Deep learning-based children's pneumonia lung image analysis and evaluation system and method

Abstract

The present invention discloses a system and method for analyzing and evaluating children's pneumonia lung images based on deep learning, which relates to the technical field of children's pneumonia analysis and evaluation, and aims to solve the technical problem that the accuracy of manual diagnosis in the prior art is difficult to guarantee. It includes: a lung image detection module for taking lung CT images of the detected children with pneumonia patients and transmitting them to the data calculation module; a data calculation module connected to the lung image detection module for receiving the images of the lung image detection module and calculating data such as alveolar density, clarity, quantity, and echo intensity; an image comparison module for comparing and analyzing the data obtained by the data calculation module to judge the treatment condition of children with pneumonia patients; an evaluation module for evaluating whether the pneumonia of children with pneumonia patients deteriorates or improves by using a comprehensive evaluation algorithm according to the results of the image comparison module; and a treatment processing module. The present invention has the advantages of accurately identifying and judging children's lung images and improving the accuracy of disease condition evaluation.

Description

Technical Field

[0001] The present invention relates to the technical field of analysis and evaluation of childhood pneumonia, and more specifically, to a system and method for analyzing and evaluating childhood pneumonia lung images based on deep learning. Background Art

[0002] In the field of diagnosis of childhood pneumonia, the technology of judging the condition based on medical images is crucial. At present, the existing technology mainly relies on doctors to manually observe and analyze childhood lung images (such as X-ray and CT images). Doctors rely on their professional knowledge and experience to identify the lesion characteristics in the images, and then judge whether a child has pneumonia and the severity of the condition. For example, doctors will observe typical features such as whether the lung texture in the image is thickened and whether there are patchy shadows to judge the condition. However, this manual analysis method has many deficiencies. In terms of condition judgment, manual interpretation is greatly affected by factors such as doctors' personal experience and fatigue level. There may be differences in the diagnostic results among different doctors, resulting in difficulty in ensuring the accuracy of diagnosis. Especially when faced with a large amount of image data, doctors are prone to fatigue, and the risks of missed diagnosis and misdiagnosis increase significantly, seriously affecting the efficiency and quality of childhood pneumonia diagnosis and possibly delaying the treatment time of children.

[0003] In view of this, we propose a system and method for analyzing and evaluating childhood pneumonia lung images based on deep learning. Summary of the Invention

[0004] The purpose of the present invention is to provide a system and method for analyzing and evaluating childhood pneumonia lung images based on deep learning to solve the technical problem that it is difficult to ensure the accuracy of manual diagnosis in the prior art.

[0005] To solve the above technical problem, the present invention provides the following technical solution: A system for analyzing and evaluating childhood pneumonia lung images based on deep learning, comprising:

[0006] A lung image detection module, configured to take a lung CT image of a detected childhood pneumonia patient and send it to a data calculation module;

[0007] A data calculation module, connected to the lung image detection module, configured to receive the image from the lung image detection module and calculate data such as alveolar density, clarity, quantity, and echo intensity;

[0008] An image comparison module, configured to compare and analyze the data obtained by the data calculation module to judge the treatment condition of a childhood pneumonia patient;

[0009] An evaluation module, configured to evaluate whether the pneumonia of a childhood pneumonia patient has deteriorated or improved by using a comprehensive evaluation algorithm according to the result of the image comparison module;

[0010] The treatment processing module analyzes the lung changes and the clarity of the lung shadow based on the evaluation results of the evaluation module, and uses a decision algorithm to generate treatment suggestions;

[0011] The data recording module stores the results of data calculation and image comparison, as well as the data generated by the evaluation module and the treatment processing module;

[0012] The terminal device analyzes and summarizes the data of the data recording module, generates the patient's treatment data and change curves, and assists the doctor in treatment judgment and recording.

[0013] Preferably, the method for calculating the alveolar density, clarity, quantity, and echo intensity data in the data calculation module is as follows:

[0014] After the data calculation module receives the image transmitted by the lung shadow detection module, it preprocesses the image using an image analysis algorithm based on the U-Net deep learning network to obtain the preprocessed image , and then uses the trained U-Net network to locate the alveolar regions, mark each alveolar space, where is the convolution kernel weight, is the bias term, is the activation function, is the pooling window;

[0015] The method for calculating the alveolar quantity by the data calculation module is as follows:

[0016] After completing the alveolar region location, analyze the marked alveolar spaces. Let the total number of marked alveolar spaces be , and by counting the alveoli in each alveolar space, obtain the number of alveoli in the th alveolar space, then the total number of alveoli in the entire lung image;

[0017] The method for calculating the alveolar density by the data calculation module is as follows:

[0018] Record the number of alveoli in each alveolar space and the area occupied by this alveolar space in the lung image. The total area of the lung image is , and through training with a large number of sample data, obtain the coefficient . The calculation formula for the alveolar density is:

[0019] .

[0020] Preferably, the method for calculating the alveolar clarity by the data calculation module is as follows:

[0021] Use the image texture extraction algorithm to obtain lung texture information from the preprocessed image Let the average gray value in the -th alveolar space be , the number of alveoli in this alveolar space be , and the standard deviation of its gray value be , and define the fuzziness index;

[0022] ;

[0023] By weighted averaging the fuzziness indices of all alveolar spaces, obtain the alveolar fuzziness of the entire lung image ;

[0024] ;

[0025] Among them, is the weight of the -th alveolar space. Judge the alveolar clarity according to the fuzziness . The lower the fuzziness, the higher the alveolar clarity;

[0026] The calculation method of the data calculation module for the alveolar echo intensity is as follows:

[0027] Let the alveolar density detected this time be , the alveolar density detected last time be , the alveolar clarity detected this time be , the alveolar clarity detected last time be , define the density change rate , the clarity change rate , and the judgment logic of the echo intensity is:

[0028] When and , ;

[0029] When and , .

[0030] Preferably, the method for the image comparison module to compare and analyze the data obtained by the data calculation module is:

[0031] Let the current data vector obtained by the data calculation module be , where represents alveolar density, clarity, quantity, and echo intensity, and the historical data vector set is , and use the cosine similarity and the Mahalanobis distance Make a comprehensive judgment;

[0032] ;

[0033] ;

[0034] Among them, represents the covariance matrix of the data. By setting the similarity threshold and the Mahalanobis distance threshold , when and , it is considered that the current data and the historical data have similar treatment conditions;

[0035] The method for the image comparison module to judge the treatment condition of children with pneumonia is:

[0036] Set weights for each index , and , calculate the comprehensive distance between the current data vector and the historical optimal data vector , and the formula is:

[0037] ;

[0038] Among them, is the th element in , and judge the treatment condition according to the comprehensive distance ;

[0039] When , judge that the treatment condition is getting better. Among them, is the pre-set improvement threshold;

[0040] When , judge that the treatment condition is deteriorating. Among them, is the pre-set deterioration threshold;

[0041] When , judge that the treatment condition is stable.

[0042] Preferably, the method for the evaluation module to evaluate whether the pneumonia of children with pneumonia is deteriorating or getting better is:

[0043] Establish an evaluation index system;

[0044] Based on the judgment result of the image comparison module, comprehensively consider the alveolar density , clarity , quantity , echo intensity And the comprehensive distance of the image comparison module , let these indicators and The weights of are , , , , , and ;

[0045] First, perform normalization processing on each indicator. Let , , , Be the normalization functions of alveolar density, clarity, quantity, and echo intensity respectively, and map them to Interval. For the comprehensive distance Use the normalization function ;

[0046] ;

[0047] Among them, And Are the maximum and minimum values of the comprehensive distance in the historical data respectively;

[0048] Define the evaluation score ;

[0049] According to the evaluation score Compare with the preset deterioration threshold And the improvement threshold To judge the deterioration, improvement or stable condition of pneumonia;

[0050] According to the evaluation score Further divide the risk level. Let the risk level be divided into For low - risk, For medium - risk, For high - risk, and the corresponding score intervals are , , .

[0051] Preferably, the method for the treatment processing module to generate treatment suggestions using the decision algorithm is:

[0052] Combining the evaluation results and risk levels of the evaluation module, deeply mining the clinical cases and treatment plans in the medical database. Let the set of cases stored in the medical database be , and use the nearest neighbor algorithm to calculate the distance Between the current evaluation score And the evaluation scores of each case in the database For each case, the recommended treatment plan is determined by weighted voting. ;

[0053] ;

[0054] ;

[0055] Among them, is the set of the most recent cases, is an indicator function, represents the treatment plan variable, represents the treatment plan of the th case in the medical database;

[0056] According to different risk levels and the changes in the condition, emergency treatment measures are preferentially recommended for highly risk patients, suggestions for adjusting the existing treatment plan are provided for moderately risk patients, and a maintenance treatment or tapering treatment plan is given for low-risk patients.

[0057] Preferably, the data recording module stores data on alveolar density, clarity, quantity, and echo intensity, as well as the corresponding imaging images, details the patient's treatment situation in a structured table form, and simultaneously stores the evaluation results of the evaluation module, the risk level, and the treatment suggestions generated by the treatment processing module.

[0058] A method for analyzing and evaluating children's pneumonia lung images based on deep learning includes the following steps:

[0059] S1. Image acquisition;

[0060] Take lung images through the lung image detection module;

[0061] S2. Image processing and data calculation;

[0062] Transmit the acquired images to the data calculation module and calculate the data on alveolar density, clarity, quantity, and echo intensity;

[0063] S3. Index analysis and condition judgment;

[0064] Receive the data on alveolar density, clarity, quantity, and echo intensity, and use the cosine similarity, Mahalanobis distance, and comprehensive distance algorithms to judge the development of the condition;

[0065] S4. Condition evaluation and risk grading;

[0066] According to the results of the image comparison module, calculate the evaluation score, evaluate the deterioration or improvement of pneumonia, and determine the risk level;

[0067] S5. Image evaluation and treatment decision-making;

[0068] According to the results of the evaluation module, treatment suggestions are generated using a decision-making algorithm.

[0069] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0070] 1. By constructing a deep learning-based children's pneumonia lung image analysis and evaluation system and method, the present invention utilizes the powerful image feature extraction and analysis capabilities of deep learning to accurately identify and judge lung images, can accurately extract lesion features, avoids the subjectivity and individual differences of manual interpretation, greatly improves the accuracy of disease judgment, and effectively solves the problem that it is difficult to guarantee the accuracy of artificial diagnosis in the prior art.

[0071] 2. Compared with manual interpretation, the present invention greatly improves the diagnostic efficiency. Analyzing and evaluating a large number of images in a short time can provide diagnostic suggestions for doctors in a timely manner, reduce the waiting time of children for diagnostic results, and further solve the problem of low efficiency of artificial diagnosis in the face of a large amount of image data.

[0072] 3. The present invention can perform quantitative analysis on the lesion features in lung images, not only can judge whether pneumonia is present, but also can more accurately evaluate the severity of the condition, providing more detailed and accurate basis for doctors to formulate personalized treatment plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 is a schematic diagram of the system framework of the present invention;

[0074] Figure 2 is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0075] To facilitate the understanding of the technical solution of the present invention by those skilled in the art, the technical solution of the present invention will be further described below in conjunction with the accompanying drawings of the specification.

[0076] Example 1, as Figure 1 shown, a deep learning-based children's pneumonia lung image analysis and evaluation system includes:

[0077] A lung image detection module, used to take lung CT images of the detected children with pneumonia patients and send them to the data calculation module;

[0078] A data calculation module, connected to the lung image detection module, used to receive the images from the lung image detection module and calculate data such as alveolar density, clarity, quantity, and echo intensity;

[0079] An image comparison module, used to compare and analyze the data obtained by the data calculation module to judge the treatment situation of children with pneumonia patients;

[0080] An evaluation module, according to the results of the image comparison module, uses a comprehensive evaluation algorithm to evaluate whether the pneumonia of a child pneumonia patient is deteriorating or improving;

[0081] A treatment processing module, based on the evaluation results of the evaluation module, analyzes the lung changes and the clarity of the lung shadow, and uses a decision algorithm to generate treatment suggestions;

[0082] A data recording module stores the results of data calculation and image comparison, as well as the data generated by the evaluation module and the treatment processing module;

[0083] A terminal device analyzes and summarizes the data of the data recording module, generates patient treatment data and change curves, and assists doctors in treatment judgment and recording.

[0084] In an embodiment of the present invention, the method for calculating alveolar density, clarity, quantity, and echo intensity data in the data calculation module is as follows:

[0085] After the data calculation module receives the image transmitted by the lung shadow detection module, it preprocesses the image using an image analysis algorithm based on the U-Net deep learning network, through convolutional operations 、max pooling operations and interpolation upsampling to obtain the preprocessed image ;

[0086] During the training process, the model continuously adjusts these parameters according to the feedback of the loss function to minimize the difference between the prediction result and the true label;

[0087] Taking the mean squared error loss function as an example, for a training set with samples, the loss function is , where, is the true label, is the model prediction value. By calculating the gradients of the loss function with respect to and , optimization algorithms such as gradient descent method are used to update them, and the update formula is , ;

[0088] The Sigmoid function is adopted, ;

[0089] In the max pooling operation, the pooling window is used to determine the area size for pooling on the input feature map, and its size is 2×2;

[0090] After that, the trained U-Net network is used to locate the alveolar regions and mark each alveolar gap, where, is the convolution kernel weight, is the updated convolution kernel weight, is the bias term, is the updated bias term, is the learning rate, represents the loss function with respect to the convolution kernel weight gradient, represents the loss function with respect to the bias term gradient, is the activation function, represents the output value of the Sigmoid activation function, is the input of the activation function, is the natural constant, is the pooling window;

[0091] The calculation method of the data calculation module for the number of alveoli is as follows:

[0092] After the alveolar region is located, the marked alveolar spaces are analyzed. Let the total number of marked alveolar spaces be , by counting the alveoli in each alveolar space, the number of alveoli in the -th alveolar space is obtained, then the total number of alveoli in the entire lung image;

[0093] The calculation method of the data calculation module for the alveolar density is as follows:

[0094] Record the number of alveoli in each alveolar space and the area occupied by this alveolar space in the lung image. The total area of the lung image is . After training with a large number of sample data, the coefficient is obtained. The calculation formula of the alveolar density is:

[0095] .

[0096] In the embodiment of the present invention, the calculation method of the data calculation module for the alveolar clarity is as follows:

[0097] Use the image texture extraction algorithm to obtain the lung texture information from the preprocessed image . Let the average gray value in the -th alveolar space be , the number of alveoli in this alveolar space be , and the standard deviation of its gray value be ;

[0098] ;

[0099] By performing a weighted average on the blurriness degree indexes of all alveolar spaces, the alveolar blurriness degree of the entire lung image is obtained ;

[0100] ;

[0101] Among them, is the weight of the th alveolar space. The alveolar clarity is judged according to the blurriness degree. The lower the blurriness degree, the higher the alveolar clarity;

[0102] The calculation method of the alveolar echo intensity by the said data calculation module is as follows:

[0103] Let the alveolar density detected this time be , the alveolar density detected last time be , the alveolar clarity detected this time be , the alveolar clarity detected last time be , define the density change rate , the clarity change rate , and the judgment logic of the echo intensity is as follows:

[0104] When and , ;

[0105] When and , .

[0106] In the embodiment of the present invention, the method for the image comparison module to compare and analyze the data obtained by the data calculation module is as follows:

[0107] Let the current data vector obtained by the data calculation module be , where represents alveolar density, clarity, quantity, and echo intensity, and the historical data vector set is . The cosine similarity and the Mahalanobis distance are used for comprehensive judgment;

[0108] ;

[0109] ;

[0110] Among them, Represents the covariance matrix of the data. By setting the similarity threshold and the Mahalanobis distance threshold , when and , it is considered that the current data and the historical data have similar treatment conditions;

[0111] The method for the image comparison module to judge the treatment condition of children with pneumonia is as follows:

[0112] Set weights for each index , and , calculate the comprehensive distance between the current data vector and the historical optimal data vector (i.e., the historical data vector corresponding to the best treatment effect), and the formula is:

[0113] ;

[0114] Among them, is the th element in . Judge the treatment condition according to the comprehensive distance ;

[0115] When , judge that the treatment condition is improving. Among them, is the pre-set improvement threshold;

[0116] When , judge that the treatment condition is deteriorating. Among them, is the pre-set deterioration threshold;

[0117] When , judge that the treatment condition is stable.

[0118] In the embodiments of the present invention, the method for the evaluation module to evaluate whether the pneumonia of children with pneumonia is deteriorating or improving is as follows:

[0119] Establish an evaluation index system;

[0120] Based on the judgment result of the image comparison module, comprehensively consider the alveolar density , clarity , quantity , echo intensity and the comprehensive distance of the image comparison module. Let the weights of these indexes and be , , , , , and ;

[0121] First, normalize each index. Let , , , be the normalization functions for alveolar density, clarity, quantity, and echo intensity respectively, and map them to the interval. For the comprehensive distance , use the normalization function ;

[0122] ;

[0123] Among them, and are the maximum and minimum values of the comprehensive distance in historical data respectively;

[0124] Define the evaluation score ;

[0125] According to the evaluation score and the preset deterioration threshold and improvement threshold for comparison to judge the deterioration, improvement, or stability of pneumonia;

[0126] According to the evaluation score further divide the risk level. Let the risk levels be for low risk, for medium risk, for high risk, and the corresponding score intervals are , , .

[0127] In the embodiment of the present invention, the method for the treatment processing module to generate treatment suggestions using a decision algorithm is as follows:

[0128] Combining the evaluation results and risk levels of the evaluation module, deeply mine the clinical cases and treatment plans in the medical database. Let the set of cases stored in the medical database be , and use the nearest neighbor algorithm to calculate the distance between the current evaluation score and the evaluation scores of each case in the database, and select the nearest cases, and determine the recommended treatment plan by weighted voting;

[0129] ;

[0130] ;

[0131] Among them, is the set of the most recent case collections, is an indicator function, represents the treatment plan variable, represents the treatment plan of the th case in the medical database;

[0132] According to different risk levels and the changes in the condition, emergency treatment measures are preferentially recommended for high-risk patients, suggestions for adjusting the existing treatment plan are provided for medium-risk patients, and maintenance treatment or stepwise dose reduction treatment plans are given for low-risk patients.

[0133] In the embodiment of the present invention, the data recording module stores the alveolar density, clarity, quantity and echo intensity data and the corresponding imaging images, details the treatment conditions of the patient in the form of a structured table, and simultaneously stores the evaluation results of the evaluation module, the risk level and the treatment suggestions generated by the treatment processing module.

[0134] Embodiment 2, as Figure 2 shown, a method for analyzing and evaluating children's pneumonia lung images based on deep learning includes the following steps:

[0135] S1, Image acquisition;

[0136] Take lung images through the lung image detection module;

[0137] S2, Image processing and data calculation;

[0138] Transmit the acquired images to the data calculation module, and calculate the alveolar density, clarity, quantity and echo intensity data;

[0139] S3, Index analysis and condition judgment;

[0140] Receive the alveolar density, clarity, quantity and echo intensity data, and use the cosine similarity, Mahalanobis distance and comprehensive distance algorithm to judge the development of the condition;

[0141] S4, Condition evaluation and risk grading;

[0142] According to the results of the image comparison module, calculate the evaluation score, evaluate the deterioration or improvement of pneumonia and determine the risk level;

[0143] S5, Image evaluation and treatment decision-making;

[0144] According to the results of the evaluation module, use the decision algorithm to generate treatment suggestions.

[0145] The embodiments disclosed in the present invention are preferred embodiments, but not limited thereto. Those of ordinary skill in the art can easily understand the spirit of the present invention based on the above embodiments and make different extensions and changes. As long as they do not depart from the spirit of the present invention, they are within the protection scope of the present invention.

Claims

1. A deep learning-based children's pneumonia lung image analysis and evaluation system, characterized in that Including: A lung image detection module, which is used to take lung CT images of the detected children with pneumonia patients and transmit them to the data calculation module; A data calculation module, connected to the lung image detection module, which is used to receive the images of the lung image detection module and calculate the data of alveolar density, clarity, quantity and echo intensity; An image comparison module, which compares and analyzes the data obtained by the data calculation module to judge the treatment condition of children with pneumonia patients; An evaluation module, according to the result of the image comparison module, uses a comprehensive evaluation algorithm to evaluate whether the pneumonia of children with pneumonia patients deteriorates or improves; A treatment processing module, based on the evaluation result of the evaluation module, analyzes the lung changes and lung image clarity, and uses a decision algorithm to generate treatment suggestions; A data recording module, which stores the results of data calculation and image comparison, as well as the data generated by the evaluation module and the treatment processing module; The method for the image comparison module to compare and analyze the data obtained by the data calculation module is: Let the current data vector obtained by the data calculation module be where x i represents alveolar density, clarity, quantity, and echo intensity, and the set of historical data vectors is Using cosine similarity and Mahalanobis distance to make a comprehensive judgment; Among them, ∑ represents the covariance matrix of the data. By setting the similarity threshold S th and the Mahalanobis distance threshold D th , when and , it is considered that the current data and the historical data have similar treatment situations; The method for the image comparison module to judge the treatment condition of children with pneumonia patients is: For each indicator x i Set the weight d i , and Calculate the comprehensive distance D between the current data vector and the historical optimal data vector com , and the formula is: Among them, is the i-th element in, and judge the treatment situation according to the comprehensive distance D com ; When D com < is less than α, it is determined that the treatment condition is improved, where α is a pre-set improvement threshold; When D com > β, it is determined that the treatment condition has deteriorated, where β is a preset deterioration threshold; When α ≤ D com ≤ β, it is determined that the treatment condition is stable.

2. The child pneumonia lung image analysis and evaluation system based on deep learning according to claim 1, characterized in that, The method for the data calculation module to calculate the data of alveolar density, clarity, quantity and echo intensity is: After the data calculation module receives the images transmitted by the lung shadow detection module, it uses an image analysis algorithm based on the U-Net deep learning network to preprocess the images, obtaining the preprocessed image I pre , and then uses the trained U-Net network to process I pre for alveolar region localization, marking out each alveolar space; The method for the data calculation module to calculate the number of alveoli is: After completing the alveolar region localization, analyze the marked alveolar spaces. Let the total number of marked alveolar spaces be m. By counting the alveoli in each alveolar space, the number of alveoli n in the i-th alveolar space is obtained i (i = 1, 2,..., m), then the total number of alveoli in the entire lung image The method for the data calculation module to calculate the alveolar density is: Record the number of alveoli n in each alveolar space i and the area S occupied by the alveolar space in the lung image i , the total area of the lung image is S total , through training with a large amount of sample data, the coefficient k is obtained, and the calculation formula for the alveolar density D is:

3. The pediatric pneumonia lung image analysis and evaluation system based on deep learning according to claim 2, characterized in that, The method for the data calculation module to calculate the alveolar clarity is: The lung texture information is obtained from the preprocessed image I by using the image texture extraction algorithm pre Let the average gray value in the j-th alveolar space be G j , the number of alveoli in this alveolar space be n j , and the standard deviation of its gray value be σ j . Define the fuzziness index F j ; By weighted averaging the blurring degree indexes of all alveolar spaces, the alveolar blurring degree F of the entire lung image is obtained; Among them, w i is the weight of the j-th alveolar space. The clarity of the alveoli is judged according to the degree of fuzziness F. The lower the degree of fuzziness, the higher the clarity of the alveoli; The method for the data calculation module to calculate the alveolar echo intensity is: Let the alveolar density detected this time be D new , and the alveolar density detected last time be D old , and the alveolar clarity detected this time be C new , and the alveolar clarity detected last time be C old , define the density change rate clarity change rate The judgment logic of the echo intensity E is as follows: When ΔD < h and ΔC > h, E = high; When ΔD ≥ 0 and ΔC ≤ 0, E = low.

4. The child pneumonia lung image analysis and evaluation system based on deep learning according to claim 3, characterized in that The method for the evaluation module to evaluate whether the pneumonia of children with pneumonia patients deteriorates or improves is: Establish an evaluation index system; Based on the judgment result of the image comparison module, comprehensively consider the alveolar density D, clarity C, quantity N, echo intensity E, and the comprehensive distance D of the image comparison module com , assuming that the weights of these indicators and D com are w D , w C , w N , w E , and First, normalize each index. Let f D (D), f C (C), f N (N), f E (E) be the normalization functions for alveolar density, clarity, quantity, and echo intensity respectively, mapping them to the interval [0, 1]. For the comprehensive distance D com Use the normalization function where max(D com ) and min(D com ) are the maximum and minimum values of the comprehensive distance in the historical data respectively; Define the evaluation score According to the evaluation score S eval compare with the preset deterioration threshold S deteriorate and the improvement threshold S improve to determine whether the pneumonia has deteriorated, improved, or the condition is stable; According to the evaluation score S eval Further divide the risk levels. Let the risk levels be L1 for low risk, L2 for medium risk, and L3 for high risk. The corresponding score intervals are respectively [S improve , 1], [S deteriorate , S improve , [h, S deteriorate .

5. The child pneumonia lung image analysis and evaluation system based on deep learning according to claim 4, characterized in that, The method for the treatment processing module to generate treatment suggestions using a decision algorithm is: Combined with the evaluation results and risk levels of the evaluation module, deeply mine the clinical cases and treatment plans in the medical database. Let the set of cases stored in the medical database be Use the nearest neighbor algorithm to calculate the current evaluation score S eval The distances from the evaluation scores of each case in the database Select the K nearest cases and determine the recommended treatment plan T by weighted voting recommend ; Among them, N K is the set of the most recent K cases, δ(T - T k ) is the indicator function, T represents the treatment plan variable, and T k represents the treatment plan of the k-th case in the medical database; According to different risk levels and the changes in the condition, give priority to recommending emergency treatment measures for high-risk patients, provide suggestions for adjusting the existing treatment plan for medium-risk patients, and give a treatment plan for maintaining treatment or gradually reducing the dose for low-risk patients.

6. The pediatric pneumonia lung image analysis and evaluation system based on deep learning according to claim 5, characterized in that, The data recording module stores the data of alveolar density, clarity, quantity and echo intensity and the corresponding image, details the treatment condition of the patient in a structured table form, and stores the evaluation result of the evaluation module, the risk level and the treatment suggestions generated by the treatment processing module at the same time.

7. An evaluation method applied to the child pneumonia lung image analysis and evaluation system based on deep learning as described in claim 6, characterized in that, Including the following steps: S1. Image acquisition; Take lung images through the lung image detection module; S2. Image processing and data calculation; Transmit the collected images to the data calculation module and calculate the data of alveolar density, clarity, quantity and echo intensity; S3. Index analysis and condition judgment; Receive the data of alveolar density, clarity, quantity and echo intensity, and use the cosine similarity, Mahalanobis distance and comprehensive distance algorithm to judge the development of the condition; S4. Condition evaluation and risk classification; According to the result of the image comparison module, calculate the evaluation score, evaluate the deterioration or improvement of pneumonia and determine the risk level; S5. Image evaluation and treatment decision; According to the result of the evaluation module, use a decision algorithm to generate treatment suggestions.

Images