution, resulting in reduced labeling efficiency and local omissions. The acquisition of optical signals relies on a single signal channel, and it is unable to fully separate signals of different wavelengths, which is prone to signal superposition and artifacts, reducing imaging clarity and resolution. There is a lack of spatial mapping and trend annotation methods, making it difficult to form an intuitive and systematic distribution map, resulting in low regional consistency in the results of margin judgment, resulting in inaccurate margin judgment and increased risk of postoperative recurrence. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings in the prior art and to propose a method for localizing fluorescence detection of oral cancer resection margin specimens.

[0006] In order to achieve the above object, the present invention adopts the following technical scheme: a method for localizing fluorescence detection of oral cancer resection margin specimens, comprising the following steps:

[0007] Step 1: Mix the nano-graphene oxide and the fluorescent targeting peptide solution, and mix them evenly in a stirring device according to a fixed ratio, separate the unbound components from the solution using a high-speed centrifugal device, repeatedly wash to remove impurities, and perform ultrapure water washing and filtering operations on the washed sample to obtain a fluorescent targeting probe solution;

[0008] Step 2: Based on the fluorescent targeting probe solution, the solution is injected into the specimen tissue through a microneedle system to ensure that the injection depth and diffusion range meet the specified standards. After the injection, the specimen is kept under constant environmental conditions, and the binding characteristics of the targeting peptide in the solution with the target molecule are used for labeling to generate a targeted binding area label;

[0009] Step 3: Based on the targeted binding area marking, a near-infrared fluorescence imaging device is used to collect signal data from the specimen surface point by point, and the wavelength range of the light signal is separated by a regional scanning method. The regional contour is extracted according to the gradient change of the collected signal, the separation results are integrated and the boundary line is marked to generate a marker signal contour map;

[0010] Step 4: Based on the marker signal contour mapping, the part with signal strength lower than the set threshold is removed, and spatial segmentation is performed according to the distribution characteristics of the optical signal strength in the area, the location of the mutation point is determined and the partition boundary is delineated, and the boundary line of the significant signal change is determined in combination with the spatial segmentation result to generate signal partition boundary data;

[0011] Step 5: Based on the signal partition boundary data, the change rate is calculated according to the intensity gradient of the boundary signal, the part with abnormal signal fluctuation range is checked point by point, the signal contour is redefined after eliminating the interference signal, and the region is continuously adjusted and optimized based on the characteristics of the signal boundary to generate an optimized tumor tissue boundary map;

[0012] Step 6: Based on the optimized tumor tissue boundary mapping, a convolutional neural network is used to extract signals from multiple cutting edge sites on the specimen in turn, and the signal distribution and intensity center of each site are calculated by a spatial mapping method, and the signal data is converted into a quantifiable point set in a two-dimensional space. The position relationship of the site signal is annotated according to the overall distribution trend of the point set to generate a cutting edge site signal distribution map;

[0013] Step 7: Based on the signal distribution map of the cutting edge site, the spatial distribution positions of the signal center and the boundary points are compared point by point, the positioning points are adjusted according to the differences in the overlapping boundary areas, the spatial consistency of the signal boundary is calibrated in combination with the distribution characteristics of the cutting edge site, and a cutting edge area positioning map is generated.

[0014] As a further solution of the present invention, the specific steps of generating the fluorescent targeting probe solution are:

[0015] The nano-graphene oxide and the fluorescent targeting peptide solution are mixed and placed in a stirring device for constant stirring at a fixed ratio, and a high-speed centrifugal device is used to separate unbound components through multiple high-speed rotations to generate a primary mixed solution;

[0016] Based on the primary mixed solution, a high-speed centrifugal device is used to gradually remove unbound impurities through high-frequency centrifugal operation, the washing operation is repeated multiple times, and ultrapure water is used to remove remaining impurities through continuous washing and filtration to generate a purified mixed solution;

[0017] Based on the purified mixed solution, the fluorescent targeting peptide is supplemented in proportion and stirred evenly, and the solution is diluted and concentrated by multiple washing operations combined with ultrapure water to complete the final binding treatment to generate a fluorescent targeting probe solution;

[0018] The fluorescent targeting probe solution includes the combined fluorescent targeting peptide, residual graphene oxide and uncombined fluorescent molecules.

[0019] As a further embodiment of the present invention, the specific steps of generating the targeted binding region marker are:

[0020] Based on the fluorescent targeting probe solution, a microneedle system is used to inject the sample tissue through uniform injection operation, the injection depth and diffusion range are controlled, and the injection uniformity is observed to generate an injection distribution structure;

[0021] Based on the injection distribution structure, the specimen is placed in a constant temperature and humidity environment, and the target peptide is gradually combined with the target molecule to complete the formation of the marked area and record the distribution, thereby generating a targeted binding area mark;

[0022] The target binding region marker includes a fluorescent region formed by the binding of the targeting peptide and the tumor-specific molecule, a distribution range of the marker signal intensity, and a boundary of an unmarked region.

[0023] As a further solution of the present invention, the specific steps of generating the marker signal contour map are:

[0024] Based on the target binding region marker, a near-infrared fluorescence imaging device is used to collect light signal data on the surface of the specimen point by point, and light signal intensity is collected at fixed time intervals and position parameters are recorded to generate light signal data on the surface of the specimen;

[0025] Based on the optical signal data on the surface of the specimen, the wavelength range of the optical signal is separated by using a regional scanning method, and stray light signals are eliminated by signal partition processing and repeated scanning to generate separated wavelength signal data;

[0026] Based on the separated wavelength signal data, extract the specimen contour by screening the peak intensity area according to the gradient change of the collected signal, generate the marked boundary line in combination with the contour boundary connection process, and generate the marked signal contour mapping;

[0027] The marker signal contour mapping includes the signal boundary position, the spatial distribution of different fluorescence intensity areas and the marker results of the optical signal wavelength range.

[0028] As a further solution of the present invention, the specific steps of generating the signal partition boundary data are:

[0029] Based on the marked signal contour mapping, the portion whose signal strength is lower than a set threshold is eliminated, low-intensity areas are excluded one by one through a signal screening operation, and the remaining data points are re-marked to generate screening signal data;

[0030] Based on the screening signal data, the spatial area is divided by continuity analysis of the light signal intensity distribution characteristics, and the mutation point position is located by step-by-step signal screening to generate spatial partition contour data;

[0031] Based on the spatial partition contour data, a complete dividing line is generated by integrating the partition contour information, and the accurate boundary of the area where the signal changes significantly is marked by boundary correction to generate signal partition boundary data;

[0032] The signal partition boundary data includes the boundary line of the significant signal partition, the signal strength distribution in the area and the spatial position of the abnormal signal.

[0033] As a further solution of the present invention, the specific steps of generating the optimized tumor tissue boundary map are:

[0034] Based on the signal partition boundary data, the boundary signal strength gradient change is measured point by point, and the signal strength difference is calculated and recorded. The change rate is analyzed point by point for the abnormal fluctuation area and the position is marked to generate abnormal signal position data;

[0035] Based on the abnormal signal position data, the signal contour shape is adjusted by gradually eliminating abnormal points and interference signals, and the breakpoint area is reconnected and the contour curve is corrected through signal continuity analysis to generate an adjusted signal boundary contour;

[0036] Based on the adjustment of the signal boundary contour, by optimizing and adjusting the continuity of the signal intensity in the boundary curve area, by correcting and connecting the signal boundary line segment by segment to generate a complete area map, an optimized tumor tissue boundary map is generated;

[0037] The optimized tumor tissue boundary mapping includes an optimized boundary contour, an adjusted boundary point set, and a regional distribution after eliminating interference signals.

[0038] As a further solution of the present invention, the measurement process of the boundary signal intensity gradient change obtains the signal intensity value of each sampling point from the signal partition boundary data, collects the signal value point by point at a fixed interval, calculates the intensity difference between adjacent points, and records the measurement result as a set of continuous gradient change data;

[0039] The analysis and determination method of the fluctuation abnormal area sets a threshold as the determination standard according to the recorded gradient change data. If the intensity difference of multiple consecutive points exceeds the threshold, it is determined to be a fluctuation abnormal area. The starting point and ending point of each abnormal area are recorded as the boundary of the area.

[0040] As a further solution of the present invention, the specific steps of generating the cutting edge site signal distribution map are:

[0041] Based on the optimized tumor tissue boundary mapping, the specimen resection margin sites are located one by one and the optical signal data of each site is extracted, and the relationship between the signal and the position is established by recording the signal intensity value and the site coordinates to generate a resection margin site optical signal point set;

[0042] Based on the optical signal point set of the cutting edge site, the center position of the signal is located by analyzing the optical signal intensity distribution, a preliminary spatial area is delineated in combination with the distribution characteristics of each point, and the point set is reintegrated by the site distribution characteristics to generate cutting edge site distribution data;

[0043] Based on the distribution data of the cutting edge sites, a convolutional neural network is used to connect the regions by analyzing the relationship between the distribution trend of the point set and the signal intensity, and a two-dimensional distribution map is generated by marking the position relationship of the sites in the spatial distribution range to generate a cutting edge site signal distribution map;

[0044] The cutting edge site signal distribution diagram includes the spatial coordinates of the site signal, the intensity value of the signal center and the distance relationship between each site.

[0045] As a further solution of the present invention, the convolutional neural network is according to the formula:

[0046]

[0047] Where: Y represents the fluorescence intensity value of each pixel in the generated cut edge site signal distribution map, w i represents the weight parameter in the convolution kernel, x i represents the fluorescence signal intensity value of the point set in the cutting edge site distribution data, and α represents the local density d i The weight coefficient, d i represents the local density of the point set, and β represents the signal intensity gradient z i The weight coefficient, z i represents the gradient value of the fluorescence signal intensity, and γ represents the distance p between the point and the center of the window i The weight coefficient, p i represents the distance between the cutting edge site in the point set and the center of the convolution window, b represents the bias term, and f represents the activation function.

[0048] As a further solution of the present invention, the specific steps of generating the cutting edge area positioning map are:

[0049] Based on the signal distribution map of the cutting edge site, the spatial distribution positions of the signal center and the boundary points are measured point by point, the position offset is gradually marked by calculating the distance difference between each center point and the adjacent boundary point, and the offset characteristics of the overlapping area are recorded to generate boundary offset data;

[0050] Based on the boundary offset data, by gradually adjusting the positioning points with larger offsets, a correction operation is performed on the site offset direction, and by rearranging and adjusting the positioning points so that they match the signal distribution characteristics in space, a calibration positioning point distribution map is generated;

[0051] Based on the calibration positioning point distribution map, the corrected signal boundary points are connected segment by segment, the complete boundary line is regenerated in combination with the overall boundary information, and the connection logic is adjusted by comparing the optimized boundary morphology to generate a cutting edge area positioning map;

[0052] The cutting edge region positioning map includes the comprehensive boundary of cutting edge signal distribution, the spatial coordinates of the positioning points and the final region after the boundary adjustment.

[0053] Compared with the prior art, the advantages and positive effects of the present invention are:

[0054] 1. In the present invention, by mixing nano-graphene oxide and fluorescent targeting peptide and performing purification operation, a fluorescent targeting probe solution with high specificity and stability is obtained, and the probe is injected into the specimen tissue using a microneedle system to ensure the consistency of injection depth and diffusion range, thereby improving the accuracy and reliability of labeling;

[0055] 2. In the present invention, spatial segmentation and mutation point analysis are performed based on the intensity distribution characteristics of the optical signal, low-intensity and abnormal fluctuation signals are eliminated, and the signal area boundary is optimized.

[0056] 3. In the present invention, a convolutional neural network is used to extract multi-site cutting edge signals, analyze the signal intensity center calculation and spatial mapping, generate a two-dimensional quantifiable point set, mark the signal distribution trend and calibrate the consistency of the boundary, and generate a cutting edge positioning map with higher spatial consistency and systematicity, thereby improving the reliability of cutting edge judgment, reducing the risk of omission, and improving the accuracy of surgical auxiliary diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a schematic diagram of the main steps of the present invention. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0059] See also Figure 1 The present invention provides a technical solution: a method for localizing fluorescence detection of oral cancer resection margin specimens, comprising the following steps:

[0060] Step 1: Mix the nano-graphene oxide and the fluorescent targeting peptide solution, and mix them evenly in a stirring device according to a fixed ratio, separate the unbound components from the solution using a high-speed centrifugal device, repeatedly wash to remove impurities, and perform ultrapure water washing and filtering operations on the washed sample to obtain a fluorescent targeting probe solution;

[0061] Step 2: Based on the fluorescent targeted probe solution, the solution is injected into the specimen tissue through a microneedle system to ensure that the injection depth and diffusion range meet the specified standards. After the injection, the specimen is kept under constant environmental conditions and marked using the binding characteristics of the targeting peptide in the solution with the target molecule to generate a targeted binding area marker;

[0062] Step 3: Based on the targeted binding area marking, the near-infrared fluorescence imaging device is used to collect signal data point by point on the specimen surface, and the wavelength range of the light signal is separated by a regional scanning method. The regional contour is extracted according to the gradient change of the collected signal, the separation results are integrated and the boundary lines are marked to generate the marking signal contour mapping;

[0063] Step 4: Based on the marker signal contour mapping, the part with signal strength lower than the set threshold is eliminated, and the spatial segmentation is performed according to the distribution characteristics of the optical signal strength in the area, the location of the mutation point is determined and the partition boundary is delineated, and the boundary line of the significant signal change is determined based on the spatial segmentation result to generate the signal partition boundary data;

[0064] Step 5: Based on the signal partition boundary data, the change rate is calculated according to the intensity gradient of the boundary signal, and the abnormal part of the signal fluctuation range is checked point by point. After removing the interference signal, the signal contour is redefined, and the region is continuously adjusted and optimized based on the characteristics of the signal boundary to generate an optimized tumor tissue boundary map;

[0065] Step 6: Based on the optimized tumor tissue boundary mapping, a convolutional neural network is used to extract signals from multiple cutting edge sites on the specimen in turn. The signal distribution and intensity center of each site are calculated through a spatial mapping method, and the signal data is converted into a quantifiable point set in a two-dimensional space. The position relationship of the site signal is annotated according to the overall distribution trend of the point set to generate a cutting edge site signal distribution map;

[0066] Step 7: Based on the signal distribution map of the cutting edge site, the spatial distribution positions of the signal center and the boundary points are compared point by point, the positioning points are adjusted according to the differences in the overlapping boundary areas, and the spatial consistency of the signal boundary is calibrated in combination with the distribution characteristics of the cutting edge sites to generate a cutting edge area positioning map.

[0067] The specific steps for generating the fluorescent targeting probe solution are:

[0068] The nano-graphene oxide and the fluorescent targeting peptide solution are mixed and placed in a stirring device for constant stirring at a fixed ratio, and a high-speed centrifugal device is used to separate unbound components through multiple high-speed rotations to generate a primary mixed solution;

[0069] Based on the initial mixed solution, a high-speed centrifugal device is used to gradually remove unbound impurities through high-frequency centrifugal operation, and the washing operation is repeated multiple times. Ultrapure water is used to remove the remaining impurities through continuous washing and filtration to generate a purified mixed solution;

[0070] Based on the purified mixed solution, the fluorescent targeting peptide is added in proportion and stirred evenly, and the solution is diluted and concentrated by multiple washing operations combined with ultrapure water to complete the final binding treatment to generate a fluorescent targeting probe solution;

[0071] Based on the nano-graphene oxide and fluorescent targeting peptide solution, a fixed ratio was used to place them in a constant speed stirring device, the stirring speed was set to 300 rpm, the stirring time was 60 minutes, the solution was uniformly mixed, and the initial mixed solution was separated by a high-speed centrifugal device, the centrifugal speed was set to 12000 rpm, the centrifugal time was 15 minutes, the unbound components were removed by separation, and the above centrifugation operation was repeated 3 times to generate a primary mixed solution;

[0072] Based on the initial mixed solution, a high-speed centrifugal device was used to perform high-frequency centrifugation on the solution, the centrifugal speed was set to 15000 rpm, the centrifugal time was 10 minutes, the unbound impurity components in the solution were gradually removed, and the solution was continuously rinsed with ultrapure water, the rinse flow rate was set to 5 mL / min, the rinse time was 30 minutes, and the solution was separated from impurities and filtered by a filtration device, and the rinsing and filtering steps were repeated 5 times to generate a purified mixed solution;

[0073] Based on the purified mixed solution, the fluorescent targeting peptide was added in proportion and stirred evenly. The stirring device was set to a speed of 200 rpm for 90 minutes. The solution components were fully mixed. The solution was diluted to a target concentration with ultrapure water, and the concentration was set to 0.5 mg / mL. The solution was concentrated using a high-speed centrifugal device, and the centrifugal speed was set to 10,000 rpm for 20 minutes. The solution was dehydrated, and the dilution and concentration steps were repeated 3 times to generate a fluorescent targeting probe solution.

[0074] The fluorescent targeting probe solution includes bound fluorescent targeting peptides, residual graphene oxide and unbound fluorescent molecules.

[0075] The specific steps to generate the target binding region label are:

[0076] Based on the fluorescent targeted probe solution, a microneedle system is used to inject the sample tissue through uniform injection operation, control the injection depth and diffusion range, observe the injection uniformity, and generate the injection distribution structure;

[0077] Based on the injection distribution structure, the specimen is placed in a constant temperature and humidity environment, and the target peptide gradually binds to the target molecule to complete the formation of the labeled area and record the distribution to generate a targeted binding area marker;

[0078] Based on the fluorescent targeted probe solution, a microneedle system was used to perform injection operations on the specimen tissue. The injection equipment parameters were set to a pressure of 50 kPa, an injection depth of 2 mm, and an injection speed of 0.5 μL / s. The tissue surface was covered by a uniform injection grid with a needle spacing of 1.5 mm. A high-resolution imaging device was used to scan the injection point distribution point by point to generate an injection position data set. The injection spacing deviation was extracted from the position data set and the injection equipment parameters were corrected to generate an injection distribution structure.

[0079] Based on the injection distribution structure, the specimen was placed in a constant environment with the ambient temperature set to 37°C and the humidity set to 75%. A matching algorithm was used to monitor the targeted binding process through time-series imaging of the specimen area. Specifically, a regional positioning algorithm based on differential analysis was used, and the binding time limit between the targeting peptide and the target molecule was set to 30 minutes. The imaging data was collected frame by frame and differentially calculated according to the time node. The binding was determined to be complete when the standard deviation of the regional statistics was less than 0.05. The marked area was located in the corresponding regional map of the specimen surface to generate a targeted binding area marker.

[0080] The target binding region marker includes the fluorescent region formed by the binding of the targeting peptide to the tumor-specific molecule, the distribution range of the marker signal intensity and the boundary of the unlabeled region.

[0081] The specific steps to generate the marker signal contour map are:

[0082] Based on the targeted binding area marking, a near-infrared fluorescence imaging device is used to collect the light signal data on the surface of the specimen point by point. The light signal intensity is collected at fixed time intervals and the position parameters are recorded to generate the light signal data on the surface of the specimen.

[0083] Based on the optical signal data on the specimen surface, the wavelength range of the optical signal is separated by using a regional scanning method. The stray light signal is eliminated by signal partition processing and repeated scanning to generate separated wavelength signal data.

[0084] Based on the separated wavelength signal data, the sample contour is extracted by screening the peak intensity area according to the gradient change of the collected signal, and the marked boundary line is generated by combining the contour boundary connection processing to generate the marked signal contour mapping;

[0085] Based on the targeted binding area marking, a near-infrared fluorescence imaging device was used to collect the optical signal data of the specimen surface point by point. The moving interval was set to 0.5 mm, and the optical signal collection frequency was set to 10 Hz. The optical signal intensity at each position was recorded and bound to the collection position parameters. The collected signals were preliminarily separated using the spectral analysis method. The analysis parameter range was 700 nm to 900 nm. The sampling interval was set to 2 nm through fixed step analysis. The analyzed optical signal data was automatically encoded and stored to generate the optical signal data of the specimen surface.

[0086] Based on the optical signal data of the specimen surface, the signal is separated step by step in wavelength range by using a regional scanning method. The scanning area is set to a 10×10mm unit block, and the scanning optical signal range is set to 700nm to 900nm. The stray optical signal is eliminated by using a noise elimination algorithm with multiple iterations. The threshold of the elimination operation is set to 5% of the peak value of the optical signal, and the number of repeated scans is 5 times. The data after elimination is integrated by wavelength segmentation, and each segment signal is generated into a segmented signal matrix according to the wavelength sequence to generate separated wavelength signal data.

[0087] Based on the separated wavelength signal data, the peak area is extracted by a signal screening algorithm with a dynamic threshold according to the gradient change of the collected signal. The dynamic threshold is set as the signal gradient change is greater than three times the standard deviation of the mean. The signal peak of the extracted area is outlined, and the contour connection algorithm is used to connect the points with similar gradients. The connection parameter is that the distance between adjacent points is less than 1mm. The connected boundary line is fitted with a smooth curve to generate a complete contour, and a marker signal contour mapping is generated;

[0088] Among them, the marker signal contour mapping includes the signal boundary position, the spatial distribution of different fluorescence intensity areas and the marking results of the optical signal wavelength range.

[0089] The specific steps to generate signal partition boundary data are:

[0090] Based on the mapping of the marked signal contour, the part with signal strength lower than the set threshold is eliminated, the low-intensity area is eliminated one by one through the signal screening operation, and the remaining data points are re-marked to generate the screened signal data;

[0091] Based on the screening signal data, the spatial area is divided through the continuity analysis of the light signal intensity distribution characteristics, and the mutation point position is located through step-by-step signal screening to generate spatial partition contour data;

[0092] Based on the spatial partition contour data, the complete dividing line is generated by integrating the partition contour information, and the accurate boundary of the area where the signal changes significantly is marked by boundary correction to generate signal partition boundary data;

[0093] Based on the marker signal contour mapping, the signal strength screening algorithm is used to screen the data points. The screening threshold is set to 20% of the signal strength below the regional mean. The signal points below the threshold are removed point by point. The remaining signal points are re-marked in combination with the data point redistribution characteristics. The marked data are grouped using matrix indexing operations and the spatial distribution matrix of the remaining signal points is generated to generate the screened signal data.

[0094] Based on the screening signal data, the spatial area is divided by the optical signal intensity distribution characteristic analysis algorithm, and the neighborhood range is set to the signal point spacing less than 1mm by the continuity analysis method. The signal points that meet the conditions are partitioned and classified according to the regional label. The intensity change rate of the signal points in the partition is analyzed point by point by using a recursive function to locate the position of the mutation point where the intensity change amplitude is greater than two times the standard deviation of the mean. The regional contour data of the mutation point is marked and the continuous partition boundary line is generated by the curve fitting method to generate the spatial partition contour data;

[0095] Based on the spatial partition contour data, the partition contour integration algorithm is used to connect the partition boundary points one by one. The boundary integration rule is set as the distance between adjacent points is less than 2 mm. The complete dividing line is generated in the order of the contour. The boundary correction algorithm is used to re-mark the significant change points of the boundary. The marking rule is that the signal gradient mutation rate is greater than 3 times the standard deviation of the regional mean. The optimized boundary is generated through the spatial position redefined by the correction point, and the signal partition boundary data is generated.

[0096] Among them, the signal partition boundary data includes the boundary line of the significant signal partition, the signal strength distribution in the area and the spatial position of the abnormal signal.

[0097] The specific steps to generate an optimized tumor tissue boundary map are:

[0098] Based on the signal partition boundary data, measure the boundary signal strength gradient change point by point, calculate and record the signal strength difference, analyze the change rate point by point for the abnormal fluctuation area and mark the position, and generate abnormal signal position data;

[0099] Based on the abnormal signal location data, the signal contour shape is adjusted by gradually removing abnormal points and interference signals, and the breakpoint area is reconnected and the contour curve is corrected through signal continuity analysis to generate the adjusted signal boundary contour;

[0100] Based on adjusting the signal boundary contour, the signal intensity continuity of the boundary curve area is optimized and adjusted, and a complete area map is generated by correcting and connecting the signal boundary line segment by segment to generate an optimized tumor tissue boundary map;

[0101] Based on the signal partition boundary data, the gradient measurement algorithm is used to calculate the change rate of the boundary signal strength point by point. The measurement parameter is set to a neighborhood window size of 3 points. The signal strength difference between each point and its previous and next points is calculated. The signal change is quantified using the differential formula. The strength difference of each point is recorded and the local change rate is calculated. The change rate standard is that the absolute value of the difference is greater than 3 times the average difference. The fluctuation abnormal area that exceeds the threshold is marked point by point to generate abnormal signal location data.

[0102] Based on the abnormal signal position data, the abnormal points and interference signals are gradually eliminated by using the abnormal elimination and contour adjustment algorithm. The elimination rule is that the change rate of the neighborhood signal strength is less than 0.5 times the average change rate. The contour shape of the signal points after elimination is adjusted. The signal continuity analysis method is used to reconstruct the disconnected area through the breakpoint connection algorithm. The breakpoint connection parameter is to directly connect the point pairs with a distance of less than 2mm. The contour curve is corrected by using the smooth fitting algorithm. The fitting function is set to a cubic polynomial to generate the adjusted signal boundary contour.

[0103] Based on adjusting the signal boundary contour, the signal intensity continuity optimization algorithm is used to optimize the boundary curve area, the signal intensity distribution of the curve is analyzed segment by segment, and the curve is corrected by piecewise linear fitting. The fitting parameters are that the segment length is set to 5mm, and the fitting error threshold is less than 0.01. The corrected segments of the curve are connected segment by segment to generate a complete boundary curve, thereby generating an optimized tumor tissue boundary map.

[0104] The optimized tumor tissue boundary mapping includes an optimized boundary contour, an adjusted boundary point set, and a regional distribution after eliminating interference signals.

[0105] The measurement process of the boundary signal intensity gradient change is to obtain the signal intensity value of each sampling point from the signal partition boundary data, collect the signal value point by point at a fixed interval, calculate the intensity difference between adjacent points, and record the measurement results as a set of continuous gradient change data;

[0106] The analysis and determination method of the fluctuation abnormal area is to set a threshold as the determination standard according to the recorded gradient change data. If the intensity difference of multiple consecutive points exceeds the threshold, it is determined to be a fluctuation abnormal area. The starting point and end point of each abnormal area are recorded as the boundary of the area.

[0107] The specific steps for generating the signal distribution map of the cutting edge site are:

[0108] Based on the optimized tumor tissue boundary mapping, the specimen resection margin sites are located one by one and the optical signal data of each site is extracted. The relationship between the signal and the position is established by recording the signal intensity value and the site coordinates, and the optical signal point set of the resection margin site is generated.

[0109] Based on the optical signal point set of the cutting edge site, the center position of the signal is located by analyzing the optical signal intensity distribution, and the preliminary spatial area is delineated based on the distribution characteristics of each point. The point set is reintegrated based on the site distribution characteristics to generate the cutting edge site distribution data;

[0110] Based on the distribution data of the cutting edge sites, a convolutional neural network is used to connect the various regions by analyzing the relationship between the distribution trend of the point set and the signal intensity. A two-dimensional distribution map is generated by marking the position relationship of the sites in the spatial distribution range to generate a cutting edge site signal distribution map.

[0111] Based on the optimization of tumor tissue boundary mapping, a point-by-point positioning algorithm is used to locate the specimen resection margin site. The optical signal data extraction method of each site is used to collect the optical signal point by point. The sampling interval is set to 1mm. The signal intensity value of each site is recorded and bound to the site space coordinates. The signal value and the position coordinates are associated and stored using a hash mapping function to generate a resection margin site optical signal point set.

[0112] Based on the optical signal point set of the incision margin site, the signal distribution of each point was analyzed using the signal intensity distribution analysis algorithm. The signal intensity peak was calculated and the signal center was located using the Gaussian fitting model. The fitting parameters were a window size of 5 mm and an error threshold of 0.02. The center position was marked, and the initial spatial area was delineated by the clustering algorithm based on the distribution characteristics of all sites. The clustering parameters were a maximum distance of 1.5 mm and a minimum sample number of 10. The point set data was reintegrated using the marked center position to generate the incision margin site distribution data.

[0113] Based on the distribution data of the cutting edge sites, a convolutional neural network was used to analyze the distribution trend of the point set and the relationship between the signal intensity. The network model structure included three convolutional layers, each with a convolution kernel size of 3×3, a step size of 1, and an activation function of ReLU. The pooling layer was used to downsample the convolution results of each layer, with a pooling size of 2×2. The weight parameters were adjusted through back propagation based on the training data, and the classifier was used to map the point set relationship in each region to a two-dimensional space to generate a signal distribution map of the cutting edge sites.

[0114] The signal distribution diagram of the cutting edge site includes the spatial coordinates of the site signal, the intensity value of the signal center and the distance relationship between each site.

[0115] Convolutional neural network, according to the formula:

[0116]

[0117] Where: Y represents the fluorescence intensity value of each pixel in the generated cut edge site signal distribution map, w i represents the weight parameter in the convolution kernel, x i represents the fluorescence signal intensity value of the point set in the cutting edge site distribution data, and α represents the local density di The weight coefficient, d i represents the local density of the point set, and β represents the signal intensity gradient z i The weight coefficient, z i represents the gradient value of the fluorescence signal intensity, and γ represents the distance p between the point and the center of the window i The weight coefficient, p i represents the distance between the cutting edge site in the point set and the center of the convolution window, b represents the bias term, and f represents the activation function;

[0118] Execution process: First, according to the fluorescence detection data of the oral cancer resection margin specimen, the fluorescence signal intensity value of each point set is extracted and recorded as x i , and calculate the local density value d within each window i , obtained by calculating the number of points in the window and dividing it by the window area, and then calculating the gradient value z of the signal strength i , by performing partial derivative operation on the spatial distribution of the fluorescence signal in the window, extracting the direction and rate of signal change, obtaining the signal change trend, and then calculating the distance between each point set and the center of the convolution window, denoted as p i , obtained by calculating the square root of the sum of the squares of the coordinates of the point set and the coordinates of the center of the window, and introducing weight coefficients α, β, and γ for local density, signal gradient, and the distance between the point and the center, respectively, to adjust the impact on the generation of the final fluorescence distribution map. Next, the convolution kernel is used to perform weighted summation on each window, including the signal intensity x i , local density d i , signal gradient z i and the distance value p i , plus the bias term b, and then input into the activation function f. The activation function introduces nonlinear mapping to enhance the expression ability of the signal, and finally obtains the signal distribution map of the cutting edge site.

[0119] The specific steps for generating the incision margin area positioning map are as follows:

[0120] Based on the signal distribution map of the incision margin site, the spatial distribution positions of the signal center and the boundary points are measured point by point, and the position offset is gradually marked by calculating the distance difference between each center point and the adjacent boundary point, and the offset characteristics of the overlapping area are recorded to generate the boundary offset data;

[0121] Based on the boundary offset data, the offset direction of the site is corrected by gradually adjusting the positioning points with larger offsets, and the positioning points are rearranged to match the signal distribution characteristics in space to generate a calibration positioning point distribution map;

[0122] Based on the calibration positioning point distribution map, the corrected signal boundary points are connected segment by segment, and the complete boundary line is regenerated in combination with the overall boundary information. The connection logic is adjusted by comparing the optimized boundary morphology to generate the incision area positioning map.

[0123] Based on the signal distribution map of the cutting edge site, the spatial position relationship between the signal center and the boundary point is measured point by point using a distance calculation algorithm. The distance is calculated according to the spatial coordinates of adjacent points. The measurement interval is set to 1 mm. Each center point and its adjacent boundary point are calculated point by point, and the offset of each point is recorded step by step. The offset is classified and stored by region using a data structure. The offset values ​​in the overlapping area are statistically analyzed, and the offset features are stored as a multidimensional data matrix; the boundary offset data is generated;

[0124] Based on the boundary offset data, the offset correction algorithm is used to adjust the positioning points. The adjustment rule is set to trigger the correction when the offset value exceeds twice the average offset value of the area. The offset direction is determined point by point for the positioning points that exceed the threshold. The coordinate position of the offset point is adjusted to gradually align it with the signal distribution characteristics. The corrected point set is rearranged according to the spatial distribution. The arrangement rule is that the distance between the correction points is kept within 3mm. Combined with the order of the adjusted point set, complete positioning point data is generated; a calibration positioning point distribution map is generated;

[0125] Based on the calibration positioning point distribution map, the segment-by-segment connection algorithm is used to spatially connect the corrected signal boundary points. The connection rule is that the distance between two points should not exceed 2 mm. All boundary points are connected into a continuous curve by multi-segment connection. The discontinuous part of the curve is smoothly adjusted. The curve connection logic is corrected by multi-segment spline interpolation method, and the adjusted curve information is mapped to the specimen boundary space; the incision area positioning map is generated;

[0126] The cutting edge area positioning map includes the comprehensive boundary of cutting edge signal distribution, the spatial coordinates of the positioning points and the final area after boundary adjustment.

[0127] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for localizing fluorescence detection of oral cancer resection margin specimens, characterized in that: The following steps are involved: Step 1: Mix the nano-graphene oxide and the fluorescent targeting peptide solution, and mix them evenly in a stirring device according to a fixed ratio, separate the unbound components from the solution using a high-speed centrifugal device, repeatedly wash to remove impurities, and perform ultrapure water washing and filtering operations on the washed sample to obtain a fluorescent targeting probe solution; Step 2: Based on the fluorescent targeting probe solution, the solution is injected into the specimen tissue through a microneedle system to ensure that the injection depth and diffusion range meet the specified standards. After the injection, the specimen is kept under constant environmental conditions, and the binding characteristics of the targeting peptide in the solution with the target molecule are used for labeling to generate a targeted binding area label; Step 3: Based on the targeted binding area marking, a near-infrared fluorescence imaging device is used to collect signal data from the specimen surface point by point, and the wavelength range of the light signal is separated by a regional scanning method. The regional contour is extracted according to the gradient change of the collected signal, the separation results are integrated and the boundary line is marked to generate a marker signal contour map; Step 4: Based on the marker signal contour mapping, the part with signal strength lower than the set threshold is removed, and spatial segmentation is performed according to the distribution characteristics of the optical signal strength in the area, the location of the mutation point is determined and the partition boundary is delineated, and the boundary line of the significant signal change is determined in combination with the spatial segmentation result to generate signal partition boundary data; Step 5: Based on the signal partition boundary data, the change rate is calculated according to the intensity gradient of the boundary signal, the part with abnormal signal fluctuation range is checked point by point, the signal contour is redefined after eliminating the interference signal, and the region is continuously adjusted and optimized based on the characteristics of the signal boundary to generate an optimized tumor tissue boundary map; Step 6: Based on the optimized tumor tissue boundary mapping, a convolutional neural network is used to extract signals from multiple cutting edge sites on the specimen in turn, and the signal distribution and intensity center of each site are calculated by a spatial mapping method, and the signal data is converted into a quantifiable point set in a two-dimensional space. The position relationship of the site signal is annotated according to the overall distribution trend of the point set to generate a cutting edge site signal distribution map; Step 7: Based on the signal distribution map of the cutting edge site, the spatial distribution positions of the signal center and the boundary points are compared point by point, the positioning points are adjusted according to the differences in the overlapping boundary areas, the spatial consistency of the signal boundary is calibrated in combination with the distribution characteristics of the cutting edge site, and a cutting edge area positioning map is generated.

2. The method for localizing fluorescence detection of oral cancer resection margin specimens according to claim 1, characterized in that: The fluorescent targeting probe solution comprises the combined fluorescent targeting peptide, residual graphene oxide and uncombined fluorescent molecules.

3. The method for localizing fluorescence detection of oral cancer resection margin specimens according to claim 1, characterized in that: The target binding region marker includes a fluorescent region formed by the binding of the targeting peptide and the tumor-specific molecule, a distribution range of the marker signal intensity, and a boundary of an unmarked region.

4. The method for localizing fluorescence detection of oral cancer resection margin specimens according to claim 1, characterized in that: The marker signal profile mapping includes the signal boundary position, the spatial distribution of different fluorescence intensity areas and the marker results of the optical signal wavelength range.

5. The method for localizing fluorescence detection of oral cancer resection margin specimens according to claim 1, characterized in that: The signal partition boundary data includes the boundary line of the significant signal partition, the signal strength distribution in the area and the spatial position of the abnormal signal.

6. The method for localizing fluorescence detection of oral cancer resection margin specimens according to claim 1, characterized in that: The specific steps of generating the optimized tumor tissue boundary map are: Based on the signal partition boundary data, the boundary signal strength gradient change is measured point by point, and the signal strength difference is calculated and recorded. The change rate is analyzed point by point for the abnormal fluctuation area and the position is marked to generate abnormal signal position data; Based on the abnormal signal position data, the signal contour shape is adjusted by gradually eliminating abnormal points and interference signals, and the breakpoint area is reconnected and the contour curve is corrected through signal continuity analysis to generate an adjusted signal boundary contour; Based on the adjustment of the signal boundary contour, by optimizing and adjusting the continuity of the signal intensity in the boundary curve area, by correcting and connecting the signal boundary line segment by segment to generate a complete area map, an optimized tumor tissue boundary map is generated; The optimized tumor tissue boundary mapping includes an optimized boundary contour, an adjusted boundary point set, and a regional distribution after eliminating interference signals.

7. The method for localizing fluorescence detection of oral cancer resection margin specimens according to claim 6, characterized in that: The measurement process of the boundary signal intensity gradient change obtains the signal intensity value of each sampling point from the signal partition boundary data, collects the signal value point by point at a fixed interval, calculates the intensity difference between adjacent points, and records the measurement result as a set of continuous gradient change data; The analysis and determination method of the fluctuation abnormal area sets a threshold as the determination standard according to the recorded gradient change data. If the intensity difference of multiple consecutive points exceeds the threshold, it is determined to be a fluctuation abnormal area. The starting point and end point of each abnormal area are recorded as the boundary of the area.

8. The method for localizing fluorescence detection of oral cancer resection margin specimens according to claim 1, characterized in that: The specific steps of generating the signal distribution map of the cutting edge site are: Based on the optimized tumor tissue boundary mapping, the specimen resection margin sites are located one by one and the optical signal data of each site is extracted, and the relationship between the signal and the position is established by recording the signal intensity value and the site coordinates to generate a resection margin site optical signal point set; Based on the optical signal point set of the cutting edge site, the center position of the signal is located by analyzing the optical signal intensity distribution, a preliminary spatial area is delineated in combination with the distribution characteristics of each point, and the point set is reintegrated by the site distribution characteristics to generate cutting edge site distribution data; Based on the distribution data of the cutting edge sites, a convolutional neural network is used to connect the regions by analyzing the relationship between the distribution trend of the point set and the signal intensity, and a two-dimensional distribution map is generated by marking the position relationship of the sites in the spatial distribution range to generate a cutting edge site signal distribution map; The cutting edge site signal distribution diagram includes the spatial coordinates of the site signal, the intensity value of the signal center and the distance relationship between each site.

9. The method for localizing fluorescence detection of oral cancer resection margin specimens according to claim 8, characterized in that: The convolutional neural network, according to the formula: Where: Y represents the fluorescence intensity value of each pixel in the generated cut edge site signal distribution map, w i represents the weight parameter in the convolution kernel, x i represents the fluorescence signal intensity value of the point set in the distribution data of the cutting edge site, and α represents the local density d i The weight coefficient, d i represents the local density of the point set, and β represents the signal intensity gradient z i The weight coefficient, z i represents the gradient value of the fluorescence signal intensity, and γ represents the distance p between the point and the center of the window i The weight coefficient, p i represents the distance between the cutting edge site in the point set and the center of the convolution window, b represents the bias term, and f represents the activation function.

10. The method for localizing fluorescence detection of oral cancer resection margin specimens according to claim 1, characterized in that: The cutting edge region positioning map includes the comprehensive boundary of cutting edge signal distribution, the spatial coordinates of the positioning points and the final region after the boundary adjustment.

Images