An ai-based bowel preparation quality assessment system

Abstract

The application discloses an AI-based intestinal preparation quality evaluation system, belonging to the technical field of enteroscopy, comprising: a collection and processing module configured to collect the excrement fluorescence image of a patient, pre-process and feature extract the same, and determine excrement feature data; a model establishment module configured to build an intestinal preparation quality evaluation model; a quality evaluation module configured to perform feature recognition and analysis on the excrement feature data, perform quality evaluation on the intestinal preparation of the patient, and determine an intestinal preparation quality evaluation result; and an output management module configured to visually output the intestinal preparation quality evaluation report of the patient and perform subsequent enteroscopy management on the patient. The application solves the problems of existing intestinal preparation quality evaluation, such as certain inaccuracy and subjectivity, and low intestinal preparation qualification rate of up to 70%. The application can provide more objective and efficient intestinal preparation quality evaluation results, thereby improving the accuracy and efficiency of enteroscopy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of enteroscopy, in particular to an AI-based intestinal preparation quality evaluation system. BACKGROUND

[0002] Intestinal preparation before enteroscopy is a key step to ensure smooth examination and improve diagnostic accuracy. Poor intestinal preparation can lead to the following situations: 1) unclear vision, easy to miss diagnosis, literature shows that poor intestinal preparation has a miss diagnosis rate as high as 45.7%; because the feces and food residues in the intestinal tract will block the intestinal mucosa, affecting the observation of the subtle lesions of the intestinal wall by the doctor, such as polyps, ulcers, tumors, etc., especially small polyps or early cancer, delaying the treatment opportunity; 2) reducing examination efficiency and safety, poor intestinal preparation is not conducive to the smooth passage of enteroscopy, increasing the operation time and patient discomfort, increasing the incidence of complications, intestinal residues may cause difficulty in advancing the enteroscope, increasing the risk of intestinal wall injury; 3) increase the treatment cost, if polyps or bleeding points are found, the doctor can directly remove or stop bleeding during the examination to avoid secondary preparation and operation, but insufficient preparation may require rescheduling the examination, increasing the time and economic cost.

[0003] Therefore, the importance of intestinal preparation is self-evident, but the cleanliness of intestinal preparation is currently judged by medical staff subjectively, which has certain inaccuracy and subjectivity, and patients who do intestinal preparation at home often have insufficient intestinal preparation when they do enteroscopy, and the intestinal preparation qualification rate is as low as 70%. SUMMARY

[0004] The purpose of the present application is to provide an AI-based intestinal preparation quality evaluation system, which can provide more objective and efficient intestinal preparation quality evaluation results, thereby improving the accuracy and efficiency of enteroscopy and solving the problems raised in the background art.

[0005] To achieve the above purpose, the present application provides the following technical solutions:

[0006] An AI-based intestinal preparation quality evaluation system, comprising:

[0007] A collection and processing module configured to collect excrement fluorescence images of a patient and pre-process and feature extract the same to determine excrement feature data;

[0008] A model establishment module configured to construct an intestinal preparation quality evaluation model according to intestinal preparation quality evaluation requirements;

[0009] A quality evaluation module configured to perform feature recognition and analysis on the excrement feature data according to the intestinal preparation quality evaluation model, perform quality evaluation on the intestinal preparation of the patient, and determine an intestinal preparation quality evaluation result;

[0010] An output management module is configured to visualize an output of a quality assessment report of a bowel preparation of a patient and to manage a follow-up colonoscopy of the patient.

[0011] Preferably, the fluorescent image of the patient's excrement is collected, including:

[0012] The excrement sample of the patient is collected in a sterile container, and the excrement sample of the patient is diluted or centrifuged to remove impurities in the excrement sample and concentrate the target component;

[0013] Based on the fluorescence imaging technology, a fluorescent dye is selected according to the characteristics of the target molecule in the excrement sample, the fluorescent dye is mixed with the excrement sample, and the two are fully contacted;

[0014] The reaction conditions, including the pH value, temperature and time of the reaction, are selected according to the target molecule in the excrement sample, and the excrement sample is washed with a buffer to remove unbound dye in the excrement sample;

[0015] The labeled sample is purified by centrifugation or filtration, and the labeled sample is placed in a fluorescence microscope, the excitation wavelength and filter combination are selected, and the fluorescent image of the patient's excrement is obtained.

[0016] Preferably, the fluorescent image of the patient's excrement is preprocessed, including:

[0017] Based on the filter, the fluorescent image of the patient's excrement is denoised to remove noise in the fluorescent image of the excrement and reduce background interference in the fluorescent image of the excrement;

[0018] Based on histogram equalization, the contrast of the fluorescent image of the patient's excrement is enhanced to enhance the details in the fluorescent image of the excrement and highlight the fluorescent signal in the target region of the fluorescent image of the excrement;

[0019] Based on the threshold segmentation method, the fluorescent image of the patient's excrement is segmented to separate the background and target region in the fluorescent image of the excrement and eliminate non-specific fluorescent interference.

[0020] Preferably, the threshold value for image segmentation processing is set, including:

[0021] The contrast value before the contrast enhancement processing and the contrast value after the contrast enhancement processing are extracted;

[0022] The contrast enhancement amplitude is obtained according to the contrast value before the contrast enhancement processing and the contrast value after the contrast enhancement processing;

[0023] The contrast enhancement ratio is obtained by ratio processing of the contrast enhancement amplitude and the contrast value before the contrast enhancement processing;

[0024] comparing the contrast enhancement ratio with a preset contrast enhancement ratio threshold value;

[0025] when the contrast enhancement ratio does not exceed the preset contrast enhancement ratio threshold value, preset initial threshold parameters for image segmentation processing are called, and the excrement fluorescent image of the patient is subjected to image segmentation processing based on threshold segmentation method using the preset initial threshold parameters for image segmentation processing;

[0026] when the contrast enhancement ratio exceeds the preset contrast enhancement ratio threshold value, the initial threshold parameters are adjusted using the gray values of the excrement fluorescent image before and after contrast enhancement;

[0027] the excrement fluorescent image of the patient is subjected to image segmentation processing based on threshold segmentation method using the adjusted threshold parameters.

[0028] Preferably, when the contrast enhancement ratio exceeds the preset contrast enhancement ratio threshold value, the initial threshold parameters are adjusted using the gray values of the excrement fluorescent image before and after contrast enhancement, comprising:

[0029] when the contrast enhancement ratio exceeds the preset contrast enhancement ratio threshold value, the excrement fluorescent image before contrast enhancement processing is called;

[0030] the excrement fluorescent image before contrast enhancement processing is subjected to gray processing to obtain a first gray image;

[0031] the excrement fluorescent image after contrast enhancement processing is called;

[0032] the excrement fluorescent image after contrast enhancement processing is subjected to gray processing to obtain a second gray image;

[0033] each pixel point in the first gray image is extracted;

[0034] each pixel point in the second gray image is extracted;

[0035] for the same position pixel point in the first gray image and the second gray, the gray difference of the same position pixel point is obtained;

[0036] the gray difference of the same position pixel point is subjected to ratio processing with the gray value of the position pixel point in the first gray image to obtain the gray change rate of the same position pixel point;

[0037] the initial threshold parameters are adjusted using the gray change rate and the contrast enhancement ratio.

[0038] Preferably, the feature extraction of the excrement fluorescent image of the patient comprises:

[0039] Based on the convolutional neural network CNN, the excrement fluorescence image of the patient is extracted, the semantic features valuable for the evaluation of the quality of intestinal preparation are extracted from the excrement fluorescence image, and the extracted features are processed by principal component analysis to determine the excrement feature data, including excrement composition, concentration, color, contrast and texture.

[0040] Preferably, the intestinal preparation quality evaluation model is constructed, including:

[0041] According to the intestinal preparation quality evaluation requirement based on AI, the excrement historical data is collected, and the collected excrement historical data is divided into a training set and a test set;

[0042] Based on the deep learning technology, the training set is used to train the deep learning model, so that the deep learning model learns the intestinal preparation quality evaluation behavior from the training set, which is used for quality evaluation of the intestinal preparation of the patient, judges the cleaning degree of the intestinal tract of the patient, and determines the intestinal preparation quality evaluation model based on deep learning;

[0043] The test set is used to test the intestinal preparation quality evaluation model based on deep learning, evaluate the performance of the intestinal preparation quality evaluation model based on deep learning, judge whether the intestinal preparation quality evaluation model based on deep learning can achieve the expected effect of quality evaluation of the intestinal preparation of the patient and judgment of the cleaning degree of the intestinal tract of the patient, and determine the model test evaluation result;

[0044] According to the model test evaluation result, the parameters of the intestinal preparation quality evaluation model based on deep learning are adjusted and optimized until the intestinal preparation quality evaluation model based on deep learning can achieve the expected effect of quality evaluation of the intestinal preparation of the patient and judgment of the cleaning degree of the intestinal tract of the patient, and then the optimal intestinal preparation quality evaluation model is determined.

[0045] Preferably, the quality of the intestinal preparation of the patient is evaluated, including:

[0046] The optimal intestinal preparation quality evaluation model is deployed, and the optimal intestinal preparation quality evaluation model is deployed in the actual intestinal preparation quality evaluation environment;

[0047] The excrement feature data is input into the intestinal preparation quality evaluation model, the excrement feature data is identified and analyzed according to the intestinal preparation quality evaluation model, the quality of the intestinal preparation of the patient is evaluated, and the cleaning degree of the intestinal tract of the patient is judged, and the intestinal preparation quality evaluation result is determined.

[0048] Preferably, the intestinal preparation quality evaluation report of the patient is visualized and output, and the patient is managed for subsequent colonoscopy, including:

[0049] According to the evaluation result of the intestinal preparation quality and in combination with the excrement fluorescence image of the patient, an intestinal preparation quality evaluation report of the patient is formed, and the report is displayed in a visual form to the doctor in real time, so as to enable the doctor to manage the patient in subsequent colonoscopy according to the intestinal preparation quality evaluation report of the patient.

[0050] Preferably, the cleaning degree of the intestinal tract of the patient is divided into four grades based on the intestinal preparation quality evaluation result, including:

[0051] The first grade, that is, the intestinal mucosa is not clearly displayed, and there is a large amount of fecal residue in the intestinal tract;

[0052] The second grade, that is, the intestinal mucosa is partially clearly displayed, and there is a moderate amount of fecal residue in the intestinal tract;

[0053] The third grade, that is, the intestinal mucosa is clearly displayed, and there is a small amount of fecal residue in the intestinal tract;

[0054] The fourth grade, that is, the intestinal mucosa is clearly displayed, and there is no fecal residue in the intestinal tract;

[0055] Among them, the higher the grade, the better the intestinal preparation, and the more conducive to the colonoscopy.

[0056] Compared with the prior art, the beneficial effects of the present application are:

[0057] The present application collects the excrement fluorescence image of the patient through fluorescence imaging technology, pre-processes and extracts features from the excrement fluorescence image of the patient, determines the excrement feature data, constructs an intestinal preparation quality evaluation model according to the intestinal preparation quality evaluation requirements based on AI, performs feature recognition and analysis on the excrement feature data according to the intestinal preparation quality evaluation model, performs quality evaluation on the intestinal preparation of the patient and judges the cleaning degree of the intestinal tract of the patient, determines the intestinal preparation quality evaluation result, and in combination with the excrement fluorescence image of the patient, forms an intestinal preparation quality evaluation report of the patient, and displays the report in a visual form to the doctor in real time, so as to enable the doctor to manage the patient in subsequent colonoscopy according to the intestinal preparation quality evaluation report of the patient, and provide more objective and efficient intestinal preparation quality evaluation results, thereby improving the accuracy and efficiency of colonoscopy. BRIEF DESCRIPTION OF DRAWINGS

[0058] Fig. 1 The module diagram of the intestinal preparation quality evaluation system based on AI of the present application;

[0059] Fig. 2 The flowchart of the intestinal preparation quality evaluation system based on AI of the present application. DETAILED DESCRIPTION

[0060] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0061] In order to solve the problem that the existing intestinal preparation cleanliness is subjectively judged by medical staff, there is a certain inaccuracy and subjectivity, and patients who do intestinal preparation at home often have insufficient intestinal preparation when doing colonoscopy, and the intestinal preparation qualification rate is as low as 70%. Please refer to Figs. 1-2 The embodiment provides the following technical solutions:

[0062] An intestinal preparation quality evaluation system based on AI, comprising: a collection and processing module, a model establishment module, a quality evaluation module, and an output management module.

[0063] Specifically, through the interaction between the collection and processing module, the model establishment module, the quality evaluation module, and the output management module, more objective and efficient intestinal preparation quality evaluation results can be provided, thereby improving the accuracy and efficiency of colonoscopy.

[0064] The collection and processing module is configured to collect a fluorescent image of excrement of a patient, and to pre-process and extract features of the fluorescent image to determine excrement feature data.

[0065] In the embodiment, the fluorescent image of the excrement of the patient is collected, including:

[0066] The excrement sample of the patient is collected in a sterile container, and the excrement sample of the patient is diluted or centrifuged to remove impurities in the excrement sample and concentrate target components.

[0067] According to the characteristics of the target molecules in the excrement sample, a fluorescent dye is selected, and the fluorescent dye is mixed with the excrement sample to make them fully contact.

[0068] According to the reaction conditions of the target molecules in the excrement sample, including the pH value, temperature and time of the reaction, and using a buffer to wash the excrement sample, the unbound dye in the excrement sample is removed.

[0069] The labeled sample is purified by centrifugation or filtration, and the labeled sample is placed in a fluorescence microscope, the excitation wavelength and filter combination are selected, and the fluorescent image of the excrement of the patient is obtained.

[0070] It should be noted that the sensitivity of the intestinal preparation quality evaluation model to subtle residues is improved by combining fluorescence imaging.

[0071] In the embodiment, the excrement fluorescence image of the patient is preprocessed, including:

[0072] The excrement fluorescence image of the patient is denoised based on a filter, to remove noise in the excrement fluorescence image and reduce background interference in the excrement fluorescence image;

[0073] The excrement fluorescence image of the patient is subjected to contrast enhancement processing based on histogram equalization, to enhance details in the excrement fluorescence image and highlight the fluorescence signal of the target region in the excrement fluorescence image;

[0074] The excrement fluorescence image of the patient is subjected to image segmentation processing based on threshold segmentation, to separate the background and the target region in the excrement fluorescence image and eliminate non-specific fluorescence interference.

[0075] Specifically, the threshold value for image segmentation processing is set, including:

[0076] The contrast value before the contrast enhancement processing and the contrast value after the contrast enhancement processing are extracted;

[0077] The contrast enhancement amplitude is obtained according to the contrast value before the contrast enhancement processing and the contrast value after the contrast enhancement processing;

[0078] The contrast enhancement ratio is obtained by ratio processing of the contrast enhancement amplitude and the contrast value before the contrast enhancement processing;

[0079] The contrast enhancement ratio is compared with a preset contrast enhancement ratio threshold value;

[0080] When the contrast enhancement ratio does not exceed the preset contrast enhancement ratio threshold value, the preset initial threshold value parameter for image segmentation processing is called, and the excrement fluorescence image of the patient is subjected to image segmentation processing based on threshold segmentation using the preset initial threshold value parameter for image segmentation processing;

[0081] When the contrast enhancement ratio exceeds the preset contrast enhancement ratio threshold value, the initial threshold value parameter is adjusted using the gray values of the excrement fluorescence images before and after the contrast enhancement;

[0082] The excrement fluorescence image of the patient is subjected to image segmentation processing based on threshold segmentation using the adjusted threshold value parameter.

[0083] The technical effects of the above technical solutions are: contrast enhancement usually stretches the gray scale range of the image, making the gray scale difference between the target region (such as the excrement fluorescent signal) and the background more obvious. At this time, the original threshold value may no longer be applicable to the new gray scale distribution. By analyzing the gray scale values before and after enhancement, the threshold value needs to be adjusted in the following way: calculate the gray scale histogram of the enhanced image, find the gray scale peak values of the target region and the background, and select the middle value as the new threshold value. According to the enhancement ratio and the gray scale change trend, adjust the initial threshold value in proportion. Recalculate the optimal threshold value in combination with the local gray scale characteristics of the enhanced image. At the same time, by comparing the values before and after contrast enhancement, calculate the contrast enhancement amplitude and ratio, and determine whether to adjust the threshold value accordingly. When the contrast enhancement ratio does not exceed the threshold value, use the initial threshold value, which is a reasonable value set through certain experience or preliminary research; when it exceeds the threshold value, adjust the threshold value according to the gray scale value. In this way, the threshold value can be selected flexibly according to the actual degree of image contrast enhancement. For excrement fluorescent images with different contrast enhancement effects, appropriate strategies can be adopted, either by using the initial threshold value or by adjusting the threshold value according to the gray scale value, so that it can adapt to various imaging conditions and image characteristics, enhancing the versatility and adaptability of the method in different scenarios. When the contrast enhancement ratio does not exceed the threshold value, the pre-set initial threshold value is used directly, avoiding unnecessary threshold adjustment calculation and improving processing efficiency; when the ratio exceeds the threshold value, the threshold value is adjusted based on the gray scale value, ensuring good segmentation effect when the image contrast changes greatly. This approach achieves a good balance between processing efficiency and segmentation effect, improving the overall performance of image processing.

[0084] Specifically, when the contrast enhancement ratio exceeds the pre-set contrast enhancement ratio threshold value, the initial threshold value parameter is adjusted using the gray scale values of the excrement fluorescent images before and after contrast enhancement, including:

[0085] When the contrast enhancement ratio exceeds the pre-set contrast enhancement ratio threshold value, the excrement fluorescent image before contrast enhancement processing is retrieved;

[0086] Performing gray scale processing on the excrement fluorescent image before contrast enhancement processing to obtain a first gray scale image;

[0087] Retrieving the excrement fluorescent image after contrast enhancement processing;

[0088] Performing gray scale processing on the excrement fluorescent image after contrast enhancement processing to obtain a second gray scale image;

[0089] Extracting each pixel point in the first gray scale image;

[0090] extracting each pixel point in the second gray scale image;

[0091] For the same position pixel point in the first gray scale image and the second gray scale, the gray scale difference of the same position pixel point is obtained;

[0092] The gray scale change rate of the same position pixel point is obtained by using the gray scale difference of the same position pixel point and the gray scale value of the position pixel point in the first gray scale image for ratio processing;

[0093] The initial threshold parameter is adjusted by using the gray scale change rate and the contrast enhancement ratio;

[0094] The adjusted threshold parameter is obtained by the following formula:

[0095]

[0096] Wherein, represents the adjusted threshold parameter; represents the initial threshold parameter; represents the number of pixel points contained in the excrement fluorescence image; represents the gray scale change rate corresponding to the i-th pixel point; represents the contrast enhancement ratio. Specifically, The difference between the two is the deviation of the gray scale change of a single pixel point relative to the contrast enhancement of the whole image, which is used to measure the difference between the local and the whole change. The difference between the two is the deviation of the gray scale change of a single pixel point relative to the contrast enhancement of the whole image, which is used to measure the difference between the local and the whole change. The difference between the two is the deviation of the gray scale change of a single pixel point relative to the contrast enhancement of the whole image, which is used to measure the difference between the local and the whole change. The difference between the two is the deviation of the gray scale change of a single pixel point relative to the contrast enhancement of the whole image, which is used to measure the difference between the local and the whole change. The difference between the two is the deviation of the gray scale change of a single pixel point relative to the contrast enhancement of the whole image, which is used to measure the difference between the local and the whole change. The difference between the two is the deviation of the gray scale change of a single pixel point relative to the contrast enhancement of the whole image, which is used to measure the difference between the local and the whole change. The difference between the two is the deviation of the gray scale change of a single pixel point relative to the contrast enhancement of the whole image, which is used to measure the difference between the local and the whole change. The difference between the two is the deviation of the gray scale change of a single pixel point relative to the contrast enhancement of the whole image, which is used to measure the difference between the local and the whole change. The average is taken to eliminate the influence of the number of pixels, to obtain a value that can represent the comprehensive difference between the pixel gray scale change and the contrast enhancement of the entire image, and to reflect the overall change characteristics of the image in these two aspects, to provide a basis for threshold adjustment. Then, an adjustment factor is constructed by adding 1 to the above average difference value. When the factor is greater than 1, the adjusted threshold value will increase; when it is less than 1, the adjusted threshold value will decrease. In this way, the adjustment direction and amplitude of the threshold value are determined according to the comprehensive situation of the image pixel gray scale change and the contrast enhancement. The formula comprehensively considers the gray scale change rate of each pixel and the contrast enhancement rate of the entire image, and obtains the adjustment factor through a series of calculations, which can accurately adjust the threshold value according to the actual image change. For excrement fluorescence images with different gray scale changes and contrast enhancement, the appropriate adjustment amplitude can be found, so that the adjusted threshold value is highly matched with the image features, and the accuracy of image segmentation is improved. Since the formula is based on the pixel gray scale change and overall contrast enhancement of the image itself, it can dynamically generate a suitable adjustment factor to adjust the threshold value for excrement fluorescence images with various imaging conditions and different gray scale distribution characteristics. Whether the image pixel gray scale change is simple or complex, and whether the contrast enhancement degree is large or small, the formula can effectively adapt to different image scenarios, enhancing the universality of the method in different image scenarios. The normalization and averaging operations in the formula effectively avoid the excessive interference of individual pixel abnormal changes or contrast enhancement ratio fluctuations on threshold adjustment. The threshold adjustment process is stable and reliable, and the adjusted threshold value will not fluctuate or change unreasonably, ensuring the stability and reliability of the image segmentation effect.

[0097] The technical effects of the above technical solution are: first, the excrement fluorescence images before and after contrast enhancement are respectively subjected to gray processing to obtain a first gray image and a second gray image. Then, the same position pixel points in the two images are extracted, the gray difference is calculated, and the gray change rate is further obtained. The gray change rate reflects the relative change of the gray values of the pixel points in the image during the contrast enhancement process, and is a key indicator for judging the degree of change in the image content. The contrast enhancement ratio measures the degree of contrast enhancement of the entire image. The combination of the gray change rate and the contrast enhancement ratio is because both of them reflect the change of the image from different angles. The gray change rate reflects the local pixel point change, and the contrast enhancement ratio reflects the overall contrast change. By comprehensively considering the information of both, the characteristic changes of the image after contrast enhancement can be more comprehensively grasped, and then the initial threshold parameter is adjusted according to these changes, so that the threshold is more consistent with the actual gray distribution of the enhanced image, to meet the accurate requirement of image segmentation. By considering the gray change rate and the contrast enhancement ratio to adjust the threshold, the gray features of the image after contrast enhancement can be more accurately adapted. In image segmentation, different regions in the image can be more accurately distinguished, the missegmentation caused by unreasonable threshold can be reduced, the consistency of the segmentation result and the actual image content can be improved, and the accuracy of image segmentation can be improved. The technical solution can dynamically adjust the threshold according to the actual situation of the contrast enhancement of different excrement fluorescence images (reflected by the gray change rate and the contrast enhancement ratio). No matter how large the image contrast enhancement amplitude is or how the gray change feature is, a suitable threshold can be found, which enhances the adaptability of the method to various imaging conditions and image features, so that it can work effectively in different scenarios. Compared with fixed threshold or threshold adjustment based on only a single indicator, the threshold adjustment based on the combination of the gray change rate and the contrast enhancement ratio can avoid the problem of poor segmentation effect caused by ignoring the local or overall change feature of the image. The threshold can be optimized from multiple dimensions, so that the effect of image segmentation after segmentation is more consistent with the actual requirement, and the overall quality of image processing is improved.

[0098] In the embodiment, feature extraction is performed on the excrement fluorescence image of the patient, including:

[0099] Based on the convolutional neural network CNN, the feature extraction is performed on the excrement fluorescence image of the patient, the semantic features valuable for the quality evaluation of the intestinal preparation are extracted from the excrement fluorescence image, and the extracted features are processed by dimension reduction based on the principal component analysis to determine the excrement feature data, including the excrement component, concentration, color, contrast and texture.

[0100] The model establishing module is configured to construct the intestinal preparation quality evaluation model according to the intestinal preparation quality evaluation requirement.

[0101] In the embodiment, the intestinal preparation quality evaluation model is constructed, including:

[0102] According to the AI-based bowel preparation quality evaluation requirement, historical excrement data is collected, and the collected historical excrement data is divided into a training set and a test set;

[0103] Based on the deep learning technology, the training set is used to train the deep learning model, so that the deep learning model learns the bowel preparation quality evaluation behavior from the training set, is used for quality evaluation of the patient's bowel preparation, judges the cleaning degree in the patient's bowel, and determines the deep learning-based bowel preparation quality evaluation model;

[0104] The test set is used to test the deep learning-based bowel preparation quality evaluation model, evaluate the performance of the deep learning-based bowel preparation quality evaluation model, judge whether the deep learning-based bowel preparation quality evaluation model can achieve the expected effect of quality evaluation of the patient's bowel preparation and judgment of the cleaning degree in the patient's bowel, and determine the model test evaluation result;

[0105] According to the model test evaluation result, the parameters of the deep learning-based bowel preparation quality evaluation model are adjusted and optimized until the deep learning-based bowel preparation quality evaluation model can achieve the expected effect of quality evaluation of the patient's bowel preparation and judgment of the cleaning degree in the patient's bowel, and then the optimal bowel preparation quality evaluation model is determined.

[0106] The quality evaluation module is used for feature recognition and analysis of the excrement feature data according to the bowel preparation quality evaluation model, quality evaluation of the patient's bowel preparation, and determination of the bowel preparation quality evaluation result.

[0107] In this embodiment, the quality evaluation of the patient's bowel preparation includes:

[0108] The optimal bowel preparation quality evaluation model is deployed, and the optimal bowel preparation quality evaluation model is deployed in the actual bowel preparation quality evaluation environment.

[0109] The excrement feature data is input into the bowel preparation quality evaluation model, and the excrement feature data is recognized and analyzed according to the bowel preparation quality evaluation model, the quality evaluation of the patient's bowel preparation is performed, and the cleaning degree in the patient's bowel is judged, and the bowel preparation quality evaluation result is determined.

[0110] The output management module is used for visual output of the patient's bowel preparation quality evaluation report, and subsequent colonoscopy management of the patient.

[0111] In this embodiment, the patient's bowel preparation quality evaluation report is visualized and output, and the patient's subsequent colonoscopy management includes:

[0112] According to the evaluation result of the intestinal preparation quality and in combination with the excrement fluorescence image of the patient, an intestinal preparation quality evaluation report of the patient is formed, and the report is displayed in a visual form to the doctor in real time, so as to enable the doctor to manage the patient in subsequent colonoscopy according to the intestinal preparation quality evaluation report of the patient.

[0113] In this embodiment, the cleaning degree of the intestinal tract of the patient is divided into four grades based on the evaluation result of the intestinal preparation quality, including:

[0114] Grade one, i.e., the intestinal mucosa is not clearly displayed, and there is a large amount of fecal residue in the intestinal tract;

[0115] Grade two, i.e., the intestinal mucosa is partially clearly displayed, and there is a medium amount of fecal residue in the intestinal tract;

[0116] Grade three, i.e., the intestinal mucosa is clearly displayed, and there is a small amount of fecal residue in the intestinal tract;

[0117] Grade four, i.e., the intestinal mucosa is clearly displayed, and there is no fecal residue in the intestinal tract;

[0118] Among them, the higher the grade, the better the intestinal preparation, and the more conducive to the colonoscopy.

[0119] Specifically, the excrement feature data is identified and analyzed according to the intestinal preparation quality evaluation model, the intestinal preparation of the patient is evaluated in quality, and the cleaning degree of the intestinal tract of the patient is determined to determine the intestinal preparation quality evaluation result, wherein the intestinal preparation quality evaluation result is shown in Table 1:

[0120]

[0121] Therefore, according to the intestinal preparation quality evaluation model, the excrement feature data is identified and analyzed, the intestinal preparation of the patient is evaluated in quality, and the cleaning degree of the intestinal tract of the patient is determined to determine the intestinal preparation quality evaluation result, and in combination with the excrement fluorescence image of the patient, an intestinal preparation quality evaluation report of the patient is formed, and the report is displayed in a visual form to the doctor in real time, so as to enable the doctor to manage the patient in subsequent colonoscopy according to the intestinal preparation quality evaluation report of the patient, which can provide more objective and efficient intestinal preparation quality evaluation result, thereby improving the accuracy and efficiency of the colonoscopy.

[0122] It is to be understood that the terminology used herein such as first and second, and the like, is only used to distinguish one entity or action from another entity or action, and does not necessarily require or imply any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0123] While embodiments of the present application have been shown and described with reference to particular embodiments thereof, it will be understood by those skilled in the art that various changes in form and details can be made therein without departing from the spirit and scope of the application. The scope of the application is defined by the appended claims and their equivalents.

Claims

1. An AI-based bowel preparation quality assessment system, characterized by, The method comprises the following steps: a collection processing module is configured to collect and pre-process and feature extract the excrement fluorescence image of the patient to determine excrement feature data; a model establishing module is configured to construct an intestinal preparation quality assessment model according to intestinal preparation quality assessment requirements; a quality assessment module is configured to perform feature recognition and analysis on the excrement feature data according to the intestinal preparation quality assessment model, to perform quality assessment on the intestinal preparation of the patient, and to determine an intestinal preparation quality assessment result; an output management module is configured to visually output the intestinal preparation quality assessment report of the patient and to perform subsequent colonoscopy management on the patient; the excrement fluorescence image of the patient is pre-processed, including: based on a filter, the excrement fluorescence image of the patient is denoised to remove noise in the excrement fluorescence image and reduce background interference in the excrement fluorescence image; based on histogram equalization, the excrement fluorescence image of the patient is subjected to contrast enhancement processing to enhance details in the excrement fluorescence image and highlight the fluorescent signal in the target region of the excrement fluorescence image; based on a threshold segmentation method, the excrement fluorescence image of the patient is subjected to image segmentation processing to separate the background and the target region in the excrement fluorescence image and eliminate non-specific fluorescent interference; the threshold value used for image segmentation processing is set, including: the contrast value before contrast enhancement processing and the contrast value after contrast enhancement processing are extracted; the contrast enhancement amplitude is obtained according to the contrast value before contrast enhancement processing and the contrast value after contrast enhancement processing; the contrast enhancement ratio is obtained by ratio processing of the contrast enhancement amplitude and the contrast value before contrast enhancement processing; the contrast enhancement ratio is compared with a preset contrast enhancement ratio threshold value; when the contrast enhancement ratio does not exceed the preset contrast enhancement ratio threshold value, the preset initial threshold value parameter used for image segmentation processing is called to perform image segmentation processing on the excrement fluorescence image of the patient based on the threshold segmentation method; when the contrast enhancement ratio exceeds the preset contrast enhancement ratio threshold value, the initial threshold value parameter is adjusted using the gray values of the excrement fluorescence image before and after contrast enhancement, including: when the contrast enhancement ratio exceeds the preset contrast enhancement ratio threshold value, the excrement fluorescence image before contrast enhancement processing is called; the excrement fluorescence image before contrast enhancement processing is subjected to gray processing to obtain a first gray image; the excrement fluorescence image after contrast enhancement processing is called; the excrement fluorescence image after contrast enhancement processing is subjected to gray processing to obtain a second gray image; each pixel point in the first gray image is extracted; each pixel point in the second gray image is extracted; the gray difference of the same position pixel point is obtained for the same position pixel point in the first gray image and the second gray image; the gray change rate of the same position pixel point is obtained by ratio processing of the gray difference of the same position pixel point and the gray value of the position pixel point in the first gray image; Adjust the initial threshold parameter by using the gray scale change rate and the contrast enhancement ratio; Perform image segmentation processing on the excrement fluorescence image of the patient by using the adjusted threshold parameter based on the threshold segmentation method; The adjusted threshold parameter is obtained by the following formula: Wherein, S represents the adjusted threshold parameter; S c represents the initial threshold parameter; n represents the number of pixel points contained in the excrement fluorescence image; P hi represents the gray scale change rate corresponding to the i-th pixel point; P d represents the contrast enhancement ratio.

2. The AI-based bowel preparation quality assessment system of claim 1, wherein, The excrement fluorescence image of the patient is collected, including: The excrement sample of the patient is collected in a sterile container, and the excrement sample of the patient is diluted or centrifuged to remove impurities in the excrement sample and concentrate the target component; Based on fluorescence imaging technology, a fluorescent dye is selected according to the characteristics of the target molecule in the excrement sample, the fluorescent dye is mixed with the excrement sample, and the two are fully contacted; The reaction conditions including pH value, temperature and time are selected according to the target molecule in the excrement sample, and the excrement sample is washed with a buffer solution to remove unbound dye in the excrement sample; The labeled sample is purified by centrifugation or filtration, and the labeled sample is placed in a fluorescence microscope, the excitation wavelength and filter combination are selected, and the excrement fluorescence image of the patient is obtained.

3. The AI-based bowel preparation quality assessment system of claim 2, wherein, Feature extraction is performed on the excrement fluorescence image of the patient, including: Based on the convolutional neural network (CNN), the excrement fluorescence image of the patient is extracted, the semantic features valuable for evaluating the quality of bowel preparation are extracted from the excrement fluorescence image, and the extracted features are dimensionally reduced based on principal component analysis to determine the excrement feature data, including excrement composition, concentration, color, contrast and texture.

4. The AI-based bowel preparation quality assessment system of claim 1, wherein, A bowel preparation quality evaluation model is constructed, including: According to the AI-based bowel preparation quality evaluation requirements, the excrement historical data is collected, and the collected excrement historical data is divided into a training set and a test set; Based on deep learning technology, the training set is used to train the deep learning model, so that the deep learning model learns the bowel preparation quality evaluation behavior from the training set, which is used to evaluate the quality of the patient's bowel preparation and determine the cleaning degree in the patient's intestine. The test set is used to test the deep learning-based bowel preparation quality evaluation model, evaluate the performance of the deep learning-based bowel preparation quality evaluation model, determine whether the deep learning-based bowel preparation quality evaluation model can achieve the expected effect of evaluating the quality of the patient's bowel preparation and determining the cleaning degree in the patient's intestine, and determine the model test evaluation result. According to the model test evaluation result, the parameters of the deep learning-based bowel preparation quality evaluation model are adjusted and optimized until the deep learning-based bowel preparation quality evaluation model can achieve the expected effect of evaluating the quality of the patient's bowel preparation and determining the cleaning degree in the patient's intestine, and then the optimal bowel preparation quality evaluation model is determined.

5. The AI-based bowel preparation quality assessment system of claim 4, wherein, The quality of the patient's bowel preparation is evaluated, including: The optimal bowel preparation quality evaluation model is deployed in the actual bowel preparation quality evaluation environment. The excrement feature data is input into the intestinal preparation quality evaluation model, feature recognition and analysis are performed on the excrement feature data according to the intestinal preparation quality evaluation model, the quality of intestinal preparation of the patient is evaluated, the cleaning degree in the intestinal tract of the patient is judged, and the intestinal preparation quality evaluation result is determined.

6. The AI-based bowel preparation quality assessment system of claim 1, wherein, The intestinal preparation quality evaluation report of the patient is visualized and output, and subsequent colonoscopy management is performed on the patient, including: According to the intestinal preparation quality evaluation result and in combination with the excrement fluorescence image of the patient, the intestinal preparation quality evaluation report of the patient is formed, and the intestinal preparation quality evaluation report of the patient is displayed in a visual form to the doctor in real time, so as to enable the doctor to perform subsequent colonoscopy management on the patient according to the intestinal preparation quality evaluation report of the patient.

7. The AI-based bowel preparation quality assessment system of claim 6, wherein, The cleaning degree in the intestinal tract of the patient is divided into four grades based on the intestinal preparation quality evaluation result.

Images