An artificial intelligence-based digestive tract detection system

Abstract

The application relates to the technical field of digestive tract detection, and discloses a digestive tract detection system based on artificial intelligence, which comprises a data acquisition unit, a preprocessing unit, a feature extraction unit, an abnormality detection unit and a result display unit. The application focuses on endoscopic images, accurately preprocesses the images into grayscale images and normalizes the grayscale images, reduces data complexity, highlights textures, and is beneficial to subsequent analysis; the application effectively extracts texture features, quantitatively presents subtle changes, accurately identifies lesions, scientifically judges abnormalities according to thresholds and displays the abnormalities, and realizes automatic and standardized detection. Meanwhile, physiological indexes are introduced, multidimensional data is integrated, and the function of the digestive tract is comprehensively reflected; the indexes are finely processed, individual differences and dynamic changes are considered, abnormal fluctuations are captured, and diseases are early warned; the application accurately detects abnormalities, cooperatively displays image and index results, and helps medical staff comprehensively judge. The application comprehensively and deeply detects, improves the accuracy of early screening, efficiently comprehensively diagnoses, has wide clinical applicability, promotes intelligent and accurate diagnosis and treatment, and the like.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digestive tract detection, and particularly relates to a digestive tract detection system based on artificial intelligence. BACKGROUND

[0002] With the improvement of people's living standards and the change of dietary structure, the incidence of digestive diseases is increasing year by year, which poses a serious threat to human health. The traditional digestive tract detection method mainly relies on the experience of doctors and the naked eye observation of endoscopic images, and the monitoring of physiological indicators is also scattered and lacks systematic integration, which gradually exposes many limitations in dealing with the increasingly complex digestive health problems.

[0003] In the aspect of digestive tract endoscopy, endoscopic images contain rich information about the morphology, color and texture of the digestive tract mucosa, which is the key basis for diagnosing digestive tract lesions. However, the original endoscopic image data is large and complex, and the image quality is uneven due to factors such as lighting conditions and endoscopic device imaging characteristics, which brings great challenges to doctors in accurately identifying lesion characteristics. Doctors stare at the endoscopic screen for a long time, which is easy to cause visual fatigue, and then some subtle but crucial signs of lesions may be missed, leading to delayed diagnosis.

[0004] For the monitoring of physiological indicators of the digestive tract, physiological indicators such as gastric acid concentration and intestinal peristalsis frequency are closely related to the normal function and development of digestive diseases. The previous detection method often measures these physiological indicators at a specific time and in isolation, which is difficult to capture their dynamic changes over time. Moreover, there are differences in the basic physiological state between different individuals, and simply relying on fixed normal range values to judge whether it is abnormal lacks precision and cannot meet the needs of early and accurate diagnosis of digestive diseases. SUMMARY

[0005] The present application aims to provide a digestive tract detection system based on artificial intelligence, which solves the technical problems proposed in the background art.

[0006] The purpose of the present application can be achieved by the following technical solutions:

[0007] A digestive tract detection system based on artificial intelligence, comprising:

[0008] A data acquisition unit for acquiring key data related to the digestive tract, including but not limited to: endoscopic image data and physiological indicator data of the digestive tract;

[0009] A preprocessing unit for preprocessing the key data;

[0010] The feature extraction unit is used to perform feature extraction processing on key data and extract endoscopic image texture features and digestive tract physiological indicators.

[0011] The abnormality detection unit is used to perform digestive tract abnormality detection and analysis based on the results obtained by the feature extraction unit, and to determine whether the endoscopic images are abnormal and whether the physiological index data are abnormal based on the analysis results, and to generate corresponding digestive tract abnormality signals from the abnormal results.

[0012] The results display unit is used to show the corresponding gastrointestinal abnormality signals obtained by the abnormality detection unit to relevant personnel.

[0013] As a further aspect of the present invention, the preprocessing method is as follows:

[0014] Step Y1: Endoscopic image data preprocessing:

[0015] Step Y1.1: Convert each endoscopic image into a grayscale image, as follows:

[0016] Select an endoscopic image;

[0017] Obtain the RGB values ​​of each pixel in the endoscopic image and record them as RGB. k ={R k G k B k} value, k=1, 2, ... v, v represents the total number of pixels in the endoscopic image, R k G represents the value of the R channel at the k-th pixel. k B represents the value of the G channel at the k-th pixel. k This represents the value of the B channel at the k-th pixel;

[0018] Subsequently passed: H k =0.299×R k +0.587×G k +0.114×B k ;

[0019] Calculate the gray value H of the k-th pixel in the endoscopic image. k ;

[0020] Step Y1.2: Select an endoscopic image and extract the maximum and minimum grayscale values ​​from each pixel in the image, and label them as H respectively. max and H min ;

[0021] Subsequently passed:

[0022] Calculate the normalized gray value H1 at each pixel.k ;

[0023] Step Y2, physiological index data preprocessing:

[0024] Step Y2.1, select a kind of physiological index data;

[0025] Mark the physiological index data as P r , r=1, 2, ……t, wherein t is the number of each physiological index value collected at multiple collection time points within a specified period;

[0026] Step Y2.2, then through: , the average value PP of the physiological index data is calculated;

[0027] At the same time, through: , the standard deviation BP of the physiological index data is calculated;

[0028] Step Y2.3, physiological index data P r Combined with PP and BP, the abnormal value in the physiological index data P r is analyzed;

[0029] Specifically: if P r -PP>3×BP, the corresponding physiological index data P r is marked as an abnormal value;

[0030] Step Y2.4, extract the maximum and minimum physiological index values from the physiological index data, and mark them as P max and P min ;

[0031] Step Y2.5, through: , the normalized value P1 r of the corresponding physiological index value in the physiological index data is calculated.

[0032] As a further scheme of the application: the endoscopic image texture feature extraction method is as follows:

[0033] Step T1.1, taking each pixel point in the endoscopic image as a center pixel point, simultaneously acquiring the pixel points adjacent to the center pixel point as neighborhood pixel points, and then forming a pixel domain according to the center pixel point and the neighborhood pixel points;

[0034] Step T1.2, in a pixel domain, the gray value of the center pixel point and the gray value of the neighborhood pixel points are acquired, and are marked as H0 k and HL k，u , respectively, u=0, 1, 2, ……c-1, c represents the number of neighborhood pixel points;

[0035] Step T1.3, calculate the binary pattern value LBP of the pixel field by:

[0036] ; k ;

[0037] wherein, 2 u is a weight coefficient, used for converting the binary number into a decimal number to represent the binary pattern value;

[0038] Step T1.4, extract all binary pattern values LBP in all pixel fields in the endoscopic image k ;

[0039] According to the frequency of the occurrence of different binary pattern values, the LBP histogram of the endoscopic image is obtained and marked as Z=[z0, z1, …, zw], w is the number of different binary pattern values; w-1 ;

[0040] Then, calculate the average LBP value ZP of the endoscopic image by:

[0041] ;

[0042] wherein, d=0, 1, 2, …, w-1, d represents the binary pattern value, z d is the number of pixels with the binary pattern value d;

[0043] Meanwhile, calculate the dispersion value ZB of the LBP value in the endoscopic image by:

[0044] ;

[0045] As a further scheme of the present application: the way of extracting the physiological index features of the digestive tract is as follows:

[0046] Step T2.1, select a kind of physiological index data, and extract the physiological index data P r and P r-1 collected at two adjacent collection time points in a specified period;

[0047] Step T2.2, calculate the corresponding physiological index change rate E r by: ;

[0048] ;

[0049] As a further scheme of the present application: the types of physiological index data include gastric acid concentration and intestinal peristalsis frequency.

[0050] As a further scheme of the present application: the abnormal detection analysis method is as follows:

[0051] Step J1, endoscope image abnormality detection:

[0052] Step J1.1, extract the pre-set LBP mean threshold ZPy and LBP dispersion degree threshold ZBy;

[0053] Step J1.2, extract the pre-set LBP mean threshold ZPc and LBP dispersion degree threshold ZBc at the same time;

[0054] Step J1.3, then compare the results between |ZP-ZPy| and ZPc, and the results between |ZB-ZBy| and ZBc, to determine whether the endoscope image is abnormal;

[0055] If |ZP-ZPy|>ZPc and |ZB-ZBy|>ZBc are true at the same time, it is determined that the endoscope image is abnormal, and a digestive tract abnormality signal one is generated, otherwise, no digestive tract abnormality signal one is generated;

[0056] Step J2, physiological index data abnormality detection:

[0057] Step J2.1, extract the pre-set change rate threshold Ey according to the corresponding physiological index data;

[0058] Step J2.2, extract the pre-set change rate threshold Ec at the same time;

[0059] Step J2.3, then compare the results between |E r -Ey| and Ec to determine whether the physiological index data is abnormal;

[0060] If |E r -Ey|>Ec is true, it is determined that the related physiological index data is abnormal, and a digestive tract abnormality signal two is generated, otherwise, no digestive tract abnormality signal two is generated.

[0061] Advantages of the present application:

[0062] The present application can obtain key data including endoscope image data of the digestive tract and physiological index data such as gastric acid concentration, intestinal peristalsis frequency, etc., and corresponding preprocessing procedures are set for different types of data, such as endoscope image data and physiological index data, so that the data is more standardized and accurate before entering the subsequent analysis link, which helps to improve the reliability and accuracy of the entire detection system. For example, the endoscope image is converted into a grayscale image, normalized, etc. The physiological index data is processed by calculating the mean value, standard deviation, analyzing the abnormal value and normalizing, etc. The interference factors in the original data can be better removed and the effective features can be highlighted.

[0063] The present application can extract the texture features of the endoscope image in detail and quantitatively by calculating the binary pattern value based on the central pixel point and the neighborhood pixel points, and then obtaining the LBP histogram and further calculating the average LBP value and the dispersion degree value, which provides a strong basis for accurately judging whether the endoscope image is abnormal, and helps to accurately find the abnormal conditions such as the signs of lesions in the endoscope image.

[0064] The present application can capture the dynamic changes of the physiological index data over time by calculating the change rate of the physiological index at adjacent collection time points, which plays an important role in understanding the stability of the digestive tract physiological function and timely detecting potential physiological abnormalities, and is convenient for subsequent judgment of whether the corresponding physiological index data is abnormal based on the change rate.

[0065] The present application sets up corresponding abnormality detection mechanisms for the endoscope image and the physiological index data respectively, compares the extracted features such as the average LBP value and the dispersion degree value of the endoscope image, the change rate of the physiological index, etc., with the pre-set threshold value and its allowable deviation value, can scientifically and rigorously judge whether the endoscope image is abnormal and whether the physiological index is abnormal, and thus accurately generates the corresponding digestive tract abnormality signal, realizes effective monitoring of the digestive tract abnormality, and helps to find related problems such as digestive diseases as early as possible.

[0066] The present application sets up a result display unit, which can intuitively display the digestive tract abnormality signal obtained by the abnormality detection unit to relevant personnel, so that relevant medical staff, researchers, etc. can know the detection result in time, and then make further diagnosis, treatment or research decision, etc. according to the abnormality.

[0067] In summary, the digestive tract detection system can improve the comprehensiveness, accuracy and early warning ability of the digestive tract detection through the cooperation of multiple links, and has a positive significance for the early detection and diagnosis of digestive tract related diseases. BRIEF DESCRIPTION OF DRAWINGS

[0068] The present application will be further described below in conjunction with the drawings.

[0069] Figure 1 is a system block diagram of a digestive tract detection system based on artificial intelligence. DETAILED DESCRIPTION

[0070] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings of the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0071] Embodiment one

[0072] Please refer to Figure 1 As shown in the figure, the present application is a kind of based on artificial intelligence's digestive tract detection system, comprising:

[0073] Data acquisition unit, for obtaining the key data related to digestive tract, key data includes the endoscopic image data of digestive tract;

[0074] Pretreatment unit, for pretreatment to endoscopic image data, pretreatment mode as follows:

[0075] Step Y1.1, each endoscopic image is converted into gray scale image, its mode as follows:

[0076] Select an endoscopic image;

[0077] Get the RGB value of each pixel point on the endoscopic image, and mark it as RGB k ={R k , G k , B k} value, k = 1, 2, …… v, v indicates the number of all pixel points on endoscopic image, R k It represents the value of R channel on the kth pixel point, G k It represents the value of G channel on the kth pixel point, B k It represents the value of B channel on the kth pixel point;

[0078] Then through: H k =0.299×R k +0.587×G k +0.114×B k ;

[0079] Calculate the gray value H k of the kth pixel point on the endoscopic image;

[0080] Step Y1.2, select an endoscopic image, and get the maximum and minimum gray value from each pixel point on the endoscopic image, and mark them as H max And H min ;

[0081] Then through:

[0082] Calculate the normalized gray value H1 k of each pixel point;

[0083] Feature extraction unit, for key data feature extraction processing, feature extraction processing is to extract endoscopic image texture feature:

[0084] Step T1.1: Take each pixel in the endoscopic image as the center pixel, and at the same time obtain the pixels surrounding the center pixel and adjacent to the center pixel as the neighboring pixels. Then, form a pixel domain based on the center pixel and the neighboring pixels.

[0085] Step T1.2: Within a pixel domain, obtain the grayscale value of the center pixel and the grayscale values ​​of the neighboring pixels, and label them as H0 respectively. k and HL k，u u = 0, 1, 2, ..., c-1, where c represents the number of neighboring pixels;

[0086] Step T1.3, Pass:

[0087] Calculate the binary mode value LBP for this pixel domain. k ;

[0088] In this embodiment, 2 u These are weighting coefficients used to convert binary numbers into decimal numbers to represent binary pattern values;

[0089] Step T1.4: Extract all binary pattern values ​​(LBP) from all pixel domains in the endoscopic image. k ;

[0090] Subsequently, based on the frequency of occurrence of different binary pattern values, the LBP histogram of the endoscopic image was obtained and labeled as Z=[z0, z1, ..., z...]. w-1 ], where w is the number of different binary pattern values;

[0091] Then proceed as follows:

[0092] Calculate the average LBP value ZP of the endoscopic image;

[0093] In the formula, d = 0, 1, 2, ..., w-1, where d represents the binary pattern value, and z d This represents the number of pixels with a value of d in binary mode.

[0094] Simultaneously through:

[0095] Calculate the dispersion value ZB of LBP in the endoscopic image;

[0096] The anomaly detection unit is used to perform gastrointestinal anomaly detection and analysis based on the results obtained from the feature extraction unit. The anomaly detection and analysis involves detecting anomalies in endoscopic images, and the method is as follows:

[0097] Step J1.1: Extract the pre-set LBP mean threshold ZPy and LBP dispersion threshold ZBy;

[0098] Step J1.2, simultaneously extracting the allowable deviation value ZPc of the preset LBP mean threshold value and the allowable deviation value ZBc of the LBP dispersion degree threshold value;

[0099] Step J1.3, then determining whether the endoscopic image is abnormal according to the comparison result between |ZP-ZPy| and ZPc, and the comparison result between |ZB-ZBy| and ZBc;

[0100] If |ZP-ZPy|>ZPc and |ZB-ZBy|>ZBc are both true, it is determined that the endoscopic image is abnormal, and a digestive tract abnormality signal one is generated;

[0101] A result display unit is configured to display the digestive tract abnormality signal one obtained by the abnormality detection unit to relevant personnel.

[0102] In this embodiment, the endoscopic image is converted into a grayscale image, the grayscale value of each pixel point is accurately calculated by using a specific formula, and the grayscale value is normalized. This series of operations effectively reduces the complexity of the image data, removes the interference caused by color information redundancy, highlights the texture and structural features of the image, and lays a foundation for accurately detecting the abnormality of the endoscopic image. The pixel domain is constructed based on the pixel point, the grayscale values of the central pixel point and the neighborhood pixel points are combined to calculate the binary pattern value, and then the LBP histogram is counted, and the average LBP value and the dispersion degree value are obtained. This method can deeply mine the texture features of the endoscopic image, quantize the subtle and complex texture changes in the image, and enable the system to have stronger recognition ability for the lesion features in the endoscopic image, greatly improving the accuracy of the endoscopic image abnormality detection, and enabling the system to more timely discover potential digestive tract mucosa lesions such as early-stage polyps and ulcers. According to the LBP mean threshold value, the dispersion degree threshold value and the allowable deviation value, which are preset and verified by a large number of experiments and clinical data, the endoscopic image is rigorously determined to be abnormal or not, and a digestive tract abnormality signal one is generated and displayed to relevant personnel. This process realizes the automation and standardization of the detection process, avoids the subjective observation and visual fatigue caused by the missed diagnosis problem, and ensures that the abnormal situation can be known in the first time, thereby providing a key basis for subsequent diagnosis and treatment decisions.

[0103] Embodiment Two

[0104] Please refer to Figure 1 As shown in FIG. 2, as embodiment two of the present application, compared with embodiment one, the technical solution of the present embodiment is only different from that of embodiment one in that in the present embodiment,

[0105] The key data further includes physiological index data of the digestive tract;

[0106] The preprocessing unit is also used to preprocess the physiological indicator data, and the preprocessing manner is as follows:

[0107] Step Y2.1, selecting a kind of physiological indicator data;

[0108] In this embodiment, the kinds of physiological indicator data include gastric acid concentration and intestinal peristalsis frequency;

[0109] The physiological indicator data is marked as P r , r = 1, 2, …… t, wherein t is the number of collection time points in a specified period for each physiological indicator value;

[0110] Step Y2.2, then calculating the average value PP of the physiological indicator data by:

[0111] At the same time, the standard deviation BP of the physiological indicator data is calculated by:

[0112] Step Y2.3, the abnormal value in the physiological indicator data P r is analyzed by combining PP and BP; r

[0113] Specifically, if P r -PP > 3×BP, the corresponding physiological indicator data P r is marked as an abnormal value;

[0114] Step Y2.4, the maximum and minimum physiological indicator values of the physiological indicator data are extracted from the physiological indicator data, and are marked as P max and P min ;

[0115] Step Y2.5, the normalized value P1 r of the corresponding physiological indicator value in the physiological indicator data is calculated by:

[0116] The feature extraction processing is also used to extract physiological indicator features:

[0117] Step T2.1, selecting a kind of physiological indicator data, and extracting the physiological indicator data P r and P r-1 collected at two adjacent collection time points in a specified period;

[0118] Step T2.2, calculating the corresponding physiological indicator change rate E r by:

[0119] ​​​​​​

[0120] The abnormality detection analysis is to detect the abnormality of the physiological index data in the following manner:

[0121] Step J2.1, extract the change rate threshold Ey preset according to the corresponding physiological index data;

[0122] Step J2.2, extract the allowable deviation value Ec of the preset change rate threshold at the same time;

[0123] Step J2.3, then determine whether the physiological index data is abnormal according to the comparison result between |E r -Ey| and Ec;

[0124] If |E r -Ey|>Ec is established, it is determined that the related physiological index data is abnormal, and then the gastrointestinal abnormality signal two is generated;

[0125] The result display unit is also used to display the gastrointestinal abnormality signal one obtained by the abnormality detection unit to the relevant personnel.

[0126] The embodiment in Embodiment One is only based on the detection of endoscopic image data, and introduces the physiological index data of the digestive tract, such as the key indicators of gastric acid concentration and intestinal peristalsis frequency. It realizes the multi-dimensional detection from single image dimension to the combination of image and physiological index, makes the evaluation of the health status of the digestive tract more comprehensive and stereoscopic, and can more comprehensively reflect the functional state of the digestive tract, avoiding missing some functional lesions or early disease signs due to relying only on endoscopic images. The physiological index data is systematically preprocessed, including calculating the average value, standard deviation, accurately identifying abnormal values, and normalization processing. These operations fully consider the differences between individuals and the dynamic change characteristics over time of the physiological index, making the subsequent change rate-based feature extraction and abnormality detection more scientific and reliable, and being able to accurately capture the slight abnormal fluctuations of the physiological index, and early warning the possible digestive disease risks, such as gastritis caused by abnormal gastric acid secretion, and early stage of intestinal obstruction implied by abnormal intestinal peristalsis. By calculating the change rate of the physiological index at adjacent collection time points, and comparing with the preset threshold and allowable deviation value, it is accurately determined whether the physiological index is abnormal, and the gastrointestinal abnormality signal two is generated. The result display unit displays both signal one and signal two, realizes the collaborative presentation of image and physiological index detection results, and enables medical staff to obtain comprehensive information at one time, quickly and comprehensively judge the digestive tract condition of the patient, and develop accurate diagnosis and treatment plan.

[0127] Embodiment Three

[0128] Please refer to Figure 1 As shown in FIG. 3, as Embodiment Three of the present application, compared with Embodiments One and Two, the technical solution of the present embodiment is to combine the schemes of Embodiments One and Two for implementation.

[0129] This embodiment combines the schemes of Embodiment One and Embodiment Two, possessing both the high-precision endoscopic image anomaly detection capability of Embodiment One and the comprehensive physiological indicator monitoring function of Embodiment Two, realizing full-range, deep-level detection of the digestive tract from morphological structure to physiological function. Whether it is a small lesion on the surface of the digestive tract mucosa or a potential anomaly in internal physiological function, the system can accurately capture it, greatly improving the accuracy and comprehensiveness of early screening for digestive tract diseases. By integrating endoscopic image texture feature analysis and dynamic monitoring of multiple physiological indicators, doctors are provided with richer and more valuable diagnostic evidence. Different types of data complement each other, making the diagnosis of complex digestive tract diseases more accurate and efficient, reducing the probability of misdiagnosis and missed diagnosis. Especially when facing some atypical symptoms and complex digestive tract diseases, the system can quickly lock the root cause of the problem, helping patients to seize the best treatment opportunity. This embodiment combines the strengths of the previous two embodiments and is suitable for various clinical scenarios with different needs, whether it is initial screening of the digestive tract in routine physical examination or in-depth examination of patients with digestive discomfort symptoms. It can play its advantages and meet the diversified medical detection needs, promoting the intelligent and precise diagnosis and treatment of digestive tract diseases.

[0130] The above formulas are dimensionless numerical calculations, and the formulas are obtained by software simulation of a large number of data to obtain a formula closest to the real situation. The preset parameters and threshold values in the formula are set by a person skilled in the art according to the actual situation.

[0131] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A digestive tract detection system based on artificial intelligence, characterized in that, include: The data acquisition unit is used to acquire key data related to the digestive tract, including but not limited to: endoscopic image data and physiological indicator data of the digestive tract; The preprocessing unit is used to preprocess key data; The feature extraction unit is used to perform feature extraction processing on the preprocessed key data and extract the texture features of the endoscopic image and the physiological indicators of the digestive tract. The method for extracting texture features from endoscopic images is as follows: Step T1.1: Take each pixel in the endoscopic image as the center pixel, and at the same time obtain the pixels surrounding the center pixel and adjacent to the center pixel as the neighboring pixels. Then, form a pixel domain based on the center pixel and the neighboring pixels. Step T1.2: Within a pixel domain, obtain the grayscale value of the center pixel and the grayscale values ​​of the neighboring pixels, and label them as H0 respectively. k and HL k，u u = 0, 1, 2, ..., c-1, where c represents the number of neighboring pixels; Step T1.3, Pass: Calculate the binary mode value LBP for this pixel domain. k ; among them, 2 u These are weighting coefficients used to convert binary numbers into decimal numbers to represent binary pattern values; Step T1.4: Extract all binary pattern values ​​(LBP) from all pixel domains in the endoscopic image. k ; Subsequently, based on the frequency of occurrence of different binary pattern values, the LBP histogram of the endoscopic image was obtained and labeled as Z=[z0, z1, ..., z...]. w-1 ], where w is the number of different binary pattern values; Then proceed as follows: The average LBP value ZP of the endoscopic image was calculated; where d = 0, 1, 2, ..., w-1, d represents the binary mode value, and z d This represents the number of pixels with a value of d in binary mode. Simultaneously through: The dispersion value ZB of LBP in the endoscopic image was calculated. The methods for extracting digestive tract physiological indicators are as follows: Step T2.1: Select a physiological indicator data set and extract the physiological indicator data P collected at two adjacent collection time points within a specified period. r and P r-1 ; Step T2.2, Pass: Calculate the corresponding physiological indicator change rate E r ; The abnormality detection unit is used to perform digestive tract abnormality detection and analysis based on the results obtained by the feature extraction unit, and to determine whether the endoscopic images are abnormal and whether the physiological indicators are abnormal based on the analysis results, and to generate corresponding digestive tract abnormality signals from the abnormal results. Gastrointestinal abnormality detection and analysis includes endoscopic image abnormality detection and analysis and physiological indicator data abnormality detection and analysis; The methods for detecting and analyzing abnormalities in endoscopic images are as follows: Step J1.1: Extract the pre-set LBP mean threshold ZPy and LBP dispersion threshold ZBy; Step J1.2: Simultaneously extract the allowable deviation value ZPc of the pre-set LBP mean threshold and the allowable deviation value ZBc of the LBP dispersion threshold; Step J1.3: Next, compare the results between |ZP-ZPy| and ZPc, and between |ZB-ZBy| and ZBc, to determine whether the endoscopic image is abnormal. The methods for detecting and analyzing abnormal physiological indicator data are as follows: Step J2.1: Extract the pre-set rate of change threshold Ey based on the corresponding physiological indicator data; Step J2.2: Simultaneously extract the allowable deviation value Ec of the pre-set rate of change threshold; Step J2.3, then |E r The comparison between -Ey| and Ec determines whether physiological indicators are abnormal. The results display unit is used to show the corresponding gastrointestinal abnormality signals obtained by the abnormality detection unit to relevant personnel.

2. The digestive tract detection system based on artificial intelligence according to claim 1, characterized in that, Endoscopic image data preprocessing is performed as follows: Step Y1.1: Convert each endoscopic image into a grayscale image, as follows: Select an endoscopic image; Obtain the RGB values ​​of each pixel in the endoscopic image and record them as RGB. k ={R k G k B k } value, k=1, 2, ... v, v represents the total number of pixels in the endoscopic image, R k G represents the value of the R channel at the k-th pixel. k B represents the value of the G channel at the k-th pixel. k This represents the value of the B channel at the k-th pixel; Subsequently passed: H k =0.299×R k +0.587×G k +0.114×B k ; Calculate the gray value H of the k-th pixel in the endoscopic image. k ; Step Y1.2: Select an endoscopic image and extract the maximum and minimum grayscale values ​​from each pixel in the image, and label them as H respectively. max and H min ; Subsequently passed: Calculate the normalized gray value H1 at each pixel. k .

3. The digestive tract detection system based on artificial intelligence according to claim 1, characterized in that, The preprocessing of physiological indicator data is performed as follows: Step Y2.1: Select a physiological indicator data; The physiological index data is labeled as P. r r = 1, 2, ..., t, where t is the number of times each physiological indicator value is collected at multiple collection time points within a specified period; Step Y2.2, followed by: Calculate the average value PP of this physiological indicator data; Simultaneously through: Calculate the standard deviation (BP) of this physiological indicator data; Step Y2.3: Transfer the physiological index data P r By combining PP and BP, the physiological index data P was analyzed. r Outliers in; Step Y2.4: Extract the maximum and minimum values ​​of the physiological indicator from this type of physiological indicator data, and label them as P. max and P min ; Step Y2.5, Pass: Calculate the normalized value P1 of the corresponding physiological indicator value in this type of physiological indicator data. r .

4. The digestive tract detection system based on artificial intelligence according to claim 3, characterized in that, In step Y2.3, if P r -PP>3×BP, then the corresponding physiological index data P r Mark as an outlier.

5. The digestive tract detection system based on artificial intelligence according to claim 1, characterized in that, In step J1.3, if both |ZP-ZPy|>ZPc and |ZB-ZBy|>ZBc are true, then the endoscopic image is determined to be abnormal, and a digestive tract abnormality signal 1 is generated accordingly; otherwise, a digestive tract abnormality signal 1 is not generated.

6. The digestive tract detection system based on artificial intelligence according to claim 1, characterized in that, In step J2.3, if |E r If -Ey|>Ec is true, then the relevant physiological indicators are determined to be abnormal, and a second abnormal signal for the digestive tract is generated accordingly; otherwise, a second abnormal signal for the digestive tract is not generated.

Images