Lung segment function evaluation method based on basin characteristics

Abstract

The present invention relates to the technical field of medical image processing. More specifically, it relates to a method for evaluating lung segment function based on watershed characteristics, including: obtaining medical image data of the lungs; constructing digital models of the pulmonary vascular network and the airway network based on the medical image data; calculating the network characteristics of the pulmonary vascular network and the airway network based on the digital models; generating a blood supply function score and a ventilation function score according to the network characteristics; calculating a comprehensive function score based on the blood supply function score and the ventilation function score; and outputting the comprehensive function score as the evaluation result of the lung segment function. The method of the present invention realizes the multi-dimensional evaluation of the lung segment function by comprehensively analyzing the characteristics of the pulmonary vascular network and the airway network, combining advanced network analysis and machine learning techniques, constructing digital models of the pulmonary blood vessels and airways, and then comprehensively characterizing the characteristics of the lung structure by calculating a series of network characteristics, such as centrality, efficiency, and density.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and more specifically, to a method for evaluating lung segment function based on watershed characteristics. Background Art

[0002] With the continuous development of medical imaging technology, the methods for evaluating lung function have also been continuously improved. Traditional lung function evaluations mainly rely on lung function tests, such as indicators like vital capacity and forced expiratory volume in one second. However, these methods often can only provide information on overall lung function and cannot accurately evaluate the function of specific lung segments.

[0003] In recent years, methods for evaluating lung segment function based on medical images have gradually become a research hotspot. These methods usually utilize imaging means such as CT or MRI, combined with image processing techniques, to analyze the lung structure. For example, some studies evaluate lung ventilation function by analyzing the density of lung parenchyma in CT images, or evaluate lung blood perfusion by contrast-enhanced CT. These methods have improved the accuracy of lung segment function evaluation to a certain extent, but there are still some limitations.

[0004] Firstly, existing methods often only focus on certain aspects of the lungs, such as only considering the density of lung parenchyma or only analyzing vascular structures, and it is difficult to comprehensively reflect the functional status of lung segments. Secondly, these methods usually use simple threshold segmentation or morphological analysis and fail to fully utilize the rich information contained in medical images. Moreover, most existing methods only perform static analysis and it is difficult to reflect the dynamic change characteristics of lung function.

[0005] In addition, existing methods for evaluating lung segment function also have deficiencies in prognosis prediction. Most methods only provide an assessment of the current functional status and cannot effectively predict the future prognosis of patients. This limitation makes it lack a reliable reference basis for clinicians when formulating long-term treatment plans.

[0006] Facing these problems, there is an urgent need for new methods that can comprehensively and accurately evaluate lung segment function and can perform prognosis prediction. An ideal method should be able to comprehensively consider various characteristics of the lungs, fully utilize medical image information, and combine advanced data analysis techniques to achieve dynamic evaluation and prediction of lung segment function. Summary of the Invention

[0007] The present invention is precisely proposed in view of the above technical problems. The present invention provides a method for evaluating lung segment function based on watershed characteristics, aiming to achieve comprehensive and accurate evaluation of lung segment function and provide reliable prognosis prediction.

[0008] To solve the above technical problems, the present invention adopts the following technical solutions:

[0009] A method for evaluating lung segment function based on watershed characteristics, comprising:

[0010] Acquisition steps, including:

[0011] Acquire medical image data of the lungs;

[0012] Processing steps, including:

[0013] Based on the medical image data, construct digital models of the pulmonary vascular network and the airway network;

[0014] Based on the digital models, calculate the network characteristics of the pulmonary vascular network and the airway network;

[0015] Generate blood supply function scores and ventilation function scores according to the network characteristics;

[0016] Calculate a comprehensive function score based on the blood supply function score and the ventilation function score;

[0017] Output steps, including:

[0018] Output the comprehensive function score as the lung segment function evaluation result.

[0019] Preferably, the construction of the digital models of the pulmonary vascular network and the airway network specifically includes:

[0020] Segment the medical image data to extract the pulmonary vascular and airway structures;

[0021] Based on the extracted pulmonary vascular and airway structures, construct a three-dimensional network model;

[0022] Perform topological analysis on the three-dimensional network model to determine network nodes and connection relationships.

[0023] Preferably, the calculation of the network characteristics of the pulmonary vascular network and the airway network includes calculating one or more of the following characteristics:

[0024] Network centrality;

[0025] Network efficiency;

[0026] Network density;

[0027] Network connectivity;

[0028] Network robustness.

[0029] Preferably, the calculation of network centrality specifically includes:

[0030] Count the number of nodes and edges of the pulmonary vascular network and the airway network;

[0031] Calculate the degree of each node and perform normalization;

[0032] Based on the normalized node degree, the PageRank algorithm is used to calculate the weights of network nodes.

[0033] Preferably, the generation of the blood supply function score specifically includes:

[0034] Based on the network characteristics of the pulmonary vascular network, calculate the blood supply weights of each lung segment;

[0035] Normalize the blood supply weights;

[0036] According to the normalized blood supply weights, calculate the blood supply function scores of each lung segment.

[0037] Preferably, the generation of the ventilation function score specifically includes:

[0038] Based on the network characteristics of the airway network and the lung parenchyma density information, calculate the ventilation weights of each lung segment;

[0039] Normalize the ventilation weights;

[0040] According to the normalized ventilation weights, calculate the ventilation function scores of each lung segment.

[0041] Preferably, it further includes a prognosis prediction step:

[0042] Based on the comprehensive function score, establish a machine learning model;

[0043] Use the machine learning model to generate the patient prognosis probability evaluation result.

[0044] Preferably, the establishment of the machine learning model specifically includes:

[0045] Collect patient historical data, including the comprehensive function score and the actual prognosis;

[0046] Use the patient historical data to train the machine learning model;

[0047] Verify and optimize the machine learning model.

[0048] Preferably, it further includes a data preprocessing step:

[0049] Denoise the medical image data;

[0050] Standardize the medical image data;

[0051] Use the principal component analysis method to reduce the dimension of the processed medical image data.

[0052] Preferably, the output step further includes:

[0053] Generate a visual lung segment function evaluation report, including the blood supply function score, ventilation function score, and comprehensive function score of each lung segment;

[0054] Send the lung segment function evaluation report to a medical terminal device for doctors to refer to for clinical decision-making.

[0055] The method of the present invention has the following remarkable technical effects:

[0056] The method of the present invention realizes the multi-dimensional evaluation of lung segment function by comprehensively analyzing the characteristics of the pulmonary vascular network and airway network, and combining advanced network analysis and machine learning techniques. This method not only overcomes the limitations of only focusing on single features in the prior art, but also can make full use of the rich information in medical images to improve the accuracy and comprehensiveness of the evaluation.

[0057] Specifically, the method of the present invention first constructs digital models of the pulmonary blood vessels and airways, and then comprehensively characterizes the characteristics of the pulmonary structure by calculating a series of network features, such as centrality, efficiency, density, etc. This network analysis-based method can reveal complex structural information that is difficult to capture by traditional methods, providing a new perspective for function evaluation.

[0058] More importantly, the method of the present invention generates a comprehensive function score by integrating the blood supply function score and ventilation function score, realizing the comprehensive evaluation of lung segment function. This comprehensive score not only reflects the blood supply and ventilation conditions of the lung segment, but also considers the mutual influence of these two aspects, thus providing a more accurate and reliable function evaluation result.

[0059] In addition, the method of the present invention also introduces machine learning techniques to establish a prognostic prediction model based on the comprehensive function score. This innovation enables this method to not only evaluate the current lung segment function status, but also predict the future prognosis, providing more comprehensive support for clinical decision-making.

[0060] Generally speaking, the method of the present invention realizes the comprehensive and accurate evaluation and prognostic prediction of lung segment function through innovative technologies such as multi-dimensional analysis, network feature extraction, comprehensive scoring, and machine learning prediction. This method not only overcomes the limitations of the prior art, but also provides a new tool for the diagnosis, treatment, and prognostic evaluation of respiratory diseases. It can help doctors better understand the patient's lung function status, formulate more personalized and precise treatment plans, thereby improving the treatment effect and patient prognosis.

[0061] The method of the present invention also has good scalability and adaptability. By adjusting the selection and weights of network features and optimizing the machine learning model, this method can adapt to different types of lung diseases and different clinical needs. This flexibility makes this method promising to play an important role in the diagnosis and treatment of various respiratory diseases and contribute to the development of precision medicine. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is the overall flowchart of the method of the present invention.

[0063] Figure 2 It is the flowchart of constructing the digital model of the present invention.

[0064] Figure 3 It is the flowchart of network feature calculation of the present invention.

[0065] Figure 4 It is the flowchart of blood supply function scoring of the present invention.

[0066] Figure 5 It is the flowchart of ventilation function scoring of the present invention.

[0067] Figure 6 It is the flowchart of comprehensive function scoring of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0068] As Figure 1-6 shown, the present invention provides a method for evaluating lung segment function based on watershed features. This method makes full use of medical imaging technology and network analysis theory to achieve a comprehensive evaluation of lung segment function. The following will describe the present invention in detail in conjunction with specific embodiments.

[0069] First, the method of the present invention includes an acquisition step, a processing step, and an output step. In the acquisition step, this method acquires medical imaging data of the lungs. Preferably, these medical imaging data can be CT (Computed Tomography) scan images or MRI (Magnetic Resonance Imaging) images. These high-resolution medical images provide a basis for subsequent analysis.

[0070] In the processing step, the method of the present invention first constructs digital models of the pulmonary vascular network and the airway network based on the acquired medical imaging data. Specifically, this step can be achieved through image segmentation techniques. For example, threshold segmentation method or region growing method can be used to extract vascular and airway structures. Then, through three-dimensional reconstruction technology, these structures are transformed into three-dimensional network models.

[0071] Next, based on the constructed digital model, this method calculates the network characteristics of the pulmonary vascular network and the airway network. These network characteristics include, but are not limited to, network centrality, network efficiency, network density, network connectivity, and network robustness. For example, network centrality can be calculated by the PageRank algorithm, network efficiency can be measured by the average shortest path length, and network density can be evaluated by the tightness of connections between nodes.

[0072] After obtaining these network characteristics, this method generates a blood supply function score and a ventilation function score based on these characteristics. The blood supply function score mainly reflects the performance of the pulmonary vascular network, while the ventilation function score reflects the performance of the airway network. Preferably, the calculation of these two scores can adopt a weighted summation method, where the weights can be optimized by machine learning algorithms.

[0073] Finally, based on the blood supply function score and the ventilation function score, this method calculates a comprehensive function score. This comprehensive score comprehensively reflects the overall functional status of the lung segment. For example, the following formula can be used to calculate the comprehensive score:

[0074] ,

[0075] where, is the comprehensive function score, is the blood supply function score, is the ventilation function score, and are the weight coefficients, and + = 1.

[0076] In the output step, this method outputs the calculated comprehensive function score as the lung segment function evaluation result. This result can provide important diagnostic references for doctors.

[0077] Furthermore, when constructing the digital models of the pulmonary vascular network and the airway network, the method of the present invention also includes segmenting medical image data to extract the pulmonary vascular and airway structures. This step can use a variety of image segmentation algorithms, such as region-growing methods, level set methods, or deep learning-based segmentation methods. For example, using a U-Net network for the segmentation of blood vessels and airways, the accuracy can reach over 95%.

[0078] After extracting the pulmonary vascular and airway structures, this method constructs a three-dimensional network model based on these structures. This step can use skeletonization algorithms, such as the distance transformation method or thinning algorithms. Preferably, the Lee algorithm can be used for skeletonization, which can well preserve the topological structure of the network.

[0079] Finally, the method performs a topological analysis on the constructed three-dimensional network model to determine the network nodes and connection relationships. This step can use algorithms in graph theory, such as depth-first search (DFS) or breadth-first search (BFS). In this way, a complete network topology structure can be obtained, laying a foundation for subsequent network feature calculations.

[0080] When calculating the network features of the pulmonary vascular network and airway network, the method of the present invention includes calculating one or more of network centrality, network efficiency, network density, network connectivity, and network robustness. These features reflect the performance of the network from different perspectives.

[0081] For example, network centrality reflects the importance of nodes in the network. It can be measured using metrics such as degree centrality, betweenness centrality, or eigenvector centrality. Among them, the calculation formula for degree centrality is:

[0082] ,

[0083] where is the degree centrality of node , is the degree of node , is the total number of nodes in the network.

[0084] Network efficiency reflects the ability of the network to transmit information. It can be measured by the average shortest path length, and the calculation formula is:

[0085] ,

[0086] where E is the network efficiency, n is the number of nodes, is the shortest path length between nodes i and j.

[0087] Network density reflects the tightness of network connections and can be calculated by the ratio of the actual number of edges to the maximum possible number of edges:

[0088] ,

[0089] where D is the network density, m is the actual number of edges, and n is the number of nodes.

[0090] The calculation of these network features provides an important basis for subsequent functional scoring. By comprehensively considering these features, the functional status of the lung segments can be comprehensively evaluated, providing strong support for clinical diagnosis and treatment decisions. In a preferred embodiment of the present invention, the step of calculating network centrality is further refined. First, the method counts the number of nodes and edges in the pulmonary vascular network and airway network. This step lays a foundation for subsequent analysis and can intuitively reflect the scale and complexity of the network.

[0091] Next, this method calculates the degree of each node and performs normalization. The degree of a node reflects the number of direct connections it has with other nodes and is a basic indicator of node importance. Normalization eliminates the effect of network size on the degree value, allowing networks of different sizes to be compared. The normalization formula is as follows:

[0092] ,

[0093] in, is the normalized degree value of node v, is the original degree value of node v, and are the minimum and maximum degree values in the network respectively.

[0094] Finally, this method uses the PageRank algorithm to calculate the weights of network nodes based on the normalized node degrees. The PageRank algorithm was originally used to rank web pages, but it can also be effectively applied to other types of network analysis. In this invention, the use of the PageRank algorithm can more comprehensively consider the importance of nodes, considering not only the direct connections of nodes, but also the influence of indirect connections. The calculation formula of PageRank is as follows:

[0095] ,

[0096] Among them, PR(v) is the PageRank value of node v, d is the damping factor (usually 0.85), N(v) is the set of nodes pointing to node v, and L(u) is the out-degree of node u.

[0097] The method of the present invention calculates network centrality in this way, which can more accurately reflect the importance of each node in the pulmonary vascular and airway network, and provide an important basis for subsequent functional evaluation.

[0098] In generating the blood supply function score, the method of the present invention first calculates the blood supply weight of each lung segment based on the network characteristics of the pulmonary vascular network. The network characteristics here may include the aforementioned indicators such as centrality, efficiency, and density. The blood supply weight can be calculated using a weighted summation method, for example:

[0099] ,

[0100] in, is the blood supply weight of the i-th lung segment, 、 and are the centrality, efficiency and density indexes of the lung segment, 、 and are the corresponding weight coefficients.

[0101] Next, the method normalizes the calculated blood supply weights. The purpose of normalization is to unify the blood supply weights of different lung segments to the same scale for easy comparison and subsequent processing. The normalization formula can adopt the min-max normalization method mentioned above.

[0102] Finally, the method calculates the blood supply function scores of each lung segment based on the normalized blood supply weights. The score calculation can directly use the normalized weights or perform some non-linear transformations to better reflect the characteristics of the blood supply function. For example, the sigmoid function can be used for transformation:

[0103] ,

[0104] where is the blood supply function score of the i-th lung segment, is the normalized blood supply weight, is the adjustment parameter used to control the steepness of the function.

[0105] In the process of generating the ventilation function score, the method of the present invention adopts similar steps but adds the consideration of the lung parenchyma density information. First, the method calculates the ventilation weights of each lung segment based on the network characteristics of the airway network and the lung parenchyma density information. The introduction of the lung parenchyma density information can better reflect the ventilation status of the alveoli and improve the accuracy of the score.

[0106] The ventilation weight can be calculated using the following formula:

[0107] ,

[0108] where is the ventilation weight of the i-th lung segment, is the airway network characteristic of this lung segment (which may be the weighted sum of multiple characteristics), is the lung parenchyma density of this lung segment, and are weight coefficients. Note that is used here because the lower the lung parenchyma density, the better the ventilation function usually is.

[0109] Next, the method also normalizes the ventilation weights and then calculates the ventilation function scores of each lung segment. The calculation method can be similar to that of the blood supply function score or can be appropriately adjusted according to the characteristics of the ventilation function.

[0110] In this way, the method of the present invention can comprehensively consider the two factors of blood supply and ventilation, and provide a reliable basis for the comprehensive evaluation of lung segment function. This evaluation method not only takes into account the characteristics of the anatomical structure, but also combines the manifestations of physiological functions, and can provide more comprehensive and accurate reference information for clinical diagnosis and treatment decisions. In another embodiment of the present invention, a prognosis prediction step is also included. The introduction of this step enables the present method to not only evaluate the current lung segment function status, but also predict the patient's future prognosis, providing more comprehensive support for clinical decision-making.

[0111] Specifically, this method first builds a machine learning model based on the previously calculated comprehensive functional score. The purpose of this model is to correlate the comprehensive functional score with the patient's prognosis. Preferably, various machine learning algorithms can be used, such as support vector machines (SVMs), random forests, or deep neural networks (DNNs). When selecting a specific algorithm, it is important to consider the characteristics of the data and the requirements of the prediction task.

[0112] For example, if a support vector machine is used, the following decision function can be used:

[0113] ,

[0114] Where x is the input feature vector (mainly the comprehensive function score in this case), is the label of the training sample, is the Lagrange multiplier, is the kernel function, b is the bias term. The kernel function can choose the Gaussian kernel:

[0115] ,

[0116] in, is the parameter of the Gaussian kernel and needs to be tuned through methods such as cross-validation.

[0117] After building a machine learning model, this method uses it to generate a patient prognosis probability assessment result. This result can be binary (such as the probability of a good prognosis and a poor prognosis) or multi-class (such as the probability distribution of short-term, medium-term, and long-term prognosis). For example, using the Softmax function can produce a multi-class probability output:

[0118] ,

[0119] in, is the original output of the model for the jth class, and K is the total number of classes.

[0120] To improve the accuracy of prognosis prediction, the method of the present invention takes a series of measures when establishing a machine learning model. First, the method collects historical data of patients, including comprehensive function scores and actual prognosis. These historical data provide the basis for model training. During the data collection process, attention should be paid to the quality and representativeness of the data, covering different types and degrees of lung diseases as much as possible.

[0121] Next, the method uses the collected historical data of patients to train a machine learning model. During the training process, techniques such as cross-validation can be used to evaluate the performance of the model and adjust the model parameters. For example, k-fold cross-validation can be used. The data set is divided into k parts. Each time, k - 1 parts are used as the training set, and the remaining 1 part is used as the validation set, and this is repeated k times. This method can effectively avoid overfitting and improve the generalization ability of the model.

[0122] Finally, the method validates and optimizes the trained machine learning model. Validation can use an independent test set to evaluate the performance of the model on new data. If the model performs poorly, methods such as adjusting the model structure, adding features, or using ensemble learning can be considered for optimization. For example, the Adaboost algorithm can be used to construct multiple weak classifiers, and then they are combined into a strong classifier:

[0123] ,

[0124] where is the t-th weak classifier, is the weight of this classifier.

[0125] In another embodiment of the present invention, it also includes a data preprocessing step. The purpose of this step is to improve the accuracy and efficiency of subsequent analysis. First, the method performs denoising on medical image data. Medical images, especially CT and MRI images, often contain a certain degree of noise, which may affect subsequent analysis. Multiple methods can be used for denoising, such as Gaussian filtering, median filtering, or wavelet transform, etc. For example, the formula for using Gaussian filtering is as follows:

[0126] ,

[0127] where (x,y) are pixel coordinates, is the standard deviation of the Gaussian function.

[0128] Next, the method normalizes the denoised medical image data. The purpose of normalization is to unify image data obtained from different sources and different devices to the same scale for subsequent processing. Commonly used normalization methods include Z-score normalization and Min-Max normalization. For example, the formula for Z-score normalization is as follows:

[0129] ,

[0130] where \(x\) is the original data, is the mean value, is the standard deviation.

[0131] Finally, this method uses the principal component analysis (PCA) method to reduce the dimension of the processed medical image data. PCA can effectively reduce the dimension of the data while retaining most of the information, thereby improving the efficiency of subsequent processing. The core of PCA is to solve the eigenvalues and eigenvectors of the covariance matrix:

[0132] ,

[0133] where \(C\) is the covariance matrix, is the eigenvector, is the corresponding eigenvalue.

[0134] In the output step of the present invention, in addition to outputting the comprehensive function score as the result of the pulmonary segment function evaluation, it also includes generating a visual pulmonary segment function evaluation report. This report includes the blood supply function score, ventilation function score, and comprehensive function score of each pulmonary segment. The visual way can be a color-coded 3D lung model, where different colors represent different function scores. For example, three colors, red, yellow, and green, can be used to represent the pulmonary segments with poor, medium, and good functions respectively.

[0135] In addition, this method also sends the generated pulmonary segment function evaluation report to the medical terminal device for doctors to refer to for clinical decision-making. This step can be achieved through the hospital internal network or a secure cloud service. In this way, doctors can view the pulmonary segment function evaluation results of patients at any time and place, providing convenience for diagnosis and treatment decisions.

[0136] Generally speaking, the pulmonary segment function evaluation method based on watershed characteristics provided by the present invention comprehensively considers the characteristics of the pulmonary vascular network and airway network, combines advanced data processing and machine learning technologies, and realizes the comprehensive evaluation and prognosis prediction of the pulmonary segment function. This method not only improves the accuracy of the pulmonary segment function evaluation but also provides more comprehensive and reliable reference information for clinical decision-making, and is expected to play an important role in the diagnosis and treatment of respiratory diseases.

[0137] The above; only the preferred specific embodiments of the present invention; but the protection scope of the present invention is not limited thereto; any person familiar with the art within the scope disclosed by the present invention; according to the solution and improvement concept of the present invention; making equivalent substitutions or changes; should be covered within the protection scope of the present invention.

Claims

1. A lung segment function assessment method based on watershed characteristics, characterized in that: include: The acquisition steps include: Acquire medical imaging data of the lungs; Processing steps include: constructing a digital model of the pulmonary vascular network and airway network based on the medical imaging data; calculating network characteristics of the pulmonary vascular network and airway network based on the digital model; generating a blood supply function score and a ventilation function score according to the network characteristics; Calculating a comprehensive function score based on the blood supply function score and the ventilation function score; First, based on the network characteristics of the pulmonary vascular network, the blood supply weight of each lung segment is calculated. Next, the calculated blood supply weight is normalized. Based on the normalized blood supply weight, the blood supply function score of each lung segment is calculated. Based on the network characteristics of the airway network and lung parenchymal density information, the ventilation weight of each lung segment is calculated; next, the ventilation weight is normalized, and then the ventilation function score of each lung segment is calculated; Output steps include: The comprehensive functional score is output as a lung segment function assessment result.

2. The method according to claim 1, characterized in that The digital model of the pulmonary vascular network and the airway network is constructed as follows: Segmenting the medical image data to extract lung blood vessels and airway structures; Based on the extracted lung blood vessels and airway structures, a three-dimensional network model is constructed; Perform a topological analysis on the three-dimensional network model to determine network nodes and connection relationships.

3. The method according to claim 1, characterized in that Calculating the network features of the pulmonary vascular network and the airway network includes calculating one or more of the following features: network centrality; Network efficiency; Network density; Network connectivity; Network robustness.

4. The method according to claim 3, characterized in that Computing network centrality specifically includes: Counting the number of nodes and edges of the pulmonary vascular network and the airway network; Calculate the degree of each node and perform normalization; Based on the normalized node degree, the PageRank algorithm is used to calculate the weight of the network node.

5. The method according to claim 1, wherein It also includes the prognosis prediction step: establishing a machine learning model based on the comprehensive functional score; The machine learning model is used to generate a patient prognosis probability assessment result.

6. The method according to claim 5, characterized in that The establishment of the machine learning model specifically includes: Collect patient historical data, including comprehensive functional scores and actual prognosis; training a machine learning model using the patient historical data; Validate and optimize the machine learning model.

7. The method according to claim 1, characterized in that It also includes data preprocessing steps: performing denoising processing on the medical image data; performing standardization processing on the medical imaging data; The principal component analysis method is used to reduce the dimensionality of the processed medical imaging data.

8. The method according to claim 1, characterized in that The output step further includes: Generate a visual lung segment function assessment report, including the blood supply function score, ventilation function score and comprehensive function score of each lung segment; The lung segment function assessment report is sent to a medical terminal device for doctors to use as a reference for clinical decision making.

Images