Method and system for identifying intestinal tract cleaning level of enteroscopy patient

Abstract

The invention discloses a method and a system for identifying the intestinal tract cleaning level of an enteroscopy patient. The method comprises the following steps: acquiring a multi-angle dirt image set of intestinal tract excretion of a patient to be subjected to enteroscopy; for each dirt image in the dirt image set, extracting a feature vector of each dimension feature related to the intestinal tract cleaning degree in the dirt image; for each dimension feature, combining all feature vectors of the extracted dimension feature to obtain a combined feature vector; and based on the combined feature vector corresponding to each dimension feature, utilizing a machine learning algorithm to evaluate the intestinal tract cleaning level of the patient. According to the embodiment of the invention, by comprehensively considering the multi-dimensional features and fully utilizing the advanced machine learning technology, objective evaluation of the intestinal tract cleaning level of the enteroscopy patient is realized, and the accuracy and reliability of the evaluation result are enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of image recognition, and in particular to a method and system for recognizing the intestinal cleanliness level of a patient with a colonoscopy. Background Art

[0002] Colonoscopy is a commonly used medical diagnostic tool for detecting and treating intestinal diseases such as colon cancer, inflammatory bowel disease, and other intestinal abnormalities. Before undergoing a colonoscopy, patients usually need to perform intestinal cleansing to ensure that there is no fecal residue inside the intestine, thereby improving the accuracy and safety of the examination. However, traditional intestinal cleansing preparation methods rely on patient self-reporting and the doctor's subjective judgment, resulting in large human factors and errors in the assessment of the degree of cleanliness. Summary of the invention

[0003] The purpose of the present invention is to provide a method and system for identifying the intestinal cleanliness level of patients undergoing colonoscopy, so as to address the deficiencies in the prior art. By comprehensively considering multidimensional features and making full use of advanced machine learning technology, the present invention can achieve an objective evaluation of the intestinal cleanliness level of patients undergoing colonoscopy, thereby enhancing the accuracy and reliability of the evaluation results.

[0004] One embodiment of the present application provides a method for identifying the intestinal cleanliness level of a patient undergoing colonoscopy, the method comprising:

[0005] Collecting a set of waste images from multiple angles of intestinal excretion from patients who are about to undergo colonoscopy;

[0006] For each dirt image in the dirt image set, extracting a feature vector of each dimensional feature related to the intestinal cleanliness degree in the dirt image;

[0007] For each dimensional feature, all feature vectors extracted from the dimensional feature are combined to obtain a combined feature vector;

[0008] Based on the combined feature vector corresponding to each dimensional feature, a machine learning algorithm is used to evaluate the patient's intestinal cleanliness level.

[0009] Optionally, the dimensional features include:

[0010] Solid morphology characteristics, liquid morphology characteristics and color distribution characteristics;

[0011] The step of extracting the feature vector of each dimensional feature related to the intestinal cleanliness degree in each dirt image in the dirt image set includes:

[0012] applying morphological operations to the dirt image to remove noise and enhance the boundaries of solid objects, and using edge detection algorithms to identify and enhance the outlines of solid parts;

[0013] Extracting a shape descriptor of a solid in the dirt image to quantify the geometric properties of the solid morphology, and converting the shape descriptor into a solid feature vector corresponding to the solid morphological feature;

[0014] The dirt image is preprocessed by using Gaussian filtering to remove noise, clarify the boundary between the liquid and other parts, and use threshold segmentation or K-means clustering to segment the image to identify the liquid area;

[0015] The optical flow method is used to detect the movement of the liquid area, and the speed and direction features of each pixel are calculated. The speed and direction features are encoded into dynamic texture feature vectors to describe the dynamic characteristics of the liquid.

[0016] From the identified liquid area, the spectrum characteristics of the liquid contour are analyzed by Fourier transform to extract morphological features, and a morphological feature vector corresponding to the morphological features is encoded to describe the geometric and morphological changes of the liquid, wherein the morphological features include contour length and curvature changes;

[0017] The flow characteristics of the liquid are simulated by the difference method and image registration, and the deformation of the liquid at different time points in the image sequence is calculated to extract the flow velocity distribution and viscosity characteristics of the fluid, and the flow velocity distribution and viscosity characteristics are encoded in numerical form as a fluid feature vector;

[0018] Combining the dynamic texture feature vector, the morphological feature vector, and the fluid feature vector into a liquid feature vector corresponding to the liquid morphological feature;

[0019] Converting the dirt image from the RGB color space to the Lab color space, and generating a color histogram of the dirt image based on the Lab color space to quantify the distribution frequency of each color in the image;

[0020] A color feature vector is constructed based on the values ​​in the color histogram to represent the proportion of different colors in the image.

[0021] Optionally, for each dimensional feature, all feature vectors extracted from the dimensional feature are combined to obtain a combined feature vector, including:

[0022] For each dimensional feature, all feature vectors extracted for the dimensional feature are weighted summed to obtain a combined feature vector.

[0023] Optionally, the step of evaluating the intestinal cleanliness level of the patient using a machine learning algorithm based on the combined feature vector corresponding to each dimensional feature includes:

[0024] The combined feature vector corresponding to each dimensional feature is input into the corresponding pre-trained intestinal cleanliness assessment sub-model based on machine learning, and the sub-score of the intestinal cleanliness degree corresponding to each dimensional feature is output, wherein one dimensional feature corresponds to the training of one intestinal cleanliness assessment sub-model;

[0025] Perform a weighted summation of the sub-scores corresponding to the output dimensional features to obtain the total score of the patient's intestinal cleanliness;

[0026] The intestinal cleanliness level corresponding to the total score is determined according to a preset score and level mapping table.

[0027] Another embodiment of the present application provides a system for identifying the intestinal cleanliness level of a patient undergoing colonoscopy, the system comprising:

[0028] A collection module, used to collect a set of waste images from multiple angles of intestinal excretion of a patient who is about to undergo colonoscopy;

[0029] An extraction module, for extracting, for each dirt image in the dirt image set, a feature vector of each dimensional feature related to the intestinal cleanliness degree in the dirt image;

[0030] A combination module is used to combine all feature vectors extracted from each dimension feature to obtain a combined feature vector;

[0031] The evaluation module is used to evaluate the intestinal cleanliness level of the patient based on the combined feature vector corresponding to each dimensional feature and using a machine learning algorithm.

[0032] Yet another embodiment of the present application provides a storage medium, wherein the storage medium stores a computer program, wherein the computer program is configured to execute any of the above methods when running.

[0033] Yet another embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute any of the methods described above.

[0034] Compared with the prior art, the present invention provides a method for identifying the intestinal cleanliness level of patients undergoing colonoscopy, which collects a set of multi-angle dirt images of intestinal excretion of patients who are about to undergo colonoscopy; for each dirt image in the dirt image set, extracts the feature vector of each dimensional feature related to the intestinal cleanliness level in the dirt image; for each dimensional feature, combines all the extracted feature vectors of the dimensional feature to obtain a combined feature vector; based on the combined feature vector corresponding to each dimensional feature, utilizes a machine learning algorithm to evaluate the intestinal cleanliness level of the patient, thereby being able to comprehensively consider multidimensional features and make full use of advanced machine learning technology to achieve objective evaluation of the intestinal cleanliness level of patients undergoing colonoscopy, thereby enhancing the accuracy and reliability of the evaluation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 A hardware structure block diagram of a computer terminal for a method for identifying the intestinal cleanliness level of a colonoscopy patient provided by an embodiment of the present invention;

[0036] Figure 2 A schematic flow chart of a method for identifying the intestinal cleanliness level of a colonoscopy patient provided by an embodiment of the present invention;

[0037] Figure 3 A schematic diagram of the structure of a system for identifying the intestinal cleanliness level of colonoscopy patients provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, but should not be construed as limiting the present invention.

[0039] The embodiment of the present invention first provides a method for identifying the intestinal cleanliness level of a colonoscopy patient. The method can be applied to electronic devices, such as computer terminals, specifically ordinary computers, etc.

[0040] The following describes it in detail by taking running on a computer terminal as an example. Figure 1 The hardware structure block diagram of a computer terminal for a method for identifying the intestinal cleanliness level of a colonoscopy patient provided by an embodiment of the present invention. Figure 1 As shown, the computer device includes a processor, a memory, and a network interface connected via a system bus, wherein the memory may include a non-volatile storage medium and an internal memory.

[0041] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any method for identifying the intestinal cleanliness level of a colonoscopy patient.

[0042] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.

[0043] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any method for identifying the intestinal cleanliness level of a colonoscopy patient.

[0044] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will appreciate that Figure 1 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0045] It should be understood that the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0046] See also Figure 2 , an embodiment of the present invention provides a method for identifying the intestinal cleanliness level of a colonoscopy patient, which may include the following steps:

[0047] S201, collecting a set of waste images from multiple angles of intestinal excretion of a patient who is to undergo colonoscopy;

[0048] Before a colonoscopy, the cleanliness of the intestine is directly related to the accuracy of the test results, so collecting a multi-angle set of waste images from the intestinal excretion of patients who are about to undergo colonoscopy is the first step of this method. This process aims to comprehensively record the characteristics of the patient's excrement through high-quality image data for subsequent accurate cleanliness level assessment. Specifically, patients need to use a mobile phone application (APP) to take real-time photos of the waste after intestinal excretion. The APP will guide patients to take photos at different angles (such as front, side, and top view), and provide real-time feedback at each shooting stage to ensure data quality. In this way, the collected images will provide comprehensive basic data for subsequent image processing and feature extraction, ensuring the accuracy and reliability of the evaluation results.

[0049] The significance of collecting waste images from multiple angles is that it greatly improves the comprehensiveness and accuracy of intestinal cleanliness assessment. The traditional single-view or single-shot method may miss important features of intestinal waste, resulting in inaccurate assessment results. By shooting from multiple angles, the morphology, distribution, and properties of the waste can be fully reflected, thereby achieving a more comprehensive analysis. In addition, this assessment method can effectively reduce the workload of medical staff, enable patients to prepare their intestines independently, and improve patient participation and self-management capabilities. At the same time, timely obtaining feedback on the cleaning assessment can help patients understand their own preparation and ensure that qualified cleaning standards are met before colonoscopy, thereby improving the success rate and safety of the examination.

[0050] In the specific implementation process, patients can collect intestinal waste images through patient terminal applications or APPs. For example, after defecating in their own bathroom, the patient opens the APP and selects the "Photograph Intestinal Waste" function. The APP will provide detailed instructions to remind patients to shoot under appropriate lighting conditions to ensure that the waste image is clear and visible. For example, patients can use the flash of the mobile phone or adjust the ambient light source to eliminate shadows and increase contrast. In order to obtain the best angle, the APP can provide a schematic diagram to guide patients on how to position the mobile phone, such as shooting from multiple angles such as directly above and from the side.

[0051] After completing the photo shoot, the patient can select the appropriate image to upload. During the upload process, the APP will perform simple pre-processing on the image, such as compression and format conversion, to ensure that the uploaded file can be quickly transmitted to the background. Once the image upload is completed, the background system will automatically receive these multi-angle dirt images and apply image recognition and processing algorithms for analysis. These algorithms will extract key features in the image, such as solid form, liquid state and color features, to provide basic data for subsequent intestinal cleanliness level assessment.

[0052] For example, if a patient takes multiple images during the filming process, one of which clearly shows the boundaries of solid dirt and another shows the distribution of liquids, the backend system can analyze these images to comprehensively evaluate the patient's intestinal cleanliness and, after the evaluation is completed, directly feedback the results and suggestions to the patient's APP in the form of a report. This simplified process not only improves the efficiency of information transmission, but also allows patients to fully understand their own health preparation status, thereby increasing the overall success rate and safety of colonoscopy.

[0053] S202, for each dirt image in the dirt image set, extracting a feature vector of each dimensional feature related to the intestinal cleanliness degree in the dirt image;

[0054] For each dirt image in the dirt image set, extracting the feature vector of each dimensional feature related to the intestinal cleanliness level in the dirt image is a key step in evaluating the intestinal cleanliness level of the patient. This step quantifies the cleanliness of the intestine by analyzing and extracting different features of solid and liquid dirt in the image. Specifically, feature extraction involves a variety of image processing techniques, including morphological operations and edge detection, which are used to emphasize the boundaries of solid objects. In addition, the analysis of the liquid part involves techniques such as Gaussian filtering, threshold segmentation and optical flow, so as to capture the dynamic characteristics and geometric properties of the liquid. The extracted feature vector will reflect information such as the morphology of the dirt, the liquid state and its color distribution, forming a feature set containing multiple dimensional information, which provides the necessary data basis for the subsequent cleanliness assessment.

[0055] The implementation of this step is of great significance to ensure the accuracy and comprehensiveness of the intestinal cleanliness assessment. By extracting multi-dimensional features from each dirt image, we can have a deeper understanding of the nature of the intestinal contents, thereby providing more detailed data support for the evaluation. By quantifying the morphology, movement characteristics, and color distribution of solids and liquids, it will help medical personnel accurately judge the cleanliness of the intestine, thereby improving the success rate of colonoscopy and reducing repeated examinations due to insufficient intestinal preparation. Overall, this step can not only improve the scientific nature of the evaluation, but also provide patients with clear self-management guidance to ensure more accurate intestinal preparation.

[0056] Specifically, the various dimensional features may include: solid form features, liquid form features and color distribution features.

[0057] Morphological operations may be applied to the dirt image to remove noise and enhance the boundaries of solid objects, and edge detection algorithms may be used to identify and enhance the outlines of solid parts;

[0058] This step processes the dirt image through morphological operations to eliminate noise and emphasize the boundaries of solid objects. Morphological operations usually include techniques such as dilation and erosion, which can remove small interferences in the image and clearly identify the shape of the solid. In addition, the edge detection algorithm (such as the Canny algorithm) is combined to further strengthen the boundary, making the outline of the solid part more obvious, providing a high-quality image foundation for subsequent feature extraction.

[0059] Accurately identifying solid boundaries not only improves the reliability of subsequent feature extraction, but also ensures a more accurate judgment of the degree of intestinal cleanliness during data analysis. If the boundaries are blurred, the extracted features may be distorted, affecting the overall evaluation results. Therefore, this step provides key preliminary support for the entire intestinal cleanliness evaluation.

[0060] During the implementation process, the system first performs Gaussian filtering on the dirt image to smooth it and reduce the random noise of the image. Subsequently, morphological operations are performed, firstly corrosion processing is performed to remove small noise, and then the dilation method is applied to restore the boundaries of solid objects. Next, the Canny algorithm is used for edge detection to further improve the clarity of the solid part contour. For example, if the solid dirt in the image has blurred edges due to noise, after the above processing, the solid contour will become clear and discernible, which provides a solid foundation for subsequent feature extraction.

[0061] Extracting a shape descriptor of a solid in the dirt image to quantify the geometric properties of the solid morphology, and converting the shape descriptor into a solid feature vector corresponding to the solid morphological feature;

[0062] During this process, the system will extract the geometric features of the solid object, including shape descriptors such as perimeter, area, center of mass, etc. These features help to quantify the morphology of the solid. The extracted shape descriptors can be converted into feature vectors to form solid feature data for subsequent analysis.

[0063] By quantifying the geometric properties of solid matter, a more precise reference to intestinal cleanliness can be provided. These shape descriptors not only facilitate detailed analysis of solid matter, but also help medical professionals better judge whether a patient has achieved acceptable cleanliness standards.

[0064] In the specific implementation, it is first necessary to extract the contour of the dirt image after edge detection to obtain the boundary information of the solid. Then, the perimeter and area of ​​the solid contour, as well as other geometric features, are calculated using image processing algorithms. Specifically, the perimeter can be calculated by traversing the pixels on the contour boundary, while the area is obtained by counting the number of pixels inside the contour.

[0065] The dirt image is preprocessed by using Gaussian filtering to remove noise, clarify the boundary between the liquid and other parts, and use threshold segmentation or K-means clustering to segment the image to identify the liquid area;

[0066] In this step, the system preprocesses the dirt image to clearly identify the liquid area. Gaussian filtering can remove background noise and improve image quality, and then threshold segmentation or K-means clustering technology can be used to identify and extract the liquid area. This process ensures that the characteristics of the liquid can be accurately extracted, providing a clear basis for subsequent analysis.

[0067] Clear identification of liquid areas is crucial for the assessment of intestinal cleanliness, as the state and characteristics of the liquid are directly related to the patient's intestinal health. By accurately extracting the liquid area, the effectiveness of subsequent feature analysis is ensured, which helps to improve the accuracy of the assessment.

[0068] In the specific implementation, Gaussian filtering is first applied to the dirt image to remove random noise in the image. Then, the image is binarized using the threshold segmentation method to clearly separate the liquid area from the solid part, or the K-means clustering method is used to cluster the liquid areas of similar colors in the image. For example, if the liquid area is obviously dark, after threshold segmentation or K-means clustering, the system will be able to clearly distinguish the liquid from the solid part, ensuring the accuracy and completeness of subsequent feature extraction.

[0069] The optical flow method is used to detect the movement of the liquid area, and the speed and direction features of each pixel are calculated. The speed and direction features are encoded into dynamic texture feature vectors to describe the dynamic characteristics of the liquid.

[0070] This step uses the optical flow method to analyze the motion characteristics of the liquid area in the image and obtain the speed and direction of each pixel moving in the image sequence. These dynamic features will be encoded into feature vectors to describe the dynamic characteristics of the liquid and help understand the flow of liquid in the intestine.

[0071] The extraction of dynamic features provides a detailed observation of liquid behavior. Properties such as fluidity and viscosity are important for assessing intestinal cleanliness. Accurate flow characteristics can help doctors determine the actual condition of the patient's intestines and make more scientific medical decisions.

[0072] In the specific implementation process, the system will use optical flow methods (such as Lucas-Kanade optical flow method) to analyze the liquid area. By comparing adjacent frames in the video sequence, the system will detect the displacement and movement direction of each pixel at different time points.

[0073] From the identified liquid area, the spectrum characteristics of the liquid contour are analyzed by Fourier transform to extract morphological features, and a morphological feature vector corresponding to the morphological features is encoded to describe the geometric and morphological changes of the liquid, wherein the morphological features include contour length and curvature changes;

[0074] This step uses Fourier transform to perform spectrum analysis on the liquid contour to extract its morphological features. This frequency domain analysis can capture the changing details of the liquid shape, such as contour length and curvature, so as to better describe the geometric properties of the liquid through these morphological features.

[0075] By extracting the spectral features of liquid morphology, we can have a more comprehensive understanding of the changes in liquid, which is extremely important for evaluating its impact on intestinal cleansing. Accurately acquiring these features will help improve the comprehensiveness and accuracy of the assessment and promote more effective medical management.

[0076] In the specific implementation process, the contour of the liquid area is first identified, and then the contour information is converted from the spatial domain to the frequency domain by applying Fourier transform. Next, the low-frequency and high-frequency components are analyzed to calculate the length and curvature changes of the contour. For example, if the spectrum analysis of a liquid area shows that the low-frequency component is strong, it indicates that the liquid flow is relatively stable, while the change of the high-frequency component may reflect the dynamic characteristics of its flow. The system will generate a morphological feature vector based on this data.

[0077] The flow characteristics of the liquid are simulated by the difference method and image registration, and the deformation of the liquid at different time points in the image sequence is calculated to extract the flow velocity distribution and viscosity characteristics of the fluid, and the flow velocity distribution and viscosity characteristics are encoded in numerical form as a fluid feature vector;

[0078] This step uses the difference method and image registration technology to analyze the flow characteristics of the liquid in the time series, thereby extracting fluid characteristics such as flow velocity distribution and viscosity. This analysis helps to deeply understand the movement state of the liquid in the intestine and its impact on intestinal cleansing.

[0079] Flow rate and viscosity are important parameters for evaluating intestinal cleanliness. Accurately obtaining these characteristics can help doctors more effectively judge the patient's intestinal preparation and ensure the success rate and safety of colonoscopy.

[0080] In the specific implementation, the system first performs image registration on the image sequence to ensure that the liquid area in each frame can accurately correspond. Then, the difference method is used to compare the morphological changes of the liquid between adjacent frames to calculate the flow rate and viscosity at each time point.

[0081] Combining the dynamic texture feature vector, the morphological feature vector, and the fluid feature vector into a liquid feature vector corresponding to the liquid morphological feature;

[0082] In this step, the system combines, for example, concatenates, the previously extracted multiple feature vectors to form a comprehensive liquid feature vector. This combination integrates information from multiple dimensions and provides comprehensive data support for subsequent cleanliness assessment.

[0083] The comprehensive liquid feature vector can more comprehensively reflect the diversity and complexity of liquids, thus providing strong data support for the evaluation of intestinal cleanliness. Ensuring the richness and accuracy of the feature vector can greatly improve the reliability of the evaluation results.

[0084] Converting the dirt image from the RGB color space to the Lab color space, and generating a color histogram of the dirt image based on the Lab color space to quantify the distribution frequency of each color in the image;

[0085] This step converts the dirt image in RGB color space to Lab color space through color space conversion. Lab color space can better reflect the human eye's perception of color. Therefore, by generating a color histogram in Lab color space, the distribution frequency of each color in the image can be accurately quantified. The color histogram shows the frequency of occurrence of each color in the image, laying the foundation for subsequent color feature extraction.

[0086] This process is crucial to understanding the color distribution of intestinal waste, because color characteristics are often closely related to the degree of intestinal cleanliness. By quantifying the distribution of different colors, it can provide data support for the assessment of intestinal cleanliness and help medical staff better judge the patient's intestinal preparation status.

[0087] In the specific implementation process, the system first converts the dirt image from RGB color space to Lab color space. This conversion process is usually achieved by applying a specific color conversion formula to map the RGB value of each pixel to the Lab value. After that, based on the converted Lab image, the system will calculate the number of occurrences of each color and generate a color histogram. For example, assuming that in the Lab image, the frequency of occurrence of color A is 200 times, the frequency of color B is 150 times, and the frequency of color C is 100 times, then these values ​​will be used as the basis of the histogram to reflect the distribution characteristics of each color in the image, which is convenient for subsequent analysis.

[0088] A color feature vector is constructed based on the values ​​in the color histogram to represent the proportion of different colors in the image.

[0089] In this step, the system will use the values ​​in the generated color histogram to construct a color feature vector representing the proportion of different colors in the image. This feature vector will provide quantitative data for subsequent analysis, so that the distribution of different colors in the image can be effectively quantified.

[0090] By establishing a color feature vector, the proportion of each color in the image can be presented in numerical form, providing richer feature information for the assessment of intestinal cleanliness. Color features are very important for understanding the nature and cleanliness of intestinal contents, and can help medical personnel analyze the patient's intestinal condition more scientifically.

[0091] In the specific implementation process, the system normalizes the frequency of each color in the color histogram to calculate the proportion of each color in the entire image. This is usually achieved by dividing the number of occurrences of each color by the total number of pixels. Assuming that the histogram shows that the frequency of color A is 200, the frequency of color B is 150, the frequency of color C is 100, and the total number of pixels is 450, the color feature vector will be constructed as [color A ratio = 200 / 450, color B ratio = 150 / 450, color C ratio = 100 / 450]. The final generated color feature vector will serve as an important part of the intestinal cleansing assessment and help the accuracy of the overall analysis.

[0092] S203, for each dimensional feature, combining all feature vectors extracted from the dimensional feature to obtain a combined feature vector;

[0093] In this step, the system combines multiple feature vectors extracted for each dimensional feature (such as solid morphology, liquid morphology, and color distribution) to form a comprehensive feature vector. Through this combination method, information from multiple dimensions can be integrated to provide comprehensive data support for subsequent cleanliness assessment. This process ensures that the interconnections between the features of each dimension are reflected in the feature vector, allowing the model to more effectively capture the contribution of different features to intestinal cleanliness during subsequent learning and analysis.

[0094] The formation of the combined feature vector is a key link in the entire evaluation system. It enables features of different dimensions to be analyzed in the same space, thus providing a richer information basis for subsequent machine learning algorithms. This comprehensive data not only improves the accuracy and reliability of the evaluation, but also enhances the generalization ability of the model, helping doctors to more scientifically judge the patient's intestinal cleanliness status.

[0095] Specifically, for each dimensional feature, all feature vectors extracted from the dimensional feature may be weighted summed to obtain a combined feature vector.

[0096] In this step, for each dimensional feature, a more comprehensive combined feature vector is obtained by weighted summing up all the extracted feature vectors. The feature vectors of each dimensional feature may have different importance in expression, so the weighted processing can dynamically adjust the weight of each feature in the combined feature vector according to its relative importance in the actual evaluation. This method not only enhances the expressiveness of the features, but also enables the entire system to better adapt to the conditions of different patients, making the evaluation results more accurate and reliable.

[0097] The weighted summation method ensures that the importance of each dimension feature in the comprehensive analysis can be reasonably reflected, making the entire intestinal cleanliness level assessment process more targeted and accurate. In this way, the system can better capture the correlation between features and provide richer feature information for subsequent machine learning model training. This process directly affects the quality of intestinal cleanliness assessment and can effectively improve the accuracy of clinical decision-making.

[0098] In actual operation, for each dimensional feature, the system will first collect all the feature vectors extracted under this feature. For example, for solid morphological features, there may be three feature vectors: perimeter, area, and shape descriptor. Then, a weight value is assigned to each feature vector. These weights can be obtained through expert experience, statistical analysis, or training of machine learning models. Next, the system multiplies each feature vector by its corresponding weight and calculates the sum of these weighted feature vectors to form a combined feature vector.

[0099] S204, based on the combined feature vector corresponding to each dimensional feature, using a machine learning algorithm, evaluate the intestinal cleanliness level of the patient.

[0100] In this step, the integrated feature vector is used as input to evaluate the patient's intestinal cleanliness level using a machine learning algorithm. Machine learning algorithms, such as support vector machines (SVM), random forests, or deep learning models, can automatically identify complex relationships between input features and expected outputs based on training data. In this step, the feature vector provides multidimensional information about the degree of intestinal cleanliness, allowing the algorithm to accurately evaluate and classify the patient's intestinal status.

[0101] The implementation of this step can utilize the powerful processing capabilities of machine learning to efficiently and accurately evaluate the patient's intestinal cleanliness level. This can not only improve the scientific nature of clinical decision-making, but also provide doctors with reliable data support to help them develop more personalized strategies before colonoscopy, thereby improving the success rate of the examination and the safety of patients.

[0102] Specifically, the combined feature vector corresponding to each dimensional feature can be input into the corresponding pre-trained intestinal cleanliness assessment sub-model based on machine learning, and the sub-score of the intestinal cleanliness degree corresponding to each dimensional feature can be output, wherein one dimensional feature corresponds to training one intestinal cleanliness assessment sub-model;

[0103] This step aims to input the combined feature vector generated for each dimensional feature (such as solid form, liquid form, and color distribution) into a specific machine learning model to evaluate the impact of this feature on the degree of intestinal cleanliness. Each dimensional feature has an independent model, which is built from historical data and relevant expert knowledge, and can relatively accurately evaluate the contribution of each feature to the intestinal cleanliness status, and ultimately provide support for medical decision-making.

[0104] By inputting the combined feature vector into an independent evaluation sub-model, fine-grained analysis of different dimensional features can be achieved. This method can help identify the specific situation of the patient's intestinal cleansing, thereby providing a basis for subsequent clinical intervention measures. In theory, the characteristics exhibited by different dimensional features can help the medical team better understand the patient's intestinal status, optimize intestinal preparation strategies, and improve the effectiveness and safety of colonoscopy.

[0105] During the implementation process, it is first necessary to prepare a pre-trained evaluation sub-model for each dimensional feature (such as solid morphology, liquid morphology, and color distribution). These models are usually trained using machine learning algorithms, such as random forests, support vector machines, or deep learning models, to learn the relationship between different features and intestinal cleanliness based on historical annotated data. When training each model, the input is a feature vector and the output is a score of intestinal cleanliness.

[0106] During the actual evaluation process, the system will input the combined feature vector into the corresponding sub-model. For example, for the solid morphology combined feature vector, assuming that it contains features such as perimeter, area, and shape descriptors, after input, the model may return a score of 8 points, indicating that this dimension feature indicates a good level of cleanliness. At the same time, the liquid morphology features and color distribution features will also be evaluated through their respective models, returning 7 points and 6 points respectively. In this way, the output sub-score of each feature model will provide data support for subsequent summary evaluations, further enabling doctors to more effectively judge the patient's overall cleanliness status.

[0107] Perform a weighted summation of the sub-scores corresponding to the output dimensional features to obtain the total score of the patient's intestinal cleanliness;

[0108] In this step, the system will weight the sub-scores obtained from each dimension feature model to generate a comprehensive total score that represents the patient's intestinal cleanliness. Each sub-score will be given a different weight based on its importance in the overall assessment, in order to ensure that the influence of each feature in the final score is reasonably reflected.

[0109] The weighted summation process allows the characteristics of each dimension to be comprehensively considered to form an overall score. This method not only improves the accuracy of the assessment, but also provides doctors with a clearer basis for judgment, helping them to make scientific decisions on how to improve the patient's intestinal cleansing status. By generating a total score, the medical team can quickly draw a panoramic view of the intestinal preparation status and effectively guide subsequent clinical interventions.

[0110] In specific implementations, the system first assigns a weight value to the sub-score of each dimensional feature. The setting of the weight value can be determined by expert opinion, statistical analysis, or model-based training results. For example, assuming that the solid morphology feature is considered to be a key factor in evaluating intestinal cleanliness, a higher weight (such as 0.5) can be given, while the weights of the liquid morphology and color distribution features are set to 0.3 and 0.2, respectively.

[0111] Next, the system multiplies all sub-scores by the corresponding weights and sums them up to get the total score. For example, if the solid form feature returns 8 points, the liquid form feature returns 7 points, and the color distribution feature returns 6 points, the calculation process of the total score is: total score = (8*0.5) + (7*0.3) + (6*0.2) = 4 + 2.1 + 1.2 = 7.3 points. The total score obtained will provide an important basis for the subsequent cleaning level judgment.

[0112] The intestinal cleanliness level corresponding to the total score is determined according to a preset score and level mapping table.

[0113] The purpose of this step is to convert the total score calculated previously into the patient's intestinal cleanliness level through a preset score and level mapping table. This mapping table is usually established based on a large amount of historical data analysis, which can intuitively correspond the total score to the actual cleanliness status, ensuring the rationality and accuracy of the evaluation results.

[0114] By combining the total score with the level mapping, the system can convert complex numerical scores into easy-to-understand intestinal cleanliness levels, which provides a clear reference for medical staff and helps them make clinical decisions quickly. This method can effectively improve the operating efficiency of doctors and enable patients to get a more accurate status assessment before colonoscopy.

[0115] When implementing this step, the system needs to prepare a score and grade mapping table in advance. For example, the following mapping relationship may be set: 0-5 points correspond to "poor intestinal cleaning", 6-7 points correspond to "average intestinal cleaning", and 8-10 points correspond to "good intestinal cleaning". In specific operations, the system will compare the previously calculated total score with the mapping table.

[0116] Assuming that the total score obtained in the above steps is 7.3 points, the system compares it with the mapping table and finds that the score falls into the range of "average intestinal cleanliness", so the patient's intestinal cleanliness status is marked as "average". This process not only improves the efficiency of the evaluation, but also provides doctors with a clear description of the patient's condition, enabling them to better plan subsequent examinations and treatment plans. Through such a mapping process, information transmission in the medical process becomes more intuitive and effective.

[0117] It can be seen that a set of multi-angle waste images of intestinal excretion of patients who are about to undergo colonoscopy is collected; for each waste image in the waste image set, the feature vector of each dimensional feature related to the intestinal cleanliness degree in the waste image is extracted; for each dimensional feature, all the extracted feature vectors of the dimensional feature are combined to obtain a combined feature vector; based on the combined feature vector corresponding to each dimensional feature, a machine learning algorithm is used to evaluate the intestinal cleanliness level of the patient, so that by comprehensively considering multi-dimensional features and making full use of advanced machine learning technology, an objective evaluation of the intestinal cleanliness level of colonoscopy patients can be achieved, thereby enhancing the accuracy and reliability of the evaluation results.

[0118] Another embodiment of the present invention provides a system for identifying the intestinal cleanliness level of a colonoscopy patient, see Figure 3 , the system may include:

[0119] The acquisition module 301 is used to acquire a set of waste images from multiple angles of intestinal excretion of a patient who is to undergo colonoscopy;

[0120] An extraction module 302 is used to extract, for each dirt image in the dirt image set, a feature vector of each dimensional feature related to the intestinal cleanliness degree in the dirt image;

[0121] A combining module 303 is used to combine all feature vectors extracted from each dimensional feature to obtain a combined feature vector;

[0122] The evaluation module 304 is used to evaluate the intestinal cleanliness level of the patient based on the combined feature vector corresponding to each dimensional feature and using a machine learning algorithm.

[0123] It can be seen that a set of multi-angle waste images of intestinal excretion of patients who are about to undergo colonoscopy is collected; for each waste image in the waste image set, the feature vector of each dimensional feature related to the intestinal cleanliness degree in the waste image is extracted; for each dimensional feature, all the extracted feature vectors of the dimensional feature are combined to obtain a combined feature vector; based on the combined feature vector corresponding to each dimensional feature, a machine learning algorithm is used to evaluate the intestinal cleanliness level of the patient, so that by comprehensively considering multi-dimensional features and making full use of advanced machine learning technology, an objective evaluation of the intestinal cleanliness level of colonoscopy patients can be achieved, thereby enhancing the accuracy and reliability of the evaluation results.

[0124] An embodiment of the present invention further provides a storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.

[0125] Specifically, in this embodiment, the above storage medium may be configured to store a computer program for performing the following steps:

[0126] S201, collecting a set of waste images from multiple angles of intestinal excretion of a patient who is to undergo colonoscopy;

[0127] S202, for each dirt image in the dirt image set, extracting a feature vector of each dimensional feature related to the intestinal cleanliness degree in the dirt image;

[0128] S203, for each dimensional feature, combining all feature vectors extracted from the dimensional feature to obtain a combined feature vector;

[0129] S204, based on the combined feature vector corresponding to each dimensional feature, using a machine learning algorithm, evaluate the intestinal cleanliness level of the patient.

[0130] It can be seen that a set of multi-angle waste images of intestinal excretion of patients who are about to undergo colonoscopy is collected; for each waste image in the waste image set, the feature vector of each dimensional feature related to the intestinal cleanliness degree in the waste image is extracted; for each dimensional feature, all the extracted feature vectors of the dimensional feature are combined to obtain a combined feature vector; based on the combined feature vector corresponding to each dimensional feature, a machine learning algorithm is used to evaluate the intestinal cleanliness level of the patient, so that by comprehensively considering multi-dimensional features and making full use of advanced machine learning technology, an objective evaluation of the intestinal cleanliness level of colonoscopy patients can be achieved, thereby enhancing the accuracy and reliability of the evaluation results.

[0131] An embodiment of the present invention further provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0132] Specifically, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0133] Specifically, in this embodiment, the processor may be configured to perform the following steps through a computer program:

[0134] S201, collecting a set of waste images from multiple angles of intestinal excretion of a patient who is to undergo colonoscopy;

[0135] S202, for each dirt image in the dirt image set, extracting a feature vector of each dimensional feature related to the intestinal cleanliness degree in the dirt image;

[0136] S203, for each dimensional feature, combining all feature vectors extracted from the dimensional feature to obtain a combined feature vector;

[0137] S204, based on the combined feature vector corresponding to each dimensional feature, using a machine learning algorithm, evaluate the intestinal cleanliness level of the patient.

[0138] It can be seen that a set of multi-angle waste images of intestinal excretion of patients who are about to undergo colonoscopy is collected; for each waste image in the waste image set, the feature vector of each dimensional feature related to the intestinal cleanliness degree in the waste image is extracted; for each dimensional feature, all the extracted feature vectors of the dimensional feature are combined to obtain a combined feature vector; based on the combined feature vector corresponding to each dimensional feature, a machine learning algorithm is used to evaluate the intestinal cleanliness level of the patient, so that by comprehensively considering multi-dimensional features and making full use of advanced machine learning technology, an objective evaluation of the intestinal cleanliness level of colonoscopy patients can be achieved, thereby enhancing the accuracy and reliability of the evaluation results.

[0139] The above describes in detail the structure, features and effects of the present invention based on the embodiments shown in the drawings. The above is only a preferred embodiment of the present invention, but the present invention is not limited to the scope of implementation shown in the drawings. Any changes made according to the concept of the present invention, or modifications to equivalent embodiments with equivalent changes, which still do not exceed the spirit covered by the description and drawings, should be within the protection scope of the present invention.

Claims

1. A method for identifying the intestinal cleanliness level of patients undergoing colonoscopy, characterized in that: The method comprises: Collecting a set of waste images from multiple angles of intestinal excretion from patients who are about to undergo colonoscopy; For each dirt image in the dirt image set, extracting a feature vector of each dimensional feature related to the intestinal cleanliness degree in the dirt image; For each dimensional feature, all feature vectors extracted from the dimensional feature are combined to obtain a combined feature vector; Based on the combined feature vector corresponding to each dimensional feature, a machine learning algorithm is used to evaluate the patient's intestinal cleanliness level.

2. The method according to claim 1, characterized in that The various dimensional features include: solid form features, liquid form features and color distribution features; The step of extracting the feature vector of each dimensional feature related to the intestinal cleanliness degree in each dirt image in the dirt image set includes: applying morphological operations to the dirt image to remove noise and enhance the boundaries of solid objects, and using edge detection algorithms to identify and enhance the outlines of solid parts; Extracting a shape descriptor of a solid in the dirt image to quantify the geometric properties of the solid morphology, and converting the shape descriptor into a solid feature vector corresponding to the solid morphological feature; The dirt image is preprocessed by using Gaussian filtering to remove noise, clarify the boundary between the liquid and other parts, and use threshold segmentation or K-means clustering to segment the image to identify the liquid area; The optical flow method is used to detect the movement of the liquid area, and the speed and direction features of each pixel are calculated. The speed and direction features are encoded into dynamic texture feature vectors to describe the dynamic characteristics of the liquid. From the identified liquid area, the spectrum characteristics of the liquid contour are analyzed by Fourier transform to extract morphological features, and a morphological feature vector corresponding to the morphological features is encoded to describe the geometric and morphological changes of the liquid, wherein the morphological features include contour length and curvature changes; The flow characteristics of the liquid are simulated by the difference method and image registration, and the deformation of the liquid at different time points in the image sequence is calculated to extract the flow velocity distribution and viscosity characteristics of the fluid, and the flow velocity distribution and viscosity characteristics are encoded in numerical form as a fluid feature vector; Combining the dynamic texture feature vector, the morphological feature vector, and the fluid feature vector into a liquid feature vector corresponding to the liquid morphological feature; Converting the dirt image from the RGB color space to the Lab color space, and generating a color histogram of the dirt image based on the Lab color space to quantify the distribution frequency of each color in the image; A color feature vector is constructed based on the values ​​in the color histogram to represent the proportion of different colors in the image.

3. The method according to claim 2, characterized in that For each dimensional feature, all feature vectors of the dimensional feature extracted are combined to obtain a combined feature vector, including: For each dimensional feature, all feature vectors extracted for the dimensional feature are weighted summed to obtain a combined feature vector.

4. The method according to claim 3, characterized in that The method of evaluating the intestinal cleanliness level of the patient based on the combined feature vector corresponding to each dimensional feature and using a machine learning algorithm includes: The combined feature vector corresponding to each dimensional feature is input into the corresponding pre-trained intestinal cleanliness assessment sub-model based on machine learning, and the sub-score of the intestinal cleanliness degree corresponding to each dimensional feature is output, wherein one dimensional feature corresponds to the training of one intestinal cleanliness assessment sub-model; Perform a weighted summation of the sub-scores corresponding to the output dimensional features to obtain the total score of the patient's intestinal cleanliness; The intestinal cleanliness level corresponding to the total score is determined according to a preset score and level mapping table.

5. A system for identifying the intestinal cleanliness level of patients undergoing colonoscopy, characterized in that: The system comprises: A collection module, used to collect a set of waste images from multiple angles of intestinal excretion of a patient who is about to undergo colonoscopy; An extraction module, for extracting, for each dirt image in the dirt image set, a feature vector of each dimensional feature related to the intestinal cleanliness degree in the dirt image; A combination module is used to combine all feature vectors extracted from each dimension feature to obtain a combined feature vector; The evaluation module is used to evaluate the intestinal cleanliness level of the patient based on the combined feature vector corresponding to each dimensional feature and using a machine learning algorithm.

6. The system according to claim 5, characterized in that The various dimensional features include: solid morphology features, liquid morphology features and color distribution features; the extraction module is specifically used to: applying morphological operations to the dirt image to remove noise and enhance the boundaries of solid objects, and using edge detection algorithms to identify and enhance the outlines of solid parts; Extracting a shape descriptor of a solid in the dirt image to quantify the geometric properties of the solid morphology, and converting the shape descriptor into a solid feature vector corresponding to the solid morphological feature; The dirt image is preprocessed by using Gaussian filtering to remove noise, clarify the boundary between the liquid and other parts, and use threshold segmentation or K-means clustering to segment the image to identify the liquid area; The optical flow method is used to detect the movement of the liquid area, and the speed and direction features of each pixel are calculated. The speed and direction features are encoded into dynamic texture feature vectors to describe the dynamic characteristics of the liquid. From the identified liquid area, the spectrum characteristics of the liquid contour are analyzed by Fourier transform to extract morphological features, and a morphological feature vector corresponding to the morphological features is encoded to describe the geometric and morphological changes of the liquid, wherein the morphological features include contour length and curvature changes; The flow characteristics of the liquid are simulated by the difference method and image registration, and the deformation of the liquid at different time points in the image sequence is calculated to extract the flow velocity distribution and viscosity characteristics of the fluid, and the flow velocity distribution and viscosity characteristics are encoded in numerical form as a fluid feature vector; Combining the dynamic texture feature vector, the morphological feature vector, and the fluid feature vector into a liquid feature vector corresponding to the liquid morphological feature; Converting the dirt image from the RGB color space to the Lab color space, and generating a color histogram of the dirt image based on the Lab color space to quantify the distribution frequency of each color in the image; A color feature vector is constructed based on the values ​​in the color histogram to represent the proportion of different colors in the image.

7. The system according to claim 6, characterized in that The combined module is specifically used for: For each dimensional feature, all feature vectors extracted for the dimensional feature are weighted summed to obtain a combined feature vector.

8. The system according to claim 7, characterized in that The evaluation module is specifically used for: The combined feature vector corresponding to each dimensional feature is input into the corresponding pre-trained intestinal cleanliness assessment sub-model based on machine learning, and the sub-score of the intestinal cleanliness degree corresponding to each dimensional feature is output, wherein one dimensional feature corresponds to the training of one intestinal cleanliness assessment sub-model; Perform a weighted summation of the sub-scores corresponding to the output dimensional features to obtain the total score of the patient's intestinal cleanliness; The intestinal cleanliness level corresponding to the total score is determined according to a preset score and level mapping table.

9. A storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program is configured to execute the method according to any one of claims 1 to 4 when executed.

10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to perform the method according to any one of claims 1 to 4.

Images