An image recognition-based digestive tract lesion analysis method and system

Abstract

The present application relates to the technical field of image analysis, in particular to a kind of digestive tract pathological change analysis method and system based on image recognition, based on the endoscope image of patient digestive tract, through image analysis, the edge intensity of pixel in image is evaluated, and by adjusting edge detection threshold, suspect pathological change area in image is identified, non-suspected pathological change area and suspected pathological change area are distinguished, and digestive tract area distinguishing result is obtained.The present application, by dynamically adjusting edge detection threshold, the identification process of suspected pathological change area and non-suspected pathological change area is optimized, by evaluating the shape, size and color of suspected pathological change area, and matching with known pathological change type in database, the accuracy of pathological change type identification is improved, and the image data of corresponding pathological change type is extracted, data support is provided for subsequent medical personnel to confirm pathological change type, by analyzing the deviation of non-suspected pathological change area and normal color data, the evaluation of healthy tissue is refined.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image analysis, in particular to a digestive tract lesion analysis method and system based on image recognition. BACKGROUND

[0002] The technical field of image analysis mainly involves the processing, analysis and understanding of images to extract useful information or features from images to support various decision-making processes. This technical field includes image preprocessing, enhancement, feature extraction, object recognition, classification and interpretation of image content. Image analysis technology is widely used in medical imaging, satellite remote sensing, automated monitoring, machine vision and many other fields, aiming to improve the visibility and interpretability of image data, thereby improving the accuracy and efficiency of decision-making.

[0003] Among them, the digestive tract lesion analysis method focuses on detecting and diagnosing various lesions in the digestive tract such as ulcers, inflammation or tumors through image recognition technology. Using endoscopic images combined with image processing algorithms can help doctors identify suspected lesion areas, assess lesion severity and monitor disease. The application of technology improves the accuracy and speed of diagnosis, reduces the subjective judgment error of doctors, and provides a more accurate diagnosis and treatment method for patients.

[0004] Traditional analysis methods lack sufficient accuracy in distinguishing between lesions and non-suspected lesion areas when dealing with complex images, leading to missed or misdiagnosed lesions, affecting early diagnosis and treatment of the disease. For color judgment, traditional methods cannot accurately distinguish the color changes between normal tissue and slightly abnormal tissue, resulting in incomplete assessment of lesions and inability to provide comprehensive digestive tract risk assessment for medical personnel, affecting the development of treatment plans and disease management. SUMMARY

[0005] The purpose of the present application is to solve the shortcomings in the prior art and to provide a digestive tract lesion analysis method and system based on image recognition.

[0006] In order to achieve the above-mentioned purpose, the present application adopts the following technical scheme: a digestive tract lesion analysis method based on image recognition, comprising the following steps:

[0007] S1: Based on the endoscopic image of the patient's digestive tract, through image analysis, the edge strength of the pixel points in the image is evaluated, and by adjusting the edge detection threshold, the suspected lesion area in the image is identified, the non-suspected lesion area and the suspected lesion area are distinguished, and the digestive tract area distinction result is obtained;

[0008] S2: Based on the digestive tract region differentiation result, the suspected lesion region characteristics, including shape, size, and color, are evaluated for matching with known lesion types at the same location in the lesion database, the matching lesion type is identified, and image data corresponding to the lesion type is extracted to obtain a matching lesion type extraction result;

[0009] S3: Based on the digestive tract region differentiation result, extracting the color data of the non-suspected lesion region and the normal color data of the corresponding part, analyzing the deviation between the non-suspected lesion region and the normal color, evaluating the abnormality degree of the non-suspected lesion region, and obtaining the abnormality evaluation result of the non-suspected lesion region;

[0010] S4: Based on the matching lesion type extraction results and the abnormality assessment results of the non-suspected lesion areas, the risk level of the digestive tract lesions is assessed according to the lesion type confirmed by the medical personnel, combined with the size of the suspected lesion areas, the number of suspected lesion areas, and the degree of abnormality of the non-suspected lesion areas.

[0011] The present invention is improved in that the method for evaluating the edge strength of pixels in an image is:

[0012] S111: converting the image based on the endoscopic image of the patient's digestive tract, converting the image into a grayscale format, and obtaining a grayscale-converted image;

[0013] S112: Based on the grayscale converted image, the formula:

[0014]

[0015] Calculate the edge strength value G of the pixel point to obtain the pixel edge strength evaluation result, where G x and G y Represents the gradient of the pixel in the horizontal and vertical directions, λ G is the weight parameter.

[0016] The present invention is improved in that the steps for obtaining the digestive tract region differentiation result are:

[0017] S121: Based on the pixel point edge intensity evaluation result, calculating the intensity mean and standard deviation of all edge points in the endoscopic image to obtain detection threshold association data;

[0018] S122: Based on the detection threshold correlation data, the formula:

[0019]

[0020] Calculate the adjusted edge detection threshold T new ,in, represents the average value of all edge intensities in the image, σ Tis the standard deviation of edge strength, k T is the adjustment coefficient, p T Is the edge strength higher than The number of pixels, q T is the total number of pixels in the image, ∈ T is the sensitivity parameter, T new is the adjusted edge detection threshold;

[0021] S123: Based on the adjusted edge detection threshold T new And the pixel edge strength evaluation result, the pixel edge strength value G and the adjusted edge detection threshold T new A comparison is performed, and the pixel points with edge strength values ​​higher than the edge detection threshold are marked as the edge of the suspected lesion area, and the non-suspected lesion area and the suspected lesion area are distinguished to obtain the digestive tract area distinction result.

[0022] The present invention is improved in that the steps of obtaining the matching lesion type extraction result are:

[0023] S211: Based on the digestive tract region differentiation result, characteristic data of the suspected lesion region is collected, including shape, size, and color;

[0024] S212: Based on the characteristic data of the suspected lesion area, the characteristics of the known lesion types at the same location in the lesion database are compared, and the formula is used:

[0025]

[0026] Calculate the matching score S, where S is the matching score, x i is the i-th eigenvalue in the current lesion eigenvector, y i is the corresponding feature value of the known lesion type in the database, w i is the weight parameter of the i-th feature, δ is a stability constant, and n is the total number of features;

[0027] S213: Based on the matching score S, compare it with the preset matching threshold. If the matching score S exceeds the matching threshold, mark the lesion type as matched, and extract the image data of the corresponding lesion type to provide data support for medical personnel to confirm the lesion type and obtain the matching lesion type extraction result.

[0028] The present invention is improved in that the method for analyzing the deviation between the non-suspected lesion area and the normal color is:

[0029] S311: Based on the digestive tract region differentiation result, extracting color values ​​of three color channels from the non-suspected lesion region, and extracting normal color data of the corresponding part to obtain color association data;

[0030] S312: Based on the color association data, the formula:

[0031]

[0032] Calculate the color deviation degree D and obtain the color deviation evaluation result, where α D , β D and γ D is the weight coefficient, R nv , G nv and B nv Represents the color values ​​of the red, green, and blue channels of the non-suspected lesion area, R std , G std and B std are the color values ​​of the red, green, and blue channels of the corresponding standard digestive tract, and D is the degree of color deviation.

[0033] The present invention is improved in that the step of obtaining the abnormality assessment result of the non-suspected lesion area is:

[0034] S321: Based on the color deviation assessment result and according to the historical deviation analysis records, extracting the deviation range allowed by the medical staff;

[0035] S322: Based on the allowable deviation range, the formula:

[0036]

[0037] Calculate the abnormality degree E of the non-suspected lesion area and obtain the abnormality assessment result of the non-suspected lesion area, where D is the color deviation degree, D max is the allowable color deviation range, κ E is the scaling factor, and E represents the abnormality of the non-suspected lesion area.

[0038] The present invention is improved in that the method for evaluating the risk level of digestive tract lesions is:

[0039] S411: Based on the matching lesion type extraction result and the non-suspected lesion area abnormality assessment result, according to the lesion type confirmed by the medical personnel, the size of the suspected lesion area, the number of suspected lesion areas, and the degree of abnormality of the non-suspected lesion area are extracted to obtain risk association data;

[0040] S412: Based on the risk association data, the formula:

[0041]

[0042] Calculate the gastrointestinal lesion risk score C to assess the risk of gastrointestinal lesions, where w j is the type weight of the jth suspected lesion area, Aj is the size of the jth suspected lesion area, e is the base of the natural logarithm, N is the number of suspected lesion areas, λ C is the adjustment factor, μ C is the quantity adjustment coefficient, E represents the abnormality of the non-suspected lesion area, and C is the risk score of the digestive tract lesion.

[0043] A digestive tract lesion analysis system based on image recognition, the digestive tract lesion analysis system based on image recognition is used to execute the above-mentioned digestive tract lesion analysis method based on image recognition, the system comprising:

[0044] The suspected lesion area differentiation module is based on the endoscopic image of the patient's digestive tract. Through image analysis, it evaluates the edge strength of the pixel points in the image, and by adjusting the edge detection threshold, it identifies the suspected lesion area in the image, distinguishes between non-suspected lesion areas and suspected lesion areas, and obtains the digestive tract area differentiation result;

[0045] The lesion type matching module evaluates the matching degree of the suspected lesion type with the known lesion type of the same part in the lesion database based on the digestive tract area differentiation result and the characteristics of the suspected lesion area, including shape, size and color, identifies the matching lesion type, and extracts the image data of the corresponding lesion type to obtain the matching lesion type extraction result;

[0046] The color deviation analysis module extracts the color of the non-suspected lesion area and the normal color data of the corresponding part based on the digestive tract area differentiation result, analyzes the deviation between the non-suspected lesion area and the normal color, and obtains a color deviation evaluation result;

[0047] The non-suspected lesion area abnormality assessment module assesses the abnormality degree of the non-suspected lesion area based on the color deviation assessment result, and obtains the non-suspected lesion area abnormality assessment result;

[0048] The gastrointestinal risk analysis module evaluates the risk level of gastrointestinal lesions based on the matching lesion type extraction results and the abnormality assessment results of the non-suspected lesion areas, according to the lesion type confirmed by medical personnel, combined with the size of the suspected lesion areas, the number of suspected lesion areas, and the degree of abnormality of the non-suspected lesion areas.

[0049] Compared with the prior art, the advantages and positive effects of the present invention are:

[0050] In the present invention, by dynamically adjusting the edge detection threshold, the identification process of suspected lesion areas and non-suspected lesion areas is optimized, and the accuracy of image recognition is enhanced. By evaluating the shape, size and color of the suspected lesion area and matching it with the known lesion types in the database, the accuracy of lesion type identification is improved, and image data of the corresponding lesion type is extracted, providing data support for subsequent medical personnel to confirm the lesion type. By analyzing the deviation between the non-suspected lesion area and the normal color data, the assessment of healthy tissue is refined, and minor lesions in other tissues can be discovered. By comprehensively evaluating the risk level of the lesion, a comprehensive disease risk assessment tool is provided for medical personnel, providing more comprehensive decision-making support for clinicians. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a flow chart of the method of the present invention;

[0052] Figure 2 This is a flow chart of the present invention for evaluating the edge strength of pixels in an image;

[0053] Figure 3 A flow chart of obtaining digestive tract region differentiation results according to the present invention;

[0054] Figure 4 A flowchart of the present invention for obtaining matching lesion type extraction results;

[0055] Figure 5 A flowchart of analyzing the deviation between the non-suspected lesion area and the normal color of the present invention;

[0056] Figure 6 A flow chart for obtaining abnormal assessment results of non-suspected lesion areas according to the present invention;

[0057] Figure 7 The figure is a flow chart for evaluating the risk level of digestive tract lesions according to the present invention. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to 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] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings and are only for the convenience of describing the present invention and simplifying the description. They do not indicate or imply that the devices or elements referred to must have a specific direction, be constructed and operate in a specific direction, and therefore should not be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0060] See also Figure 1 The present invention provides a technical solution: a method for analyzing digestive tract lesions based on image recognition, comprising the following steps:

[0061] S1: Based on the endoscopic image of the patient's digestive tract, the edge strength of the pixel points in the image is evaluated through image analysis. The suspected lesion area in the image is identified by adjusting the edge detection threshold, and the non-suspected lesion area and the suspected lesion area are distinguished to obtain the digestive tract area distinction result;

[0062] S2: Based on the digestive tract region differentiation results, the suspected lesion area features, including shape, size, and color, are evaluated for matching with known lesion types in the lesion database. The matching lesion type is identified and image data of the corresponding lesion type is extracted to provide data support for medical personnel to confirm the lesion type and obtain the matching lesion type extraction result;

[0063] S3: Based on the digestive tract region differentiation results, extract the color data of the non-suspected lesion area and the normal color data of the corresponding part, analyze the deviation between the non-suspected lesion area and the normal color, evaluate the abnormality degree of the non-suspected lesion area, and obtain the abnormality assessment result of the non-suspected lesion area;

[0064] S4: Based on the matching lesion type extraction results and the abnormality assessment results of the non-suspected lesion areas, the risk level of the digestive tract lesions is assessed according to the lesion type confirmed by the medical staff, combined with the size of the suspected lesion areas, the number of suspected lesion areas, and the degree of abnormality of the non-suspected lesion areas.

[0065] The digestive tract area differentiation results include the pixel sets of suspected lesion areas and the pixel sets of non-suspected lesion areas. The matching lesion type extraction results include the matching lesion type labels, matching scores and corresponding image data. The non-suspected lesion area abnormality assessment results include color deviation values ​​and abnormality degree indicators. The digestive tract lesion risk assessment results include risk ratings and patient digestive tract risk information.

[0066] See also Figure 2 , the method for evaluating the edge strength of pixels in an image is:

[0067] S111: converting the image based on the endoscopic image of the patient's digestive tract, converting the image into a grayscale format, and obtaining a grayscale-converted image;

[0068] S112: Based on the grayscale converted image, the formula:

[0069]

[0070] Calculate the edge strength value G of the pixel point to obtain the pixel edge strength evaluation result, where G x and G y Represents the gradient of the pixel in the horizontal and vertical directions, λ G is the weight parameter.

[0071] formula:

[0072]

[0073] Parameter details and how to obtain them:

[0074] G x and G y These are the horizontal and vertical gradients of an image. These gradients are used for edge detection. These two gradients are typically obtained by applying a Sobel filter to an image. The Sobel filter is an edge detection operator that contains two kernels (one for the horizontal direction and one for the vertical direction), which are convolved with the image region to calculate the gradient. The filter kernel is applied by sliding it across each pixel in the image, calculating the weighted sum of adjacent pixels with each shift.

[0075] λ G This parameter is used to weight the response to diagonal edges. The parameter value is selected based on predefined criteria to optimize edge detection performance, particularly in images with pronounced diagonal edges. This parameter is typically adjusted by the algorithm developer based on their needs, and the optimal value is selected through experimental comparison of the effects of different values ​​on the results.

[0076] Calculation example:

[0077] Set in the image area, use Sobel filter to calculate G x and G y Get G x =120 and G y =80, set λ G =0.5.

[0078] The process of calculating G is as follows:

[0079]

[0080] The calculated G=160 represents the strength of the image edge in the selected area. A higher G value usually indicates that the area has significant edge features.

[0081] See also Figure 3 , the steps to obtain the digestive tract region differentiation results are:

[0082] S121: Based on the pixel edge intensity evaluation result, the intensity mean and standard deviation of all edge points in the endoscopic image are calculated to obtain detection threshold association data;

[0083] S122: Correlate data based on the detection threshold, using the formula:

[0084]

[0085] Calculate the adjusted edge detection threshold T new ,in, represents the average value of all edge intensities in the image, σ T is the standard deviation of edge strength, k T is the adjustment coefficient, p T Is the edge strength higher than The number of pixels, q T is the total number of pixels in the image, ∈ T is the sensitivity parameter, T new is the adjusted edge detection threshold;

[0086] S123: Based on the adjusted edge detection threshold T new And the pixel edge strength evaluation result, the pixel edge strength value G and the adjusted edge detection threshold T new A comparison is performed, and the pixel points with edge strength values ​​higher than the edge detection threshold are marked as the edge of the suspected lesion area, and the non-suspected lesion area and the suspected lesion area are distinguished to obtain the digestive tract area distinction result.

[0087] formula:

[0088]

[0089] Parameter details and how to obtain them

[0090] It represents the average value of all edge strengths in the image, obtained by averaging the edge strength values ​​of each pixel.

[0091] σ T Is the standard deviation of edge intensity, which is used to measure the dispersion of edge intensity values. Calculated using the standard deviation formula.

[0092] kT It is an adjustment coefficient used to adjust the flexibility of the threshold and is usually set by the experimenter according to the characteristics of the image and the requirements of edge detection.

[0093] p T Is the edge strength higher than The number of pixels with edge strength values ​​greater than Pixels are obtained.

[0094] q T It is the total number of pixels in the image and can be obtained directly from the image resolution.

[0095] ∈ T It is a sensitivity parameter, and its value is usually set by the experimenter according to the requirements of edge detection accuracy and false alarm rate.

[0096] Calculation example:

[0097] Set in an image area, the data is as follows: σ T =10,k T =1.5, p T =1500,q T = 10000 (image of 100x100 pixels).

[0098] Calculation process:

[0099]

[0100] The calculated threshold T new =64.048, indicating the adjusted edge detection threshold, which takes into account the average intensity and standard deviation of the distribution of edges in the image. A high threshold will lead to stricter edge recognition and more accurate identification of truly significant edges.

[0101] See also Figure 4 , the steps to obtain the matching lesion type extraction results are:

[0102] S211: Based on the digestive tract region differentiation results, characteristic data of the suspected lesion area is collected, including shape, size, and color;

[0103] S212: Based on the characteristic data of the suspected lesion area, compare it with the characteristics of the known lesion types at the same location in the lesion database, using the formula:

[0104]

[0105] Calculate the matching score S, where S is the matching score, x i is the i-th eigenvalue in the current lesion eigenvector, y iis the corresponding feature value of the known lesion type in the database, w i is the weight parameter of the i-th feature, δ is a stability constant, and n is the total number of features;

[0106] S213: Based on the matching score S, it is compared with the preset matching threshold. If the matching score S exceeds the matching threshold, the lesion type is marked as matched, and the image data of the corresponding lesion type is extracted to provide data support for medical personnel to confirm the lesion type and obtain the matching lesion type extraction result.

[0107] formula:

[0108]

[0109] Parameter details and how to obtain them:

[0110] w i is the weight parameter for the i-th feature, reflecting the importance of different features in identifying lesion types. These weights are usually set based on historical data analysis or expert clinical experience to optimize the influence of each feature in the matching process.

[0111] x i It is the i-th eigenvalue in the current lesion feature vector, which is extracted from the patient's endoscopic image through image processing algorithms (such as edge detection, color analysis, etc.).

[0112] y i It is the corresponding characteristic value of a known lesion type in the lesion database. The value is usually entered by medical experts based on historical cases and literature records when the lesion database is established.

[0113] δ is a stability constant used to prevent the denominator from being zero and ensure the stability of the calculation. This value is usually set to a very small number to minimize its impact on the denominator while maintaining the validity of the calculation.

[0114] Calculation example:

[0115] Three features (shape, size, and color) are set for lesion type matching, with weights w1 = 0.3, w2 = 0.5, and w3 = 0.2, respectively. The feature value of the current lesion is set to x = [1.0, 2.0, 5.0], the feature value of a lesion type in the database is set to y = [0.8, 2.1, 4.8], and δ = 0.001.

[0116] Calculate the weighted similarity for each feature:

[0117] For shape i=1:

[0118]

[0119] For size i=2:

[0120]

[0121] For color i=3:

[0122]

[0123] Calculate the total similarity score S:

[0124] S=S1+S2+S3

[0125] =0.267+0.488+0.196

[0126] =0.951

[0127] The obtained similarity score S = 0.951 is high, indicating that the characteristics of the current lesion are relatively close to the lesion types in the database, supporting further medical diagnosis and treatment decisions, and helping doctors to quickly and accurately confirm the lesion type.

[0128] See also Figure 5 ,The method for analyzing the deviation between the non-suspected lesion area and the normal color is:

[0129] S311: Based on the digestive tract region differentiation result, extract the color values ​​of the three color channels from the non-suspected lesion region, and extract the normal color data of the corresponding part to obtain color association data;

[0130] S312: Based on the color association data, the formula:

[0131]

[0132] Calculate the color deviation degree D and obtain the color deviation evaluation result, where α D , β D and γ D is the weight coefficient, R nv , G nv and B nv Represents the color values ​​of the red, green, and blue channels of the non-suspected lesion area, R std , G std and B std are the color values ​​of the red, green, and blue channels of the corresponding standard digestive tract, and D is the degree of color deviation.

[0133] formula:

[0134]

[0135] Parameter details and how to obtain them:

[0136] αD , β D and γ D are the color channel weight coefficients, corresponding to the red, green, and blue channels respectively. The coefficients are usually set based on clinical research or the experience of image processing experts to adjust the importance of each color in the overall color deviation assessment.

[0137] R nv , G nv and B nv The color values ​​of the red, green, and blue channels are extracted from non-suspected lesion areas. The values ​​are measured directly from the endoscopic image using image processing software.

[0138] R std , G std and B std It is the color value of the red, green, and blue channels of the corresponding standard digestive tract, usually obtained from a standard image library of the healthy digestive tract area.

[0139] Calculation example:

[0140] In a specific endoscopy, the following color values ​​are obtained from the non-suspected lesion area: Color value R of the non-suspected lesion area nv =150, G nv =100, B nv =90, standard digestive tract area color value R std =140, G std =110, B std =95, the weight coefficient is set to α D =0.3,β D =0.4,γ D =0.3.

[0141] Calculate the color deviation contribution of the red channel:

[0142] α D ·|R nv -R std |=0.3·|150-140|=0.3·10=3

[0143] Calculate the color deviation contribution of the green channel:

[0144]

[0145] Calculate the color deviation contribution of the blue channel:

[0146] γ D ln(1+|B nv -B std |)=0.3·ln(1+|90-95|)=0.3·ln(6)≈0.54

[0147] Calculate the color deviation degree D:

[0148] D≈3+1.26+0.54≈4.8

[0149] The calculated color deviation value D≈4.8 provides a quantitative assessment of the degree of difference in visual color between non-suspected lesion areas and normal digestive tract sites.

[0150] See also Figure 6 , the steps for obtaining abnormal assessment results in non-suspected lesion areas are:

[0151] S321: Based on the color deviation assessment results and historical deviation analysis records, extract the deviation range allowed by medical staff;

[0152] S322: Based on the allowable deviation range, by the formula:

[0153]

[0154] Calculate the abnormality degree E of the non-suspected lesion area and obtain the abnormality assessment result of the non-suspected lesion area, where D is the color deviation degree, D max is the allowable color deviation range, κ E is the scaling factor, and E represents the abnormality of the non-suspected lesion area.

[0155] formula:

[0156]

[0157] Parameter details and how to obtain them:

[0158] D is the degree of color deviation, calculated in the previous step.

[0159] D max The color deviation range is usually determined based on clinical experience or historical data. This threshold reflects the maximum acceptable range of color difference between non-suspected lesion areas and normal digestive tract.

[0160] κ E is a scaling factor that adjusts the sensitivity of the abnormality calculation. It can be adjusted based on specific clinical needs to ensure the accuracy and practicality of the abnormality assessment.

[0161] Calculation example:

[0162] Set the following parameters and color deviation data:

[0163] Color deviation degree D = 50, the allowable color deviation range, D max =40, scaling factor κ E =1.5.

[0164] Calculate the degree of abnormality:

[0165]

[0166] The calculated abnormality level, E=0.375, indicates that the currently non-suspected lesion area exhibits significant abnormality relative to the maximum allowable color deviation, indicating a significant color difference in the area and requiring further medical evaluation or monitoring. This provides doctors with a quantitative tool to assess and monitor changes in digestive tract health.

[0167] See also Figure 7 The method for assessing the risk of gastrointestinal lesions is as follows:

[0168] S411: Based on the matching lesion type extraction results and the non-suspected lesion area abnormality assessment results, according to the lesion type confirmed by the medical personnel, the size of the suspected lesion area, the number of suspected lesion areas, and the degree of abnormality of the non-suspected lesion area are extracted to obtain risk association data;

[0169] S412: Based on the hazard association data, the formula:

[0170]

[0171] Calculate the gastrointestinal lesion risk score C to assess the risk of gastrointestinal lesions, where w j is the type weight of the jth suspected lesion area, A j is the size of the jth suspected lesion area, e is the base of the natural logarithm, N is the number of suspected lesion areas, λ C is the adjustment factor, μ C is the quantity adjustment coefficient, E represents the abnormality of the non-suspected lesion area, and C is the risk score of the digestive tract lesion.

[0172] formula:

[0173]

[0174] Parameter details and how to obtain them:

[0175] w j : Represents the type weight of the jth suspected lesion area. The weight is based on long-term clinical research and historical data analysis, and reflects the medical severity of different lesion types.

[0176] A j : is the size of the jth suspected lesion area, which is usually quantified from the endoscopic image by image processing software, such as calculating the pixel area of ​​the suspected lesion area for evaluation.

[0177] N: is the number of suspected lesion areas, which is directly obtained from image analysis.

[0178] λ C : is the weight adjustment factor for the abnormality level E in non-suspected lesion areas. The parameter adjusts the contribution of E to the overall assessment and is customized based on medical research and actual needs.

[0179] E: The abnormality degree of the non-suspected lesion area, calculated by the previous step.

[0180] μ C : Quantity adjustment coefficient, which adjusts the impact of the number N of suspected lesion areas. It is set through clinical data to ensure that the increase in quantity is appropriate to the increase in lesion risk.

[0181] Calculation example:

[0182] Set the following parameters:

[0183] The number of suspected lesion areas N = 3, the size of the suspected lesion area

[0184] A1=1.2, A2=0.8, A3=1.5, type weight w1=0.6, w2=0.8, w3=0.9, abnormality degree of non-suspected lesion area E=0.4, weight adjustment factor λ C =1.2, quantity influence coefficient μ C =2.

[0185] Calculation process:

[0186] Calculate the risk contribution of each suspected lesion area:

[0187]

[0188] Sum up the risk contributions of all suspected lesion areas:

[0189] 1.992+1.780+4.034=7.806

[0190] Calculate the weighted abnormality degree of non-suspected lesion areas:

[0191] λ C ·E=1.2·0.4=0.48

[0192] Calculate the total base risk plus the degree of non-lesional abnormality:

[0193] 7.806+0.48=8.286

[0194] The impact of the number of suspected lesion areas:

[0195] C=8.286·ln(1+N·μ C )

[0196] =8.286·ln(1+3·2)

[0197] =8.286·ln(7)

[0198] ≈8.286·1.946

[0199] ≈16.126

[0200] The calculated C = 16.126 indicates that, given the size and number of suspected lesions and the degree of abnormality in non-suspected lesions, the patient's risk of gastrointestinal lesions is high. This quantitative risk index can help doctors assess the severity of a patient's condition and formulate appropriate treatment plans. The calculation reflects the combined impact of the number and size of suspected lesions and the abnormality of non-suspected lesions on the overall risk level, providing a comprehensive assessment tool.

[0201] A digestive tract lesion analysis system based on image recognition, which is used to execute the above-mentioned digestive tract lesion analysis method based on image recognition, comprises:

[0202] The suspected lesion area differentiation module is based on the endoscopic image of the patient's digestive tract. Through image analysis, it evaluates the edge strength of the pixel points in the image, and by adjusting the edge detection threshold, it identifies the suspected lesion area in the image, distinguishes the non-suspected lesion area from the suspected lesion area, and obtains the digestive tract area differentiation result;

[0203] The lesion type matching module, based on the digestive tract region differentiation results, evaluates the degree of match with known lesion types in the same area in the lesion database based on the characteristics of the suspected lesion region, including shape, size, and color. It then identifies the matching lesion type and extracts the image data corresponding to the lesion type, providing data support for medical personnel to confirm the lesion type and obtain the matching lesion type extraction result.

[0204] The color deviation analysis module extracts the color of the non-suspected lesion area and the normal color data of the corresponding part based on the digestive tract area differentiation results, analyzes the deviation between the non-suspected lesion area and the normal color, and obtains the color deviation evaluation result;

[0205] The non-suspected lesion area abnormality assessment module assesses the abnormality degree of the non-suspected lesion area based on the color deviation assessment result, and obtains the non-suspected lesion area abnormality assessment result;

[0206] The gastrointestinal risk analysis module is based on the matching lesion type extraction results and the abnormality assessment results of non-suspected lesion areas. According to the lesion type confirmed by medical personnel, combined with the size of suspected lesion areas, the number of suspected lesion areas, and the degree of abnormality of non-suspected lesion areas, it assesses the risk level of gastrointestinal lesions.

[0207] The above merely describes the preferred embodiments of the present application, but does not limit the present application in other forms. Any person skilled in the art can make changes or modifications to the above disclosed technical contents 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 according to the technical essence of the present application without departing from the technical solution content of the present application still falls within the protection scope of the present application.

Claims

1. A method for analyzing digestive tract lesions based on image recognition, characterized in that: The following steps are involved: Based on the endoscopic image of the patient's digestive tract, the edge strength of the pixel points in the image is evaluated through image analysis, and the suspected lesion area in the image is identified by adjusting the edge detection threshold, and the non-suspected lesion area and the suspected lesion area are distinguished to obtain the digestive tract area distinction result; The steps for obtaining the digestive tract region differentiation result are: Based on the pixel point edge intensity evaluation result, calculating the intensity mean and standard deviation of all edge points in the endoscopic image to obtain detection threshold association data; Based on the detection threshold correlation data, the formula: ； Calculate the adjusted edge detection threshold ,in, represents the average value of all edge intensities in the image, is the standard deviation of edge strength, is the adjustment coefficient, Is the edge strength higher than The number of pixels, is the total number of pixels in the image, is the sensitivity parameter, is the adjusted edge detection threshold; Based on the adjusted edge detection threshold And the pixel edge strength evaluation result, the pixel edge strength value With the adjusted edge detection threshold Compare and mark the pixels whose edge strength value is higher than the edge detection threshold as the edge of the suspected lesion area, distinguish the non-suspected lesion area from the suspected lesion area, and obtain the digestive tract area distinction result; Based on the digestive tract region differentiation result, the suspected lesion region characteristics, including shape, size, and color, are evaluated for matching with known lesion types in the same location in the lesion database, the matching lesion type is identified, and image data corresponding to the lesion type is extracted to obtain a matching lesion type extraction result; Based on the digestive tract region differentiation result, extracting the color data of the non-suspected lesion region and the normal color data of the corresponding part, analyzing the deviation between the non-suspected lesion region and the normal color, evaluating the abnormality degree of the non-suspected lesion region, and obtaining the abnormality evaluation result of the non-suspected lesion region; The method for analyzing the deviation between the non-suspected lesion area and the normal color is: Based on the digestive tract region differentiation result, extracting color values ​​of three color channels from the non-suspected lesion area, and extracting normal color data of the corresponding part to obtain color association data; Based on the color association data, the formula: ； Calculate the degree of color deviation , and obtain the color deviation evaluation result, where 、 and is the weight coefficient, 、 and Represent the color values ​​of the red, green, and blue channels of the non-suspected lesion area, 、 and They are the color values ​​of the red, green, and blue channels of the corresponding standard digestive tract. is the degree of color deviation; Based on the matching lesion type extraction results and the non-suspected lesion area abnormality assessment results, the risk level of the digestive tract lesion is assessed according to the lesion type confirmed by the medical personnel, combined with the size of the suspected lesion area, the number of suspected lesion areas, and the degree of abnormality of the non-suspected lesion area.

2. The digestive tract lesion analysis method based on image recognition according to claim 1, characterized in that: The method for evaluating the edge strength of pixels in an image is: Based on an endoscopic image of the patient's digestive tract, converting the image into a grayscale format to obtain a grayscale converted image; Based on the grayscale converted image, the formula: ； Calculate the edge strength value of the pixel , get the pixel edge strength evaluation result, where, and Represent the gradient of the pixel in the horizontal and vertical directions respectively, is the weight parameter.

3. The method for analyzing digestive tract lesions based on image recognition according to claim 1, characterized in that: The steps for obtaining the matching lesion type extraction result are: Based on the digestive tract region differentiation result, collecting characteristic data of the suspected lesion area, including shape, size and color; Based on the characteristic data of the suspected lesion area, the characteristics of the known lesion types at the same location in the lesion database are compared, and the formula is used: ； Calculating the matching score ,in, is the matching score, is the first eigenvalues, is the corresponding feature value of the known lesion type in the database, It is The weight parameter of each feature, is the stability constant, is the total number of features; Based on the matching score , compared with the preset matching threshold, if the matching score If the matching degree threshold is exceeded, the lesion type is marked as matched, and the image data of the corresponding lesion type is extracted to provide data support for medical personnel to confirm the lesion type and obtain the matching lesion type extraction result.

4. The method for analyzing digestive tract lesions based on image recognition according to claim 1, characterized in that: The steps for obtaining the abnormality assessment results of the non-suspected lesion area are: Based on the color deviation assessment results and according to historical deviation analysis records, extracting the deviation range allowed by medical staff; Based on the allowable deviation range, the formula: ； Calculate the degree of abnormality in non-suspected lesion areas , and obtain the abnormal assessment results of non-suspected lesion areas, among which, is the degree of color deviation, is the allowable color deviation range, is the scaling factor, Indicates the degree of abnormality in non-suspected lesion areas.

5. The method for analyzing digestive tract lesions based on image recognition according to claim 1, characterized in that: The method for evaluating the risk level of digestive tract lesions is as follows: Based on the matching lesion type extraction results and the non-suspected lesion area abnormality assessment results, according to the lesion type confirmed by the medical personnel, the size of the suspected lesion area, the number of suspected lesion areas, and the degree of abnormality of the non-suspected lesion area are extracted to obtain risk association data; Based on the risk association data, the formula: ； Calculation of gastrointestinal lesion risk score , assess the risk of gastrointestinal lesions, including: It is The type weight of the suspected lesion area, It is The size of the suspected lesion area, is the base of natural logarithms, is the number of suspected lesion areas, is the regulating factor, is the quantity adjustment coefficient, Represents the degree of abnormality in non-suspected lesion areas, Score the risk of gastrointestinal lesions.

6. A digestive tract lesion analysis system based on image recognition, characterized in that: The method for analyzing digestive tract lesions based on image recognition according to any one of claims 1 to 5, wherein the system comprises: The suspected lesion area differentiation module is based on the endoscopic image of the patient's digestive tract. Through image analysis, it evaluates the edge strength of the pixel points in the image, and by adjusting the edge detection threshold, it identifies the suspected lesion area in the image, distinguishes the non-suspected lesion area from the suspected lesion area, and obtains the digestive tract area differentiation result; The lesion type matching module evaluates the matching degree of the suspected lesion type with the known lesion type of the same part in the lesion database based on the digestive tract area differentiation result and the characteristics of the suspected lesion area, including shape, size and color, identifies the matching lesion type, and extracts the image data of the corresponding lesion type to obtain the matching lesion type extraction result; The color deviation analysis module extracts the color of the non-suspected lesion area and the normal color data of the corresponding part based on the digestive tract area differentiation result, analyzes the deviation between the non-suspected lesion area and the normal color, and obtains a color deviation evaluation result; The non-suspected lesion area abnormality assessment module assesses the abnormality degree of the non-suspected lesion area based on the color deviation assessment result, and obtains the non-suspected lesion area abnormality assessment result; The gastrointestinal risk analysis module evaluates the risk level of gastrointestinal lesions based on the matching lesion type extraction results and the abnormality assessment results of the non-suspected lesion areas, according to the lesion type confirmed by medical personnel, combined with the size of the suspected lesion area, the number of suspected lesion areas, and the degree of abnormality of the non-suspected lesion areas.

Images