Image auxiliary analysis system for hospital department based on deep learning

Abstract

The invention relates to the technical field of image auxiliary analysis, and particularly discloses an image auxiliary analysis system for hospital departments based on deep learning. According to the method, the morphological evaluation index, the density evaluation index and the texture evaluation index of the focus area of each preprocessed medical image corresponding to the current time period of the hospital department are comprehensively analyzed, so that the analysis result of the abnormal degree grade of the focus area is more reasonable and reliable, and the uncertainty of traditional manual subjective grading is realized; a unified and objective standard is provided for disease assessment, the diagnosis and treatment process can be standardized, the homogenization level of medical services is improved, doctors can rapidly master key disease key points from abnormal levels and focus position information given by the system, time is saved in busy clinical work, the severity of diseases is rapidly judged, and the diagnosis and treatment efficiency is improved. Therefore, whether further examination is needed or not and which treatment measure is taken are determined, medical resource allocation is optimized, diagnosis and treatment efficiency is improved, and a patient can be treated and cured timely and properly.

Description

Technical Field

[0001] The present invention relates to the field of image-assisted analysis technology, and specifically to an image-assisted analysis system for hospital departments based on deep learning. Background Art

[0002] In the modern medical field, medical imaging technology has become a key means of disease diagnosis, treatment monitoring and prognosis assessment. From traditional X-rays and CT (computed tomography) to more advanced MRI (magnetic resonance imaging) and PET (positron emission tomography), various imaging devices continuously generate massive amounts of medical imaging data. However, in the face of these complex and huge data, medical personnel face many severe challenges.

[0003] On the one hand, different imaging devices generate different impact modalities with different formats, resolutions, grayscale ranges, etc., which poses a huge obstacle to the unified processing and analysis of images; on the other hand, the complexity and professionalism of images make it more difficult for doctors to interpret them. Faced with massive amounts of imaging information, manual analysis is not only time-consuming and labor-intensive, but also prone to missed diagnoses and misdiagnoses due to factors such as visual fatigue.

[0004] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0005] The purpose of the present invention is to provide an image-assisted analysis system for hospital departments based on deep learning to solve the problems raised in the above background.

[0006] The purpose of the present invention can be achieved by the following technical solution: An image-assisted analysis system for hospital departments based on deep learning, comprising:

[0007] The image acquisition and preprocessing module is used to acquire medical images of various imaging devices in the hospital department corresponding to the current period, obtain medical images of various imaging devices in the hospital department corresponding to the current period, and perform preprocessing operations on them to obtain preprocessed medical images of the hospital department corresponding to the current period;

[0008] The lesion area recognition and analysis module is used to perform lesion area recognition and analysis on the pre-processed medical images of the hospital department corresponding to the current period, and obtain the lesion area of ​​the pre-processed medical images of the hospital department corresponding to the current period;

[0009] A regional feature parameter acquisition module is used to extract the lesion region feature parameters of each pre-processed medical image of the hospital department corresponding to the current period based on deep learning technology, and obtain the lesion region feature parameters of each pre-processed medical image of the hospital department corresponding to the current period, wherein the feature parameters are specifically: morphological feature parameters, density feature parameters and texture feature parameters; a lesion region morphological analysis module is used to analyze the morphological feature parameters of the lesion region of each pre-processed medical image of the hospital department corresponding to the current period, and obtain the morphological evaluation index of the lesion region of each pre-processed medical image of the hospital department corresponding to the current period;

[0010] The lesion area density analysis module is used to analyze the density characteristic parameters of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period, and obtain the density evaluation index of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period;

[0011] The lesion area texture analysis module is used to analyze the texture feature parameters of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period, and obtain the texture evaluation index of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period;

[0012] The auxiliary evaluation and analysis module is used to analyze the abnormality level of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period, and obtain the abnormality level of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period;

[0013] The display terminal performs corresponding display operations based on the analysis result information of the pre-processed medical images of the hospital department corresponding to the current period.

[0014] Furthermore, the lesion area recognition and analysis is performed on the pre-processed medical images of the hospital departments corresponding to the current period. The specific execution steps are as follows:

[0015] Performing regional division on each preprocessed medical image of the hospital department corresponding to the current period according to a set division method, obtaining each sub-region of each preprocessed medical image of the hospital department corresponding to the current period, and extracting the tissue and organ structure of each sub-region of each preprocessed medical image, obtaining the tissue and organ structure of each sub-region of each preprocessed medical image of the hospital department corresponding to the current period, and matching the reference gray value interval range corresponding to the tissue and organ structure of each sub-region stored in the cloud database, obtaining the reference gray value interval range of each sub-region of each preprocessed medical image of the hospital department corresponding to the current period;

[0016] Obtain the grayscale value of each pixel in each sub-region of the preprocessed medical images of the hospital department corresponding to the current period, and compare and analyze it with the reference grayscale value range of each sub-region. If the grayscale value of a pixel in a region is greater than the maximum value of the reference grayscale value range corresponding to the sub-region, then the pixel in the sub-region is marked as a target pixel. If the grayscale value of a pixel in a region is less than the minimum value of the reference grayscale value range corresponding to the sub-region, then the pixel in the sub-region is marked as a marked pixel. In this way, the number of target pixels and the number of marked pixels in each sub-region of the preprocessed medical images of the hospital department corresponding to the current period are statistically obtained, and are marked as i represents the number of each preprocessed medical image, i=1,2,3,....,m, m represents the total number of preprocessed medical image numbers, j represents the number of each sub-region, j=1,2,3,....,n, n represents the total number of sub-region numbers;

[0017] The grayscale values ​​of each target pixel in each sub-region are extracted from the grayscale values ​​of each pixel in each sub-region of the preprocessed medical images of the hospital department corresponding to the current period, and the grayscale values ​​of each target pixel in each sub-region are calculated by difference with the maximum value of the reference grayscale value interval corresponding to each sub-region to obtain the grayscale value difference of each target pixel in each sub-region, and then the summation is performed to obtain the total grayscale value difference of the target pixel in each sub-region of the preprocessed medical images of the hospital department corresponding to the current period, which is recorded as The grayscale values ​​of each marked pixel point in each sub-region are extracted from the grayscale values ​​of each pixel point in each sub-region of the preprocessed medical images of the hospital department corresponding to the current period, and the grayscale value difference of each marked pixel point in each sub-region is obtained by subtracting it from the minimum value of the reference grayscale value interval corresponding to each sub-region, and then the summation is performed to obtain the total grayscale value difference of the marked pixel points in each sub-region of the preprocessed medical images of the hospital department corresponding to the current period, which is recorded as According to the formula Calculate the grayscale matching evaluation index of each sub-region of each preprocessed medical image of the hospital department corresponding to the current period a1, a2, a3, and a4 represent the weight factors corresponding to the set number of target pixels, the number of marked pixels, the total grayscale value difference of target pixels, and the total grayscale value difference of marked pixels, respectively;

[0018] The grayscale matching evaluation index of each sub-region of the preprocessed medical images of the hospital department corresponding to the current time period is compared and analyzed with the set reference grayscale matching evaluation index threshold. When the grayscale matching evaluation index of a region is greater than the set reference grayscale matching evaluation index threshold, the sub-region is judged to be a normal region. When the grayscale matching evaluation index of a region is less than the set reference grayscale matching evaluation index threshold, the sub-region is judged to be a lesion region. In this way, the lesion regions of the preprocessed medical images of the hospital department corresponding to the current time period are obtained through integration.

[0019] Furthermore, the morphological characteristic parameters of the lesion areas of the pre-processed medical images of the hospital departments corresponding to the current period are analyzed. The specific analysis method is as follows:

[0020] The maximum long-axis distance of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period is obtained from the morphological characteristic parameters of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, and it is used as the long-axis distance of the lesion area. At the same time, the maximum short-axis distance of the lesion area is obtained and used as the short-axis distance of the lesion area. The ratio of the long-axis distance to the short-axis distance of the lesion area is calculated to obtain the aspect ratio of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, which is recorded as AR i ;

[0021] Get the area and perimeter of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, and mark them as A i , P i , according to the formula Calculate the circularity C of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period i ;

[0022] Obtain the lobulation index of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period and mark it as LI i , according to the formula Calculate the morphological evaluation index XP of the lesion area of ​​the target patient's medical image i , b1, b2, b3, b4, and b5 represent the weight factors corresponding to the aspect ratio, area, perimeter, circularity, and lobulation index of the set lesion area, respectively.

[0023] Furthermore, the density characteristic parameters of the lesion areas of the pre-processed medical images of the hospital departments corresponding to the current period are analyzed. The specific analysis method is as follows:

[0024] The grayscale values ​​of all pixels in the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period are obtained from the density feature parameters of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, and the mean is calculated to obtain the average grayscale value of all pixels in the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, which is recorded as At the same time, the maximum grayscale value, the minimum grayscale value and the median grayscale value are extracted from the grayscale values ​​of all pixels in the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, and marked as According to the formula

[0025] Calculate the gray value dispersion LD of all pixels in the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period i , c1, c2, c3 represent the set proportional factors respectively;

[0026] The grayscale value of each pixel in the lesion area is extracted from the grayscale values ​​of all pixels in the lesion area of ​​the preprocessed medical images of the hospital department corresponding to the current period, and is recorded as f represents the number of all pixel points in the lesion area, f = 1, 2, 3, ...., g, g represents the total number of all pixel point numbers;

[0027] According to the formula Calculate the gray value variation coefficient HI of all pixels in the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period i ;

[0028] According to the formula Calculate the density assessment index MP of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period i , c4 and c5 represent the weight factors corresponding to the set gray value discreteness and gray value variation coefficient respectively.

[0029] Furthermore, the texture feature parameters of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period are analyzed. The specific analysis method is as follows:

[0030] The contrast, correlation, energy and entropy of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period are obtained from the texture feature parameters of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, and they are marked as CON i , COR i 、ASM i 、E i , according to the formula Calculate the texture evaluation index WP of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period i , d1, d2, d3, and d4 represent the weight factors corresponding to the contrast, correlation, energy, and entropy of the set lesion area, respectively.

[0031] Furthermore, the abnormal levels of the lesion areas of the pre-processed medical images of the hospital departments corresponding to the current period are analyzed. The specific analysis method is as follows:

[0032] Comprehensively analyze the morphological evaluation index, density evaluation index and texture evaluation index of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period, and obtain the abnormality evaluation coefficient of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period;

[0033] Matching the abnormality evaluation coefficients of the lesion areas of the preprocessed medical images of the hospital departments corresponding to the current period with the abnormality evaluation coefficient thresholds corresponding to the set abnormality levels, to obtain the abnormality levels of the lesion areas of the preprocessed medical images of the hospital departments corresponding to the current period;

[0034] The analysis result information of each preprocessed medical image of the hospital department corresponding to the current period is constituted by the lesion area location and abnormality degree level of each preprocessed medical image of the hospital department corresponding to the current period.

[0035] Beneficial effects of the present invention:

[0036] The present invention uses a detailed medical image area division, tissue and organ structure extraction and gray value matching process, and no longer relies solely on the doctor's experience and judgment. Instead, it accurately identifies the lesion area based on the quantitative gray interval standard, makes a fine comparison of the gray values ​​of the pixel points, combines the statistical analysis of the target pixel points and the marked pixel points, and comprehensively considers the gray matching evaluation index. It can keenly capture tiny abnormal gray changes, greatly improve the detection rate of early lesions and tiny lesions, reduce the risk of misdiagnosis and missed diagnosis, and buy precious time for early intervention of the disease. It avoids subjective judgment bias caused by factors such as artificial visual fatigue and experience differences, and ensures the accuracy and stability of diagnosis.

[0037] The present invention uses deep learning technology to mine multi-dimensional feature parameters of the morphology, density, and texture of the lesion area in medical images. These parameters reflect the characteristics of the lesions from different aspects. Morphological feature parameters such as aspect ratio, circularity, and lobulation index can assist in judging the growth mode and invasiveness of the lesions, and provide key clues for distinguishing benign and malignant lesions; density feature parameters accurately reflect the uniformity of the internal structure of the lesion by measuring the gray value discreteness, coefficient of variation, etc., which helps to infer the pathological type of the lesion tissue; texture feature parameters are based on the contrast, correlation, energy, and entropy obtained from the grayscale co-occurrence matrix, which carefully depicts the complexity and regularity of the lesion image texture, further enriches the diagnostic information, and allows doctors to have a deeper understanding of the intrinsic properties of the lesions; each feature parameter is calculated and analyzed through formulaic calculation to obtain the corresponding evaluation index, which provides doctors with an intuitive and accurate numerical reference, changing the previous limitations of naked eye observation and fuzzy experience judgment. At the same time, doctors can make more scientific and accurate diagnoses based on these quantitative indicators, combined with other information such as clinical symptoms and medical history, and formulate personalized treatment plans to improve the diagnosis and treatment effect.

[0038] The present invention makes a comprehensive analysis of the morphological evaluation index, density evaluation index and texture evaluation index of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current time period, so that the abnormality level analysis result of the lesion area is more reasonable and reliable, and the uncertainty of traditional manual subjective grading is realized, and a unified and objective standard is provided for disease assessment, which is helpful to standardize the diagnosis and treatment process and improve the homogeneity level of medical services. Doctors can quickly grasp the key points of the disease from the abnormality level and lesion location information given by the system, save time in busy clinical work, quickly judge the severity of the disease, and then decide whether further examination is needed and what treatment measures to take, such as emergency surgery, conservative treatment or close observation, etc., optimize the allocation of medical resources, improve the efficiency of diagnosis and treatment, and enable patients to receive timely and appropriate treatment. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The present invention will be further described below in conjunction with the accompanying drawings.

[0040] Figure 1 It is a system block diagram of the present invention. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] See also Figure 1As shown, the present invention is an image-assisted analysis system for hospital departments based on deep learning, including: an image acquisition and preprocessing module, a lesion area recognition and analysis module, a regional feature parameter acquisition module, a lesion area morphology analysis module, a lesion area density analysis module, a lesion area texture analysis module, an auxiliary evaluation and analysis module, a display terminal and a cloud database. Wherein, the connection mode between the modules is: the image acquisition and preprocessing module is connected to the lesion area recognition and analysis module, the lesion area recognition and analysis module is respectively connected to the regional feature parameter acquisition module and the cloud database, the regional feature parameter acquisition module is respectively connected to the lesion area morphology analysis module, the lesion area density analysis module and the lesion area texture analysis module, the lesion area morphology analysis module, the lesion area density analysis module and the lesion area texture analysis module are connected to the auxiliary evaluation and analysis module, and the auxiliary evaluation and analysis module is connected to the display terminal.

[0043] The image acquisition and preprocessing module is used to acquire medical images of various imaging devices in the hospital department corresponding to the current period, obtain medical images of various imaging devices in the hospital department corresponding to the current period, and perform preprocessing operations on them to obtain preprocessed medical images of the hospital department corresponding to the current period;

[0044] It should be noted that various imaging devices include but are not limited to CT, MRI, X-ray, and ultrasound; preprocessing operations include but are not limited to format conversion, denoising and enhancement processing, grayscale normalization, cropping and scaling; for medical images of different modalities of various imaging devices, a unified NI fTI format conversion method is executed to unify them into a standard format that can be processed by the system, and the conversion operation is performed through the NI fTI format conversion tool. Specifically: the NI fTI format conversion tool can use dcm2n iix, MRIcron; denoising and enhancement processing is to use Gaussian filtering or median filtering algorithm to remove noise interference in medical images; grayscale normalization is to map the grayscale value of medical images to a specific interval [0,1]; cropping and scaling are to focus on key areas and reduce the amount of calculation.

[0045] In a specific embodiment, the present invention collects and preprocesses medical images of various imaging devices in hospital departments corresponding to the current period, thereby breaking the isolated situation of data of different modal influences, allowing various types of influences to be summarized, providing a data basis for subsequent comprehensive and integrated analysis, and avoiding information omissions or analysis obstacles due to equipment differences.

[0046] The lesion area recognition and analysis module is used to perform lesion area recognition and analysis on the pre-processed medical images of the hospital department corresponding to the current period. The specific execution steps are as follows:

[0047] Performing regional division on each preprocessed medical image of the hospital department corresponding to the current period according to a set division method, obtaining each sub-region of each preprocessed medical image of the hospital department corresponding to the current period, and extracting the tissue and organ structure of each sub-region of each preprocessed medical image, obtaining the tissue and organ structure of each sub-region of each preprocessed medical image of the hospital department corresponding to the current period, and matching it with the reference gray value interval range corresponding to the tissue and organ structure of each sub-region stored in the cloud database, obtaining the reference gray value interval range of each sub-region of each preprocessed medical image of the hospital department corresponding to the current period;

[0048] Obtain the grayscale value of each pixel in each sub-region of the preprocessed medical images of the hospital department corresponding to the current period, and compare and analyze it with the reference grayscale value range of each sub-region. If the grayscale value of a pixel in a region is greater than the maximum value of the reference grayscale value range corresponding to the sub-region, then the pixel in the sub-region is marked as a target pixel. If the grayscale value of a pixel in a region is less than the minimum value of the reference grayscale value range corresponding to the sub-region, then the pixel in the sub-region is marked as a marked pixel. In this way, the number of target pixels and the number of marked pixels in each sub-region of the preprocessed medical images of the hospital department corresponding to the current period are statistically obtained, and are marked as i represents the number of each preprocessed medical image, i=1,2,3,....,m, m represents the total number of preprocessed medical image numbers, j represents the number of each sub-region, j=1,2,3,....,n, n represents the total number of sub-region numbers;

[0049] The grayscale values ​​of each target pixel in each sub-region are extracted from the grayscale values ​​of each pixel in each sub-region of the preprocessed medical images of the hospital department corresponding to the current period, and the grayscale values ​​of each target pixel in each sub-region are calculated by difference with the maximum value of the reference grayscale value interval corresponding to each sub-region to obtain the grayscale value difference of each target pixel in each sub-region, and then the summation is performed to obtain the total grayscale value difference of the target pixel in each sub-region of the preprocessed medical images of the hospital department corresponding to the current period, which is recorded as The grayscale values ​​of each marked pixel point in each sub-region are extracted from the grayscale values ​​of each pixel point in each sub-region of the preprocessed medical images of the hospital department corresponding to the current period, and the grayscale value difference of each marked pixel point in each sub-region is obtained by subtracting it from the minimum value of the reference grayscale value interval corresponding to each sub-region, and then the summation is performed to obtain the total grayscale value difference of the marked pixel points in each sub-region of the preprocessed medical images of the hospital department corresponding to the current period, which is recorded as According to the formula Calculate the grayscale matching evaluation index of each sub-region of the preprocessed medical images in the current period corresponding to the hospital department a1, a2, a3, and a4 represent the weight factors corresponding to the set number of target pixels, the number of marked pixels, the total grayscale value difference of target pixels, and the total grayscale value difference of marked pixels, respectively;

[0050] The grayscale matching evaluation index of each sub-region of the preprocessed medical images of the hospital department corresponding to the current time period is compared and analyzed with the set reference grayscale matching evaluation index threshold. When the grayscale matching evaluation index of a region is greater than the set reference grayscale matching evaluation index threshold, the sub-region is judged to be a normal region. When the grayscale matching evaluation index of a region is less than the set reference grayscale matching evaluation index threshold, the sub-region is judged to be a lesion region. In this way, the lesion regions of the preprocessed medical images of the hospital department corresponding to the current time period are obtained through integration.

[0051] In a specific embodiment, the present invention uses a detailed medical image area division, tissue and organ structure extraction and gray value matching process, and no longer relies solely on the doctor's experience and judgment, but accurately identifies the lesion area based on the quantitative gray interval standard, and makes a fine comparison of the gray values ​​of the pixel points, combines the statistical analysis of the target pixel points and the marked pixel points, and comprehensively considers the gray matching evaluation index, which can keenly capture the tiny abnormal gray changes, greatly improve the detection rate of early lesions and tiny lesions, reduce the risk of misdiagnosis and missed diagnosis, and buy precious time for early intervention of the disease. It avoids subjective judgment bias caused by factors such as artificial visual fatigue and experience differences, and ensures the accuracy and stability of diagnosis.

[0052] The regional feature parameter acquisition module is used to extract the lesion regional feature parameters of the pre-processed medical images of the hospital department corresponding to the current period based on the deep learning technology, and obtain the lesion regional feature parameters of the pre-processed medical images of the hospital department corresponding to the current period, wherein the feature parameters are specifically: morphological feature parameters, density feature parameters and texture feature parameters; it should be noted that deep learning mainly refers to the use of neural network models to detect and identify lesion regional features. Deep learning is a branch of machine learning that uses multi-layer neural network models to simulate the human nervous system to achieve data learning and processing. In the medical image lesion regional feature detection system, the detection and identification capabilities of medical image lesion regional features are improved by learning and training a large number of medical image lesion regional features. Generally speaking, the deep learning process includes the following: data preparation, model construction, model training, model evaluation, and model application.

[0053] The lesion area morphology analysis module is used to analyze the morphological characteristic parameters of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period. The specific analysis method is as follows:

[0054] The maximum long-axis distance of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period is obtained from the morphological characteristic parameters of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, and it is used as the long-axis distance of the lesion area. At the same time, the maximum short-axis distance of the lesion area is obtained and used as the short-axis distance of the lesion area. The ratio of the long-axis distance to the short-axis distance of the lesion area is calculated to obtain the aspect ratio of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, which is recorded as AR i ;

[0055] Get the area and perimeter of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, and mark them as A i , P i , according to the formula Calculate the circularity C of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period i ;

[0056] Obtain the lobulation index of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period and mark it as LI i , according to the formula Calculate the morphological evaluation index XP of the lesion area of ​​the target patient's medical image i , b1, b2, b3, b4, and b5 represent the weight factors corresponding to the aspect ratio, area, perimeter, circularity, and lobulation index of the set lesion area, respectively;

[0057] It should be noted that the lobulation index refers to the amount of data on the degree of concavity and convexity of the edge of the lesion area. The larger the lobulation index, the more obvious the lobulation of the lesion area, which is often related to malignant lesions. The closer the circularity of the lesion area is to 1, the closer the lesion area is to a circle, and the deviation from 1 indicates that the lesion area is more irregular in shape, and irregular lesions are more likely to be malignant. The lesion area density analysis module is used to analyze the density characteristic parameters of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period. The specific analysis method is as follows:

[0058] The grayscale values ​​of all pixels in the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period are obtained from the density feature parameters of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, and the mean is calculated to obtain the average grayscale value of all pixels in the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, which is recorded as At the same time, the maximum grayscale value, the minimum grayscale value and the median grayscale value are extracted from the grayscale values ​​of all pixels in the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, and marked as According to the formula

[0059] Calculate the gray value dispersion LD of all pixels in the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period i , c1, c2, c3 represent the set proportional factors respectively;

[0060] The grayscale value of each pixel in the lesion area is extracted from the grayscale values ​​of all pixels in the lesion area of ​​the preprocessed medical images of the hospital department corresponding to the current period, and is recorded as f represents the number of all pixel points in the lesion area, f = 1, 2, 3, ...., g, g represents the total number of all pixel point numbers;

[0061] According to the formula Calculate the gray value variation coefficient HI of all pixels in the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period i ;

[0062] According to the formula Calculate the density assessment index MP of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period i , c4 and c5 represent the weight factors corresponding to the set gray value discreteness and gray value variation coefficient respectively;

[0063] It should be noted that the gray value dispersion LD of all pixels in the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period i The larger the value, the higher the grayscale inhomogeneity inside the lesion area. The grayscale value variation coefficient HI of all pixels in the lesion area of ​​the hospital department corresponding to the preprocessed medical images in the current period i The larger it is, the more uneven the density of the lesion area is.

[0064] The lesion area texture analysis module is used to analyze the texture feature parameters of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period. The specific analysis method is as follows:

[0065] The contrast, correlation, energy and entropy of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period are obtained from the texture feature parameters of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, and they are marked as CON i , COR i 、ASM i 、E i , according to the formula Calculate the texture evaluation index WP of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period i, d1, d2, d3, and d4 represent the weight factors corresponding to the contrast, correlation, energy, and entropy of the set lesion area, respectively;

[0066] It should be noted that the texture feature parameters are obtained based on the gray level co-occurrence matrix method, which specifically includes: selecting a specific gray level distance and direction, determining adjacent pixel pairs according to the gray level distance and direction; counting the number of occurrences of these adjacent pixel pairs at a specified gray level, and constructing a co-occurrence matrix between gray levels; and calculating the texture feature parameters according to the obtained co-occurrence matrix;

[0067] Contrast refers to the difference in grayscale values ​​of adjacent pixels in the lesion area image, and its calculation formula is: N is the gray level, P(s,t) is the element in the gray level co-occurrence matrix, s represents the sth row in the gray level co-occurrence matrix, and t represents the tth column in the gray level co-occurrence matrix. The greater the contrast, the greater the gray level difference between adjacent pixels in the lesion area, and the higher the clarity and sharpness of the texture;

[0068] The correlation reflects the linear dependence of the gray value in the lesion area image in a specific direction, and its calculation formula is: μ s , μ t are the means of the sth row and tth column in the gray-level co-occurrence matrix, σ s , σ t are the standard deviations of the sth row and tth column in the gray-level co-occurrence matrix, respectively. The value of the correlation is between -1 and 1. A value close to 1 indicates that the gray-level values ​​have a strong linear correlation in this direction, a value close to -1 indicates a strong negative correlation, and a value close to 0 indicates a weak correlation.

[0069] Energy, also known as the angular second-order moment, is the sum of the squares of the elements in the gray-level co-occurrence matrix. It reflects the uniformity of the gray-level distribution of the image in the lesion area and the coarseness of the texture. Its calculation formula is: The larger the energy value, the more regular and uniform the texture of the image in the lesion area, and the more stable the lesion area. The smaller the energy value, the more non-uniform the texture of the image in the lesion area, and the more complex and heterogeneous the lesion.

[0070] Entropy refers to the randomness and uncertainty of the texture in the lesion area image, and its calculation formula is: The larger the entropy value, the higher the randomness and uncertainty of the texture in the lesion area, and the more complex and irregular the texture.

[0071] In a specific embodiment, the present invention uses deep learning technology to mine multi-dimensional feature parameters of the morphology, density, and texture of the lesion area of ​​the medical image. These parameters reflect the characteristics of the lesion from different aspects. Morphological feature parameters such as aspect ratio, circularity, and lobulation index can assist in judging the growth mode and invasiveness of the lesion, and provide key clues for distinguishing benign and malignant lesions; density feature parameters accurately reflect the uniformity of the internal structure of the lesion by measuring the gray value discreteness, coefficient of variation, etc., which helps to infer the pathological type of the lesion tissue; texture feature parameters are based on the contrast, correlation, energy, and entropy obtained from the grayscale co-occurrence matrix, which carefully depicts the complexity and regularity of the lesion image texture, further enriches the diagnostic information, and allows doctors to have a deeper understanding of the intrinsic properties of the lesion; each feature parameter is calculated and analyzed by formulaic calculation to obtain the corresponding evaluation index, which provides doctors with an intuitive and accurate numerical reference, changing the previous limitations of naked eye observation and fuzzy experience judgment. At the same time, doctors can make more scientific and accurate diagnoses based on these quantitative indicators, combined with other information such as clinical symptoms and medical history, and formulate personalized treatment plans to improve the diagnosis and treatment effect.

[0072] The auxiliary evaluation and analysis module is used to analyze the abnormality level of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period, and obtain the abnormality level of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period;

[0073] Furthermore, the abnormal levels of the lesion areas of the pre-processed medical images of the hospital departments corresponding to the current period are analyzed. The specific analysis method is as follows:

[0074] According to the formula Calculate the abnormality evaluation coefficient YC of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period i , w1, w2, w3 represent the coefficient factors corresponding to the set morphological evaluation index, density evaluation index and texture evaluation index respectively;

[0075] Matching the abnormality evaluation coefficients of the lesion areas of the preprocessed medical images of the hospital departments corresponding to the current period with the abnormality evaluation coefficient thresholds corresponding to the set abnormality levels, to obtain the abnormality levels of the lesion areas of the preprocessed medical images of the hospital departments corresponding to the current period;

[0076] The analysis result information of each preprocessed medical image of the hospital department corresponding to the current period is constituted by the lesion area location and abnormality degree level of each preprocessed medical image of the hospital department corresponding to the current period.

[0077] The display terminal performs corresponding display operations based on the analysis result information of the pre-processed medical images of the hospital department corresponding to the current period.

[0078] In a specific embodiment, the present invention comprehensively analyzes the morphological evaluation index, density evaluation index and texture evaluation index of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period, so that the abnormal degree level analysis result of the lesion area is more reasonable and reliable, and the uncertainty of the traditional manual subjective grading is realized, and a unified and objective standard is provided for the disease assessment, which is helpful to standardize the diagnosis and treatment process and improve the homogeneity level of medical services. Doctors can quickly grasp the key points of the disease from the abnormal level and lesion location information given by the system, save time in busy clinical work, quickly judge the severity of the disease, and then decide whether further examination is needed and what treatment measures to take, such as emergency surgery, conservative treatment or close observation, etc., optimize the allocation of medical resources, improve the efficiency of diagnosis and treatment, and enable patients to receive timely and appropriate treatment. The above content is only an example and explanation of the structure of the present invention. The technical personnel in the relevant technical field make various modifications or supplements to the specific embodiments described or replace them in a similar way. As long as they do not deviate from the structure of the invention or exceed the scope defined by the claims, they should all belong to the protection scope of the present invention.

Claims

1. An image-assisted analysis system for hospital departments based on deep learning, characterized in that: include: The image acquisition and preprocessing module is used to acquire medical images of various imaging devices in the hospital department corresponding to the current period, obtain medical images of various imaging devices in the hospital department corresponding to the current period, and perform preprocessing operations on them to obtain preprocessed medical images of the hospital department corresponding to the current period; The lesion area recognition and analysis module is used to perform lesion area recognition and analysis on the pre-processed medical images of the hospital department corresponding to the current period, and obtain the lesion area of ​​the pre-processed medical images of the hospital department corresponding to the current period; A regional feature parameter acquisition module is used to extract the lesion region feature parameters of each pre-processed medical image of the hospital department corresponding to the current period based on deep learning technology, and obtain the lesion region feature parameters of each pre-processed medical image of the hospital department corresponding to the current period, wherein the feature parameters are specifically: morphological feature parameters, density feature parameters and texture feature parameters; a lesion region morphological analysis module is used to analyze the morphological feature parameters of the lesion region of each pre-processed medical image of the hospital department corresponding to the current period, and obtain the morphological evaluation index of the lesion region of each pre-processed medical image of the hospital department corresponding to the current period; The lesion area density analysis module is used to analyze the grayscale characteristic parameters of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period, and obtain the density evaluation index of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period; The lesion area texture analysis module is used to analyze the texture feature parameters of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period, and obtain the texture evaluation index of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period; The auxiliary evaluation and analysis module is used to analyze the abnormality level of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period, and obtain the abnormality level of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period; The display terminal performs corresponding display operations based on the analysis result information of the pre-processed medical images of the hospital department corresponding to the current period.

2. The deep learning-based image-assisted analysis system for hospital departments according to claim 1, characterized in that: The lesion area recognition and analysis is performed on the preprocessed medical images of the hospital department corresponding to the current time period, and the specific execution steps are as follows: the preprocessed medical images of the hospital department corresponding to the current time period are divided into regions according to a set division method to obtain the sub-regions of the preprocessed medical images of the hospital department corresponding to the current time period, and the tissue and organ structures of the sub-regions of the preprocessed medical images are extracted therefrom to obtain the tissue and organ structures of the sub-regions of the preprocessed medical images of the hospital department corresponding to the current time period, and the tissue and organ structures are matched with the reference gray value interval range corresponding to the tissue and organ structures of the sub-regions stored in the cloud database to obtain the reference gray value interval range of the sub-regions of the preprocessed medical images of the hospital department corresponding to the current time period; Obtain the grayscale value of each pixel in each sub-region of the preprocessed medical images of the hospital department corresponding to the current period, and compare and analyze it with the reference grayscale value range of each sub-region. If the grayscale value of a pixel in a region is greater than the maximum value of the reference grayscale value range corresponding to the sub-region, then the pixel in the sub-region is marked as a target pixel. If the grayscale value of a pixel in a region is less than the minimum value of the reference grayscale value range corresponding to the sub-region, then the pixel in the sub-region is marked as a marked pixel. In this way, the number of target pixels and the number of marked pixels in each sub-region of the preprocessed medical images of the hospital department corresponding to the current period are statistically obtained, and are marked as i represents the number of each preprocessed medical image, i=1,2,3,....,m, m represents the total number of preprocessed medical image numbers, j represents the number of each sub-region, j=1,2,3,....,n, n represents the total number of sub-region numbers; The grayscale values ​​of each target pixel in each sub-region are extracted from the grayscale values ​​of each pixel in each sub-region of the preprocessed medical images of the hospital department corresponding to the current period, and the grayscale values ​​of each target pixel in each sub-region are calculated by difference with the maximum value of the reference grayscale value interval corresponding to each sub-region to obtain the grayscale value difference of each target pixel in each sub-region, and then the summation is performed to obtain the total grayscale value difference of the target pixel in each sub-region of the preprocessed medical images of the hospital department corresponding to the current period, which is recorded as The grayscale values ​​of each marked pixel point in each sub-region are extracted from the grayscale values ​​of each pixel point in each sub-region of the preprocessed medical images of the hospital department corresponding to the current period, and the grayscale value difference of each marked pixel point in each sub-region is obtained by subtracting it from the minimum value of the reference grayscale value interval corresponding to each sub-region, and then the summation is performed to obtain the total grayscale value difference of the marked pixel points in each sub-region of the preprocessed medical images of the hospital department corresponding to the current period, which is recorded as According to the formula Calculate the grayscale matching evaluation index of each sub-region of each preprocessed medical image of the hospital department corresponding to the current period a1, a2, a3, and a4 represent the weight factors corresponding to the set number of target pixels, the number of marked pixels, the total grayscale value difference of target pixels, and the total grayscale value difference of marked pixels, respectively; The grayscale matching evaluation index of each sub-region of the preprocessed medical images of the hospital department corresponding to the current time period is compared and analyzed with the set reference grayscale matching evaluation index threshold. When the grayscale matching evaluation index of a region is greater than the set reference grayscale matching evaluation index threshold, the sub-region is judged to be a normal region. When the grayscale matching evaluation index of a region is less than the set reference grayscale matching evaluation index threshold, the sub-region is judged to be a lesion region. In this way, the lesion regions of the preprocessed medical images of the hospital department corresponding to the current time period are obtained through integration.

3. The deep learning-based image-assisted analysis system for hospital departments according to claim 1, characterized in that: The morphological characteristic parameters of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period are analyzed in the following specific analysis method: The maximum long-axis distance of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period is obtained from the morphological characteristic parameters of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, and it is used as the long-axis distance of the lesion area. At the same time, the maximum short-axis distance of the lesion area is obtained and used as the short-axis distance of the lesion area. The ratio of the long-axis distance to the short-axis distance of the lesion area is calculated to obtain the aspect ratio of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, which is recorded as AR i ; Get the area and perimeter of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, and mark them as A i , P i , according to the formula Calculate the circularity C of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period i ; Obtain the lobulation index of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period and mark it as LI i , according to the formula Calculate the morphological evaluation index XP of the lesion area of ​​each pre-processed medical image in the hospital department corresponding to the current period i , b1, b2, b3, b4, and b5 represent the weight factors corresponding to the aspect ratio, area, perimeter, circularity, and lobulation index of the set lesion area, respectively.

4. The deep learning-based image-assisted analysis system for hospital departments according to claim 1, characterized in that: The grayscale characteristic parameters of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period are analyzed, and the specific analysis method is as follows: The grayscale values ​​of all pixels in the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period are obtained from the density feature parameters of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, and the mean is calculated to obtain the average grayscale value of all pixels in the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, which is recorded as At the same time, the maximum grayscale value, the minimum grayscale value and the median grayscale value are extracted from the grayscale values ​​of all pixels in the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, and marked as According to the formula Calculate the gray value dispersion LD of all pixels in the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period i , c1, c2, c3 represent the set proportional factors respectively; The grayscale value of each pixel in the lesion area is extracted from the grayscale values ​​of all pixels in the lesion area of ​​the preprocessed medical images of the hospital department corresponding to the current period, and is recorded as f represents the number of all pixels in the lesion area. f=1,2,3,....,g, where g represents the total number of all pixel numbers; According to the formula Calculate the gray value variation coefficient HI of all pixels in the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period i ; According to the formula Calculate the density assessment index MP of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period i , c4 and c5 represent the weight factors corresponding to the set gray value discreteness and gray value variation coefficient respectively.

5. The deep learning-based image-assisted analysis system for hospital departments according to claim 1, characterized in that: The texture feature parameters of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period are analyzed, and the specific analysis method is as follows: The contrast, correlation, energy and entropy of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period are obtained from the texture feature parameters of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period, and they are marked as CON i , COR i 、ASM i 、E i , according to the formula Calculate the texture evaluation index WP of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period i , d1, d2, d3, and d4 represent the weight factors corresponding to the contrast, correlation, energy, and entropy of the set lesion area, respectively.

6. The deep learning-based image-assisted analysis system for hospital departments according to claim 1, characterized in that: The abnormal level of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period is analyzed, and the specific analysis method is as follows: Comprehensively analyze the morphological evaluation index, density evaluation index and texture evaluation index of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period, and obtain the abnormality evaluation coefficient of the lesion area of ​​each pre-processed medical image of the hospital department corresponding to the current period; Matching the abnormality evaluation coefficient of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period with the abnormality evaluation coefficient threshold corresponding to each set abnormality level, to obtain the abnormality level of the lesion area of ​​each preprocessed medical image of the hospital department corresponding to the current period; The analysis result information of each preprocessed medical image of the hospital department corresponding to the current period is constituted by the lesion area location and abnormality degree level of each preprocessed medical image of the hospital department corresponding to the current period.

Images