General surgery department operation incision image analysis method and system

Abstract

The invention relates to the technical field of image analysis, in particular to a general surgery department surgical incision image analysis method and system, and the method comprises the following steps: removing background noise and a non-target region in an image based on an initially collected surgical incision image, adjusting the brightness and contrast of the image, and compressing and decompressing image data through a self-encoder, and stable features in the image data are extracted. According to the method, the definition of image data and the accuracy of analysis are ensured by finely adjusting the brightness and the contrast ratio of the image and accurately removing irrelevant noise, so that the observation of the incision healing process is more accurate, and the application of the auto-encoder is excellent in the aspect of extracting stable features of the image; according to the method, key information is quickly identified from complex data, an abnormal image deviating from normal healing is identified by using an isolated forest algorithm, and the development trend of abnormity is analyzed through an LSTM network, so that the ability of early judgment of complications is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image analysis, and in particular to a method and system for analyzing images of general surgical incisions. Background Art

[0002] Image analysis technology involves the use of algorithms and models to extract and parse useful information from image data. It is widely used in many fields such as computer vision, medical imaging, satellite remote sensing and robotics. In the medical field, image analysis helps doctors more accurately judge and monitor disease status, such as detecting abnormalities by analyzing X-rays, CT scans or MRI images, so that image analysis can play an important role in feature extraction, classification, pattern recognition and image reconstruction.

[0003] Among them, the general surgery incision image analysis method refers to the application of image analysis technology to evaluate and process the incision images produced in general surgery. Its main purpose is to help doctors evaluate the quality of surgery, monitor the incision healing process and detect complications early by analyzing the images of surgical incisions. For example, by analyzing the size, shape, edges and other visual features of the incision, the state of incision healing can be automatically evaluated, thereby providing patients with more timely medical services.

[0004] Existing technologies lack sufficient image quality control when processing general surgery incision images, such as effective removal of background noise and optimization of contrast, resulting in inaccurate interpretation of image details, affecting the doctor's ability to evaluate the healing process. Existing technologies fail to fully extract key information from images, making it impossible to effectively identify and predict abnormalities that occur during the healing process. The lack of a dynamic abnormality tracking and prediction mechanism makes it difficult to monitor the healing progress in real time in actual applications, increases the difficulty of intervening in the development of complications, and results in the inability to provide patients with the best management strategy, which reduces the healing progress. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a general surgery incision image analysis method and system.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a general surgical incision image analysis method, comprising the following steps:

[0007] S1: Based on the initially acquired surgical incision image, remove the background noise and non-target areas in the image, adjust the image brightness and contrast, and use the autoencoder to compress and decompress the image data to extract the stable features in the image data to obtain the image stable feature set;

[0008] S2: Based on the image stabilization feature set, the isolation forest algorithm is used to calculate the outlier score of the image features, identify abnormal images that deviate from the normal healing progress, and perform secondary screening on the identification results. By comparing with known healing images, the difference from the normal healing process is evaluated to obtain the abnormal image identification;

[0009] S3: Based on the abnormal image identification, the LSTM network is applied to analyze the occurrence frequency and change speed of abnormal features, monitor the abnormal development trend, and compare the existing complication data to identify potential complication types and obtain complication prediction data;

[0010] S4: Based on the complication prediction data, data associated with abnormality types and potential complications are collected, and the postoperative incision management and monitoring process is adjusted according to the collected data to obtain a postoperative management configuration.

[0011] The present invention is improved in that the step of adjusting the image brightness and contrast is specifically as follows:

[0012] S111: Based on the initially acquired surgical incision image, image segmentation technology is applied to identify and remove background noise and non-target areas in the image to obtain a denoised and clear incision image;

[0013] S112: Based on the denoised and cleared incision image, adjust the brightness and contrast using the formula:

[0014] C(I)=α·(II min ) / (I max -I min )+β

[0015] Linearly adjust each pixel value in the image to obtain the image C(I) after brightness and contrast adjustment, where I min and I max are the minimum and maximum brightness values ​​of the image, α is the contrast adjustment coefficient, β is the brightness adjustment coefficient, and I represents the input surgical incision image.

[0016] The present invention is improved in that the steps of acquiring the image stabilization feature set are specifically as follows:

[0017] S121: compressing the image based on the encoding part of the autoencoder, extracting key features in the image, and generating a compressed feature vector;

[0018] S122: Based on the compressed feature vector, the image is restored using the decoding part of the autoencoder, and the decoding process is optimized to capture the image stability feature, using the formula:

[0019]

[0020] Calculate the reconstruction error and stability to obtain the reconstructed image and stability feature index, where F(v) is the feature set of the image, v is the decoded vector, and k is the coefficient for adjusting the decoding sensitivity;

[0021] S123: extracting stable features from the reconstructed image, and screening and determining features with the best stability in the image in combination with the stable feature index to obtain an image stable feature set.

[0022] The present invention is improved in that the step of identifying the abnormal image that deviates from the normal healing progress is specifically:

[0023] S211: Based on the image stabilization feature set, using the isolation forest algorithm, obtaining a feature vector of each image, constructing a random decision tree to isolate abnormal points, and obtaining an original outlier score;

[0024] S212: Based on the original outlier score, standardization is performed to eliminate the scale effect, using the formula:

[0025]

[0026] Get the adjusted outlier score SD normalized , where SD represents the raw outlier score, μ S is the mean of the outlier fraction, σ S is the standard deviation of the outlier scores, C S is the adjustment constant;

[0027] S213: Using the adjusted outlier score, determine the abnormal state of the image, identify images that deviate from the normal healing progress, and obtain a confirmed abnormal image list.

[0028] The present invention is improved in that the step of obtaining the abnormal image identification is specifically as follows:

[0029] S221: extracting key healing features from the screened abnormal images, including skin edge clarity, color uniformity, and skin texture regularity, to obtain feature data information;

[0030] S222: Based on the feature data information, compare the features of each image with the known normal healing image, using the formula:

[0031]

[0032] Get the image deviation score DScore, where G i,j Represents the value of the jth feature of the current image, G ref,j is the typical value of feature j in a standard normal healing image, βg j is the comparison weight of feature j, m g is the total number of features;

[0033] S223: Based on the image deviation score, determine the difference between the image and normal healing, and obtain an abnormal image identification.

[0034] The present invention is improved in that the monitoring steps of the abnormal development trend are specifically as follows:

[0035] S311: Based on the abnormal image identification, record the change information of each feature over time and establish a feature sequence data set;

[0036] S312: Based on the feature sequence data set, the formula is used:

[0037]

[0038] Calculate the change rate and attenuation adjustment ADs of each feature to obtain the abnormal development trend, where ΔF k Represents the change value of feature k at consecutive time points, Δt k Indicates the measurement ΔF k The time interval used, γ Z is the decay rate parameter, is the exponential decay factor, n Z is the total number of features.

[0039] The present invention is improved in that the step of obtaining the complication prediction data is specifically as follows:

[0040] S321: Based on the abnormal development trend, compare with historical complication data, identify existing cases associated with the current abnormal pattern, and obtain a list of similar cases;

[0041] S322: Based on the similar case list, determine the relevance and potential of multiple cases under the current patient status, using the formula:

[0042] RH=∑(β H ×PH i,j )

[0043] Calculate the risk score RH of complications to obtain the complication prediction data, among which PH i,j represents the matching probability between the current patient status and the i-th case on feature j, β H is the weight coefficient.

[0044] The present invention is improved in that the steps of obtaining the postoperative management configuration are specifically as follows:

[0045] S411: Based on the complication prediction data, collect data related to the type and frequency of postoperative complications to obtain decision support data;

[0046] S412: Based on the decision support data, the formula is used:

[0047]

[0048] Quantify the risk of each complication RF, where Fr i represents the severity score of complication i, Rf i is the frequency of complication i, N Y is the total number of observed cases;

[0049] S413: Based on the risk level of the complications, the postoperative incision management and monitoring process is redesigned, including the frequency of incision inspection and medication adjustment, to obtain the postoperative management configuration.

[0050] A general surgical incision image analysis system, the system comprising:

[0051] The image feature processing module removes background noise and non-target areas based on the initially acquired surgical incision images, adjusts brightness and contrast parameters, applies an autoencoder to compress and decompress image data, extracts stable features from image data, and generates an image stability feature set;

[0052] The outlier feature detection module applies the isolation forest algorithm based on the image stabilization feature set to calculate the outlier score of the image features, identify images that deviate from the normal healing progress, and compare and screen with known healing image data to obtain abnormal image identification;

[0053] The abnormal dynamic analysis module uses the LSTM network to analyze the frequency and change rate of abnormal features based on the abnormal image identification, monitor the abnormal development trend, and generate abnormal trend analysis results;

[0054] The complication identification module uses the abnormal trend analysis results to compare the existing complication data, predict potential complication types, and generate complication prediction data;

[0055] The postoperative process optimization module collects data associated with abnormal types and potential complications based on the complication prediction data, adjusts the postoperative incision management and monitoring process, and obtains the postoperative management configuration.

[0056] Compared with the prior art, the advantages and positive effects of the present invention are:

[0057] In the present invention, by carefully adjusting the brightness and contrast of the image and accurately removing irrelevant noise, the clarity of the image data and the accuracy of the analysis are ensured, so that the observation of the incision healing process is more accurate. The application of the autoencoder performs excellently in extracting the stable features of the image, so that key information can be quickly identified from complex data. The isolation forest algorithm is used to identify abnormal images that deviate from normal healing, and the development trend of the abnormality is analyzed through the LSTM network, which greatly improves the ability to judge complications early, allows preventive measures to be taken in advance, and adjusts the postoperative management process in real time, so that the incision treatment can be optimized according to the clinical situation, thereby improving the incision healing effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 A flow chart of a general surgical incision image analysis method is proposed for the present invention;

[0059] Figure 2 A flowchart for adjusting image brightness and contrast in the present invention;

[0060] Figure 3 is a flow chart of obtaining an image stabilization feature set in the present invention;

[0061] Figure 4 is a flow chart for identifying abnormal images that deviate from normal healing progress in the present invention;

[0062] Figure 5 This is a flow chart for obtaining abnormal image identification in the present invention;

[0063] Figure 6 It is a monitoring flow chart of abnormal development trend in the present invention;

[0064] Figure 7 A flowchart for obtaining complication prediction data in the present invention;

[0065] Figure 8 This is a flowchart for obtaining the postoperative management configuration in the present invention. DETAILED DESCRIPTION

[0066] 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.

[0067] 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 positions or positional relationships, are based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0068] Example

[0069] See also Figure 1 The present invention provides a technical solution: a general surgical incision image analysis method, comprising the following steps:

[0070] S1: Based on the initially acquired surgical incision image, remove the background noise and non-target areas in the image, adjust the image brightness and contrast, and use the autoencoder to compress and decompress the image data to extract the stable features in the image data to obtain the image stable feature set;

[0071] S2: Based on the image stability feature set, the isolation forest algorithm is used to calculate the outlier score of the image features, identify abnormal images that deviate from the normal healing progress, and perform secondary screening on the identification results. By comparing with known healing images, the differences from the normal healing process are evaluated and the abnormal image identification is obtained;

[0072] S3: Based on abnormal image identification, the LSTM network is applied to analyze the frequency and change rate of abnormal features, monitor the abnormal development trend, and compare the existing complication data to identify potential complication types and obtain complication prediction data;

[0073] S4: Based on the complication prediction data, collect data associated with abnormality types and potential complications, and adjust the postoperative incision management and monitoring process according to the collected data to obtain the postoperative management configuration.

[0074] The image stability feature set includes brightness balance, texture clarity, and color contrast. Abnormal image identification includes abnormal area positioning, healing stage determination results, and image heterogeneity score. Complication prediction data includes infection probability, atypical healing signals, and healing time prediction results. Postoperative management configuration includes follow-up examination frequency and necessary surgical re-evaluation.

[0075] See also Figure 2 , the steps to adjust the image brightness and contrast are as follows:

[0076] S111: Based on the initially acquired surgical incision image, image segmentation technology is applied to identify and remove background noise and non-target areas in the image to obtain a denoised and clear incision image;

[0077] Image segmentation technology is applied to the initially collected surgical incision images to identify and remove background noise and non-target areas in the images. This operation uses a semantic segmentation network based on deep learning. First, the image is processed, including image normalization and noise filtering. Then the image is analyzed pixel by pixel. Through the multi-layer feature extraction of the neural network, it is determined whether each pixel belongs to the target area. Non-target pixels are marked as background and removed from the image to ensure that only the key parts of the surgical incision are retained in the image, generating a denoised and clear surgical incision image.

[0078] S112: Based on the denoised and cleared cutout image, adjust the brightness and contrast using the formula:

[0079] C(I)=α·(II min ) / (I max -I min )+β

[0080] Linearly adjust each pixel value in the image to obtain the image C(I) after brightness and contrast adjustment, where I min and I max are the minimum and maximum brightness values ​​of the image, α is the contrast adjustment coefficient, β is the brightness adjustment coefficient, and I represents the input surgical incision image;

[0081] Assume α = 1.2 as the contrast adjustment coefficient, β = 0.05 as the brightness adjustment coefficient, I min =30 and I max = 225 is the actual minimum and maximum brightness of the image. For an intermediate brightness value I = 128, the calculation process is:

[0082] C(I)=1.2·(128-30) / (225-30)+0.05

[0083] C(I)=1.2·98 / 195+0.05≈0.603+0.05=0.653

[0084] The result shows that after adjustment, the brightness of the pixel in the image will increase to 65.3% of the maximum brightness, making the image contrast more distinct and helping doctors to more clearly identify the details of the incision area.

[0085] See also Figure 3 , the steps for obtaining the image stabilization feature set are as follows:

[0086] S121: compressing the image based on the encoding part of the autoencoder, extracting key features in the image, and generating a compressed feature vector;

[0087] The image data of the surgical incision is collected and processed by the encoding part of the autoencoder. The data dimension is gradually reduced mainly through the multi-layer feature extraction mechanism of the neural network. During the compression process, the image data is nonlinearly transformed through the activation function to ensure the retention of key visual information while reducing redundant information. This ensures that the feature vector contains the main information of the image, rather than all the details, thereby achieving data compression while retaining the core content of the image. The feature vector will serve as the input for subsequent steps and provide the basis for image reconstruction and further feature extraction.

[0088] S122: Based on the compressed feature vector, the decoded part of the autoencoder is used to restore the image and optimize the decoding process to capture the image stability features using the formula:

[0089]

[0090] Calculate the reconstruction error and stability to obtain the reconstructed image and stability feature index, where F(v) is the feature set of the image, v is the decoded vector, and k is the coefficient for adjusting the decoding sensitivity;

[0091] If we have the following data, let v = 0.5 be an element in the decoded feature vector, and k = 1.5 be the coefficient for adjusting the decoding sensitivity, and substitute it into the calculation:

[0092]

[0093] The result shows that the processed eigenvalue is 0.354, which will be used to determine the features with high stability in the image and help extract a more accurate image feature set. This value is lower than the original eigenvalue, achieving effective adjustment and optimization of the eigenvalue.

[0094] S123: extracting stable features from the reconstructed image, and screening and determining the features with the best stability in the image by combining the stable feature index, to obtain a stable feature set of the image;

[0095] Stable features are extracted from the reconstructed image. Each feature in the image is evaluated based on the stable feature index, which involves comparing the feature value with the threshold to determine which features are stable. In the process, by screening the feature values, only those features that exceed the threshold are considered stable and useful. Such screening criteria ensure the stability and reliability of the feature set, thereby obtaining a high-quality image stability feature set, which provides data support for subsequent image processing or diagnosis.

[0096] See also Figure 4, the specific steps for identifying abnormal images that deviate from the normal healing progress are:

[0097] S211: Based on the image stabilization feature set, the isolation forest algorithm is used to obtain the feature vector of each image, and a random decision tree is constructed to isolate the outliers to obtain the original outlier score;

[0098] Based on the image stabilization feature set, the random forest algorithm is used to process the feature vector of each image. By constructing multiple decision trees to isolate and identify abnormal points, it does not rely on the normal distribution assumption of the data, making it highly efficient when processing image data with complex distribution characteristics. Through this method, the features of each image are evaluated and isolated, and the original outlier score of each image is generated. The score is a preliminary indicator to quantify the degree of abnormality of each image, providing a basis for further analysis and comparison. The key lies in effectively using machine learning technology to identify image features that deviate from the normal healing progress, ensuring that potential abnormal images can be accurately filtered and labeled.

[0099] S212: Based on the original outlier score, standardization is performed to eliminate the scale effect, using the formula:

[0100]

[0101] Get the adjusted outlier score SD normalized , where SD represents the raw outlier score, μ S is the mean of the outlier fraction, σ S is the standard deviation of the outlier scores, C S is the adjustment constant used to adjust the standardized outlier score to more sensitively identify anomalies;

[0102] Given the following data, SD=0.85 is a raw outlier score obtained from the isolation forest, μ S =0.5 is the average value of the scores, σ S =0.1 is the standard deviation, C S =0.05 is the adjustment constant,

[0103]

[0104] The result shows that after normalization, the outlier score of the image is 3.55, which means it is more abnormal than the average level (the average level is 2). It will be used in the subsequent steps to set the threshold to determine whether the image is abnormal.

[0105] S213: using the adjusted outlier score, determining the abnormal state of the image, identifying images that deviate from the normal healing progress, and obtaining a confirmed abnormal image list;

[0106] Using the adjusted outlier score, the abnormal threshold (2) is set. The threshold setting takes into account the distribution of outlier scores of normal healing images and the need to identify abnormal images. The threshold can effectively determine which images deviate from the normal healing progress. The process involves further analysis of the adjusted outlier score, including comparison, screening and confirmation. The result is a confirmed list of abnormal images. The images will be further used in medical diagnosis and treatment processes to provide key visual evidence to support clinical decision-making, ensuring the accuracy and practicality of the output to identify and handle cases that deviate from the normal healing progress.

[0107] See also Figure 5 , the specific steps for obtaining abnormal image identification are:

[0108] S221: extracting key healing features from the screened abnormal images, including skin edge clarity, color uniformity, and skin texture regularity, to obtain feature data information;

[0109] Key healing features are extracted from images that have been identified as abnormal, including clarity of skin edges, uniformity of color, and regularity of skin texture. Each image is scanned in detail and the above features are extracted. The features are directly evaluated through actual image data. The obtained feature data will be used to construct a healing feature profile for each image. The profile records the key visual attributes of each image, providing the necessary basis for further analysis and comparison.

[0110] S222: Based on the feature data information, compare the features of each image with the known normal healing image, using the formula:

[0111]

[0112] Get the image deviation score DScore, where G i,j Represents the value of the jth feature of the current image, G ref,j is the typical value of feature j in a standard normal healing image, βg j is the comparison weight of feature j, m g is the total number of features;

[0113] There are three key features for comparison. For a certain image, its feature value is G i,1 =0.2, G i,2 =0.3 and G i,3 =0.4, the corresponding feature average value of the normal healing image is G ref,1 =0.1, G ref,2 =0.25 and G ref,3 =0.5, the weights are set to βg1=0.5, βg2=1 and βg3=1.5, and the calculation is as follows:

[0114]

[0115] The results show that the calculated deviation score is 0.083, indicating that the image is significantly different from the normal healing image in key features, thus confirming it as an abnormal image.

[0116] S223: Based on the image deviation score, determine the difference between the image and normal healing, and obtain an abnormal image identification;

[0117] Based on the obtained feature data, the deviation score of each image is calculated. The score is calculated by comparing the healing features of each image with the known normal healing image features to obtain the degree of deviation of each image from the normal standard. By setting a threshold, those images with deviation scores exceeding the norm are identified, and it is determined which images show significant differences from normal healing. The images are identified as abnormal images, and a list of abnormal image identifications is obtained.

[0118] See also Figure 6 , the specific steps for monitoring abnormal development trends are:

[0119] S311: Based on the abnormal image identification, record the change information of each feature over time and establish a feature sequence data set;

[0120] Extract key features of each image from the abnormal image dataset, including edge clarity, color uniformity, and texture regularity. Each feature is extracted with a timestamp to ensure that changes in the feature over time can be tracked. Through this process, a feature time series dataset is created. The dataset provides the necessary input for subsequent deep learning analysis, ensuring data integrity and the feasibility of analysis.

[0121] S312: Based on the feature sequence data set, the formula is:

[0122]

[0123] Calculate the change rate and attenuation adjustment ADs of each feature to obtain the abnormal development trend, where ΔF k Represents the change value of feature k at consecutive time points, Δt k Indicates the measurement ΔF k The time interval used, γ Z is the decay rate parameter, which controls the effect of time decay, is an exponential decay factor, which decays with time interval Δt k The increase of n Z is the total number of features;

[0124] If the following data are available, ΔF k = Characteristic change in time Δt k =2,γZ =0.1 time attenuation coefficient,

[0125]

[0126] The result shows that within a given time period, the adjusted dynamic value of the feature is about 2.05, which reflects the rate of change of the feature after adjustment for time decay and is used to evaluate the development trend of abnormal features.

[0127] See also Figure 7 , the specific steps for obtaining complication prediction data are as follows:

[0128] S321: Based on the abnormal development trend, compare with historical complication data, identify existing cases associated with the current abnormal pattern, and obtain a list of similar cases;

[0129] Based on the abnormal development trend currently monitored, data matching and comparison with historical complication records are performed to screen out historical cases similar to the current situation. This comparison process includes analyzing the abnormal characteristics of each case and its development path, and accurately matching cases through time series analysis and pattern recognition to generate a list of similar cases. The list reflects those historical situations that share the same or similar complication risks as the current case, which plays a key role in providing decision support;

[0130] S322: Based on the list of similar cases, determine the relevance and potential of multiple cases under the current patient status, using the formula:

[0131] RH=Σ(β H ×PH i,j )

[0132] Calculate the risk score RH of complications to obtain the complication prediction data, among which PH i,j represents the matching probability between the current patient status and the i-th case on feature j, β H is the weight coefficient, which is adjusted according to the severity and potential of the impact of different complications on the patient's health;

[0133] There are three similar cases with matching probabilities of 0.8, 0.5, and 0.3, respectively. The weights are set to 0.5, 1.0, and 0.5, respectively. Substitute them into the formula for calculation:

[0134] RH=(0.5×0.8)+(1.0×0.5)+(0.5×0.3)=0.4+0.5+0.15=1.05

[0135] The results showed a composite risk score of 1.05, indicating that the patient had a moderate risk of complications, and the data will guide subsequent clinical decisions and patient management.

[0136] See also Figure 8 , the specific steps for obtaining postoperative management configuration are:

[0137] S411: Based on the complication prediction data, collect data related to the types and frequencies of postoperative complications to obtain decision support data;

[0138] Based on the existing complication prediction data, data specific to the types of postoperative complications and their frequency of occurrence are collected, and the occurrence data of various complications are quantitatively analyzed, including the time stamps of the data, severity classifications, and frequency records. Through the data, a decision support dataset is created. The dataset will contain key information extracted from the records, such as the patient's age, gender, preoperative condition, and specific complication details, in order to optimize postoperative management strategies.

[0139] S412: Based on decision support data, use the formula:

[0140]

[0141] Quantify the risk of each complication RF, where Fr i represents the severity score of complication i, Rf i is the frequency of complication i, N Y is the total number of observed cases;

[0142] There are three common complications, with severity scores of Fr i They are 3, 2 and 1 respectively, and the frequency of occurrence is Rf i are 0.05, 0.10 and 0.20 respectively, and the total number of observed cases N Y is 100, substitute it into the formula:

[0143]

[0144]

[0145] The results show that taking into account the severity and probability of complications, the comprehensive risk index is 0.0055, which represents that in the current patient population, based on the given complication types and incidence rates, the overall risk level of postoperative management is low. The value will guide the adjustment and optimization of postoperative management measures.

[0146] S413: Based on the risk level of complications, redesign the postoperative incision management and monitoring process, including the frequency of incision inspection and medication adjustment, to obtain the postoperative management configuration;

[0147] Using the collected complication risk data, we redesigned the postoperative incision management and monitoring process for different risk levels, including adjusting the frequency of incision inspections and the use of necessary anti-inflammatory drugs according to the potential risk of complications, and developing specific treatment guidelines for each risk level to ensure that postoperative incisions are properly managed and reduce the risk of complications, thereby obtaining a postoperative management configuration.

[0148] General surgery incision image analysis system, the system includes:

[0149] The image feature processing module removes background noise and non-target areas based on the initially acquired surgical incision images, adjusts brightness and contrast parameters, applies an autoencoder to compress and decompress image data, extracts stable features from image data, and generates an image stability feature set;

[0150] The outlier feature detection module uses the isolation forest algorithm based on the image stability feature set to calculate the outlier score of the image features, identify images that deviate from the normal healing progress, and compare and screen with known healing image data to obtain abnormal image identification;

[0151] The abnormal dynamic analysis module uses LSTM network to analyze the frequency and change speed of abnormal features based on abnormal image identification, monitor the abnormal development trend, and generate abnormal trend analysis results;

[0152] The complication identification module uses abnormal trend analysis results to compare existing complication data and predict potential complication types to generate complication prediction data;

[0153] The postoperative process optimization module collects data associated with abnormal types and potential complications based on complication prediction data, adjusts the postoperative incision management and monitoring process, and obtains the postoperative management configuration.

[0154] 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 analyzing images of general surgical incisions, characterized in that: The following steps are involved: Based on the initially acquired surgical incision image, background noise and non-target areas in the image are removed, image brightness and contrast are adjusted, and the image data is compressed and decompressed using an autoencoder to extract stable features in the image data to obtain an image stabilization feature set; Based on the image stabilization feature set, an isolation forest algorithm is used to calculate the outlier score of the image features, identify abnormal images that deviate from the normal healing progress, and perform secondary screening on the identification results. By comparing with known healing images, the difference from the normal healing process is evaluated to obtain the abnormal image identification; Based on the abnormal image identification, the LSTM network is applied to analyze the occurrence frequency and change speed of abnormal features, monitor the abnormal development trend, and compare the existing complication data to identify potential complication types and obtain complication prediction data; Based on the complication prediction data, data associated with abnormal types and potential complications are collected, and the postoperative incision management and monitoring process is adjusted according to the collected data to obtain a postoperative management configuration.

2. The general surgery incision image analysis method according to claim 1, characterized in that: The steps of adjusting the image brightness and contrast are specifically as follows: Based on the initially acquired surgical incision image, image segmentation technology is applied to identify and remove background noise and non-target areas in the image to obtain a denoised and clear incision image; Based on the denoised and sharpened incision image, brightness and contrast are adjusted using the formula: C(I)=α·(I-I min ) / (I max -I min )+β Linearly adjust each pixel value in the image to obtain the image C(I) after brightness and contrast adjustment, where I min and I max are the minimum and maximum brightness values ​​of the image, α is the contrast adjustment coefficient, β is the brightness adjustment coefficient, and I represents the input surgical incision image.

3. The general surgery incision image analysis method according to claim 1, characterized in that: The steps of acquiring the image stabilization feature set are specifically as follows: Compressing the image based on the encoding part of the autoencoder, extracting key features in the image, and generating a compressed feature vector; Based on the compressed feature vector, the decoding part of the autoencoder is used to restore the image, and the decoding process is optimized to capture the image stability features, using the formula: Calculate the reconstruction error and stability to obtain the reconstructed image and stability feature index, where F(v) is the feature set of the image, v is the decoded vector, and k is the coefficient for adjusting the decoding sensitivity; The reconstructed image is subjected to stable feature extraction, and the features with the best stability in the image are screened and determined in combination with the stable feature index to obtain an image stable feature set.

4. The general surgery incision image analysis method according to claim 1, characterized in that: The steps of identifying abnormal images that deviate from the normal healing progress are specifically as follows: Based on the image stabilization feature set, using the isolation forest algorithm, obtaining the feature vector of each image, constructing a random decision tree to isolate the outliers, and obtaining the original outlier score; Based on the original outlier score, standardization is performed to eliminate the scale effect, using the formula: Get the adjusted outlier score SD normalized , where SD represents the raw outlier score, μ S is the mean of the outlier fraction, σ S is the standard deviation of the outlier score, C S is the adjustment constant; The adjusted outlier score is used to determine the abnormal state of the image, identify images that deviate from the normal healing progress, and obtain a confirmed abnormal image list.

5. The general surgery incision image analysis method according to claim 1, characterized in that: The steps for obtaining the abnormal image identification are specifically as follows: Extracting key healing features from the screened abnormal images, including skin edge clarity, color uniformity, and skin texture regularity, to obtain feature data information; Based on the characteristic data information, the characteristics of each image are compared with the known normal healing image, using the formula: Get the image deviation score DScore, where G i,j Represents the value of the jth feature of the current image, G ref,j is the typical value of feature j in a standard normal healing image, βg j is the comparison weight of feature j, m g is the total number of features; Based on the image deviation score, the difference between the image and normal healing is determined to obtain an abnormal image identification.

6. The general surgery incision image analysis method according to claim 1, characterized in that: The specific steps of monitoring the abnormal development trend are: Based on the abnormal image identification, record the change information of each feature over time and establish a feature sequence data set; Based on the feature sequence data set, the formula is adopted: Calculate the change rate and attenuation adjustment ADs of each feature to obtain the abnormal development trend, where ΔF k Represents the change value of feature k at consecutive time points, Δt k Indicates the measurement ΔF k The time interval used, γ Z is the decay rate parameter, is the exponential decay factor, n Z is the total number of features.

7. The general surgery incision image analysis method according to claim 1, characterized in that: The steps for obtaining the complication prediction data are specifically as follows: Based on the abnormal development trend, compare with historical complication data, identify existing cases associated with the current abnormal pattern, and obtain a list of similar cases; Based on the similar case list, the relevance and potential of multiple cases under the current patient status are determined using the formula: RH=∑ ( b H ×PH i,j ) Calculate the risk score RH of complications to obtain the complication prediction data, among which PH i,j represents the matching probability between the current patient status and the i-th case on feature j, β H is the weight coefficient.

8. The general surgery incision image analysis method according to claim 1, characterized in that: The steps for obtaining the postoperative management configuration are specifically as follows: Based on the complication prediction data, data related to the types and frequencies of postoperative complications are collected to obtain decision support data; Based on the decision support data, the formula is used: Quantify the risk of each complication RF, where Fr i represents the severity score of complication i, Rf i is the frequency of complication i, N Y is the total number of observed cases; Based on the risk level of the complications, the postoperative incision management and monitoring process was redesigned, including the frequency of incision inspection and medication adjustment, to obtain the postoperative management configuration.

9. General surgery incision image analysis system, characterized in that: According to any one of claims 1 to 8, the general surgical incision image analysis method is implemented, and the system comprises: The image feature processing module removes background noise and non-target areas based on the initially acquired surgical incision images, adjusts brightness and contrast parameters, applies an autoencoder to compress and decompress image data, extracts stable features from image data, and generates an image stability feature set; The outlier feature detection module applies the isolation forest algorithm based on the image stabilization feature set to calculate the outlier score of the image features, identify images that deviate from the normal healing progress, and compare and screen with known healing image data to obtain abnormal image identification; The abnormal dynamic analysis module uses the LSTM network to analyze the frequency and change rate of abnormal features based on the abnormal image identification, monitor the abnormal development trend, and generate abnormal trend analysis results; The complication identification module uses the abnormal trend analysis results to compare the existing complication data, predict potential complication types, and generate complication prediction data; The postoperative process optimization module collects data associated with abnormal types and potential complications based on the complication prediction data, adjusts the postoperative incision management and monitoring process, and obtains the postoperative management configuration.

Images