Respiratory failure treatment scheme recommendation method and system based on artificial intelligence

Abstract

The embodiment of the invention provides a respiratory failure treatment scheme recommendation method and system based on artificial intelligence, and the method comprises the steps: obtaining the first multi-modal medical collection data of a target object, and enabling the first multi-modal medical collection data to comprise text modal data and image modal data; the text modal data comprises physiological feature data, basic disease data and blood gas analysis data; data preprocessing operation is conducted on the first multi-modal medical collection data, second multi-modal medical collection data are obtained, and data preprocessing comprises data cleaning, normalization processing and time window alignment processing. And extracting a medical feature vector set of the target object from the second multi-mode medical collection data. And performing classification processing on the target object according to the medical feature vector set, and determining an object treatment category label of the target object. And according to the object treatment category label, a recommended treatment scheme of the target object is determined, so that the efficiency and accuracy of determining the respiratory failure treatment scheme are improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to an artificial intelligence-based method and system for recommending treatment options for respiratory failure. Background Art

[0002] Respiratory failure is a clinical syndrome that seriously threatens human health, with high morbidity and mortality. Timely and accurate formulation of appropriate treatment plans for patients with respiratory failure is crucial to improving patient prognosis and reducing mortality.

[0003] In the existing technology, the determination of respiratory failure treatment plans mainly relies on the doctor's clinical experience and traditional medical judgment methods. Doctors usually make a preliminary assessment of the patient's condition based on a single type of physiological indicator data, such as heart rate, blood pressure, etc., combined with their own professional knowledge and previous treatment experience. However, this assessment method based on a single physiological indicator and subjective experience has obvious limitations. A single physiological indicator can only reflect one aspect of the patient's condition and cannot fully and accurately present the patient's overall condition. In addition, different doctors have different experience levels and judgment criteria, which may lead to a lack of consistency and accuracy in the formulation of treatment plans.

[0004] Furthermore, for some complex cases of respiratory failure, relying solely on a single physiological indicator can fail to capture subtle changes in the condition and potential influencing factors. For example, patients may suffer from multiple underlying conditions, which can interact with each other and complicate the treatment of respiratory failure. However, traditional methods often fail to comprehensively consider these multiple factors, which can easily lead to suboptimal treatment options and delay treatment for patients.

[0005] Furthermore, most existing methods rely on simple classification rules, often based on limited clinical experience, that cannot adequately adapt to the individual needs of different patients. For patients with special circumstances or complex conditions, these simple classification rules may not accurately categorize them into appropriate treatment categories, making it difficult to provide precise treatment recommendations.

[0006] In other words, existing technologies have problems in the process of formulating respiratory failure treatment plans, such as insufficient data utilization, inaccurate evaluation methods, and lack of personalization in treatment plan recommendations. Summary of the Invention

[0007] The present application provides an artificial intelligence-based method and system for recommending respiratory failure treatment plans to improve the efficiency and accuracy of determining respiratory failure treatment plans.

[0008] In a first aspect, the present application provides a method for recommending a treatment plan for respiratory failure based on artificial intelligence, comprising: Acquire first multimodal medical collected data of the target object, where the first multimodal medical collected data includes text modality data and image modality data, and the text modality data includes physiological characteristic data, basic disease data, and blood gas analysis data; Performing data preprocessing on the first multimodal medical collected data to obtain second multimodal medical collected data, wherein the data preprocessing includes data cleaning, normalization, and time window alignment; extracting a medical feature vector set of the target object from the second multimodal medical acquisition data; performing classification processing on the target object according to the medical feature vector set to determine an object treatment category label of the target object; Determine a recommended treatment plan for the target subject based on the subject's treatment category label.

[0009] Optionally, performing classification processing on the target object according to the medical feature vector set to determine the object treatment category label of the target object includes: Performing a cluster analysis based on the medical feature vector set to obtain a cluster analysis result of the target object, and determining an object treatment category label of the target object based on the cluster analysis result and a mapping relationship between the cluster analysis result and the object treatment category label; or, The medical feature vector set is input into a pre-trained decision tree model to obtain the object treatment category label of the target object.

[0010] Optionally, performing cluster analysis based on the medical feature vector set to obtain cluster analysis results of the target object includes: determining an initial cluster center according to the medical feature vectors included in the medical feature vector set; Based on the treatment risk quantification index of the medical feature vector, obtaining the medical weight coefficient corresponding to each medical feature vector; Determining the distance between each medical feature vector and the initial cluster center according to each medical feature vector and the medical weight coefficient corresponding to each medical feature vector, and generating an initial clustering result based on the distance between each medical feature vector and the initial cluster center; According to the initial clustering result, the medical feature vector of the target object and the medical weight coefficient, the initial cluster center is iteratively updated until a preset termination condition is met, thereby obtaining a cluster analysis result of the target object.

[0011] Optionally, performing cluster analysis based on the medical feature vector set to obtain cluster analysis results of the target object includes: Based on the distribution characteristics of each medical feature vector in the medical feature vector set, constructing an initial state of hierarchical clustering, wherein each medical feature vector in the initial state is an independent initial cluster; Based on the treatment risk quantification index of the medical feature vector, obtaining the medical weight coefficient corresponding to each medical feature vector; Calculating the weighted distances between the independent initial clusters based on the medical feature vectors and the medical weight coefficients corresponding to the medical feature vectors to generate a cluster distance matrix; According to each weighted distance in the cluster distance matrix, independent initial clusters whose weighted distance is less than or equal to a preset distance threshold are iteratively merged until the number of clusters of the independent initial clusters reaches a preset number, thereby obtaining a cluster analysis result of the target object.

[0012] Optionally, before performing classification processing on the target object according to the medical feature vector set and determining the treatment category label of the target object, the method further includes a training method for the decision tree model, specifically including: constructing an initial state of hierarchical clustering based on the distribution characteristics of each sample medical feature vector in the sample medical feature vector set of the sample object, wherein each sample medical feature vector in the initial state is an independent initial cluster; Based on the treatment risk quantification index of the sample medical feature vector, obtaining the medical weight coefficient corresponding to each of the sample medical feature vectors; Calculating the weighted distances between the independent initial clusters based on the sample medical feature vectors and the medical weight coefficients corresponding to the sample medical feature vectors to generate a sample cluster distance matrix; According to each weighted distance in the sample cluster distance matrix, iteratively merge independent initial clusters whose weighted distance is less than or equal to a preset distance threshold until the number of clusters reaches a preset number, thereby obtaining an initial classification result of the sample object; Obtaining constraints of an initial decision tree model, wherein the constraints include a Gini coefficient constraint and a medical feature vector threshold constraint; The initial decision tree model is trained according to the initial classification results and the constraint conditions to generate the decision tree model.

[0013] Optionally, training the initial decision tree model according to the initial classification results and the constraint conditions to generate the decision tree model includes: Obtaining the classification label corresponding to each sample medical feature vector in the initial classification result, as well as the Gini coefficient constraint and the medical feature vector threshold constraint in the constraint conditions; Based on the classification label, traverse each medical feature vector in the sample medical feature vector set, calculate the Gini coefficient of the medical feature vector under different splitting thresholds, and select the splitting feature with the smallest Gini coefficient and the corresponding splitting threshold as a candidate splitting node; According to the medical feature vector threshold constraint, a normalization check is performed on the splitting threshold in the candidate splitting node; if the splitting threshold is within the numerical range of the normalized medical feature vector, the candidate splitting node is retained; otherwise, the candidate splitting node is eliminated; Based on the retained candidate split nodes, sorting them in ascending order of the Gini coefficient to generate a split priority queue, selecting the candidate split node with the smallest Gini coefficient from the split priority queue as the current node split condition, and dividing the sample medical feature vector set into two subsets; Recursively traversing the two subsets, repeatedly calculating the Gini coefficient, screening the splitting features and the splitting threshold, verifying the splitting threshold, and generating the splitting priority queue until a preset stopping condition is met, wherein the preset stopping condition includes that the Gini coefficient of the subset is zero, the splitting thresholds of all medical feature vectors do not meet the medical feature vector threshold constraint, or the maximum recursion depth is reached; A tree structure including classification rules and splitting thresholds is generated according to all nodes after recursive splitting as the decision tree model.

[0014] Optionally, after classifying the target object according to the medical feature vector set and determining the treatment category label of the target object, the method further includes: Acquire the target subject's disease monitoring quantitative data according to preset batch configuration parameters, wherein the disease monitoring quantitative data includes vital sign quantitative data, blood gas analysis quantitative data, and pathological trend quantitative data; Obtaining progression weights corresponding to the vital sign quantitative data, the blood gas analysis quantitative data, and the pathological trend quantitative data respectively; Determining a quantitative indicator of the target subject's disease progression based on an attention mechanism according to the quantitative vital sign data, the quantitative blood gas analysis data, the quantitative lesion trend data, and the progression weight; determining whether it is necessary to adjust the subject treatment category label of the target subject based on the disease progression quantification indicator, the current subject treatment category label of the target subject, and a level conversion threshold corresponding to the disease progression quantification indicator; If the subject treatment category label of the target subject needs to be adjusted, a recommended treatment regimen for the target subject is determined based on the adjusted subject treatment category label.

[0015] Optionally, obtaining the progression weights corresponding to the vital sign quantitative data, the blood gas analysis quantitative data, and the lesion trend quantitative data, respectively, includes: Obtaining a sample medical feature vector set of each sample patient from pre-constructed disease progression history data of multiple sample patients, wherein each sample medical feature vector in the sample medical feature vector set includes vital sign features, blood gas analysis features, and pathological trend features, and has been normalized; Extracting the disease exacerbation index of the sample patient as a regression target variable, wherein the disease exacerbation index is quantitatively generated according to the occurrence time and severity of a preset exacerbation event; Using the vital sign characteristics, the blood gas analysis characteristics, and the pathological trend characteristics as independent variables, fitting a linear regression model or a logistic regression model, and outputting the regression coefficient corresponding to each characteristic; Calculate the importance index of each feature based on the absolute value of the regression coefficient of each feature; The importance index is normalized so that the sum of the progress weights of the features is 1, and progress weights corresponding to the vital sign quantitative data, the blood gas analysis quantitative data, and the pathological trend quantitative data are generated.

[0016] Optionally, the determining whether it is necessary to adjust the subject treatment category label of the target subject based on the disease progression quantification indicator, the current subject treatment category label of the target subject, and a level conversion threshold corresponding to the disease progression quantification indicator includes: Obtaining a preset level conversion threshold range associated with the current subject treatment category label, wherein the preset level conversion threshold range includes an upper deterioration threshold limit and a lower improvement threshold limit; performing a first comparison between the disease progression quantification indicator and the upper deterioration threshold, and generating a category upward adjustment instruction if the disease progression quantification indicator is greater than or equal to the upper deterioration threshold; Performing a second comparison between the disease progression quantification index and the improvement threshold lower limit, and generating a category downgrade instruction if the disease progression quantification index is less than or equal to the improvement threshold lower limit; determining an adjusted subject treatment category label for the target subject according to the category-up instruction or the category-down instruction, and verifying whether the adjusted subject treatment category label is within a preset label hierarchy range; If the adjusted object treatment category label exceeds the preset label level range, the current object treatment category label is maintained unchanged; otherwise, the adjusted object treatment category label is used as the updated object treatment category label.

[0017] In a second aspect, the present application provides an artificial intelligence-based respiratory failure treatment plan recommendation system, which includes a machine-readable storage medium and a processor, wherein the machine-readable storage medium stores machine-executable instructions. When the processor executes the machine-executable instructions, the artificial intelligence-based respiratory failure treatment plan recommendation system implements the aforementioned artificial intelligence-based respiratory failure treatment plan recommendation method.

[0018] In a third aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the method as described in any one of the first aspects.

[0019] In a fourth aspect, the present application provides a computer program product, which, when executed by a processor, is used to implement the method as described in any one of the first aspects.

[0020] The artificial intelligence-based respiratory failure treatment recommendation method and system provided in this application overcomes the limitations of existing technologies that rely solely on single physiological indicator data by acquiring first multimodal medical data comprising textual modalities (physiological characteristic data, underlying disease data, and blood gas analysis data) and image modalities. This method and system can reflect the target patient's condition from multiple dimensions. Data preprocessing, including data cleaning, normalization, and time window alignment, effectively addresses existing medical data issues such as noise, missing values, inconsistencies, and formatting differences, thereby improving data quality and usability. A set of medical feature vectors is extracted from the preprocessed second multimodal medical data, converting the multimodal data into quantifiable and analyzable feature vectors to facilitate further processing and mining of potential information within the data. Classification based on the set of medical feature vectors determines the subject's treatment category label, avoiding the drawback of existing simple classification rules that are unable to adapt to personalized needs. This allows for more accurate classification of the target patient, ultimately determining a recommended treatment plan based on the patient's treatment category label. This enables personalized and precise recommendations for respiratory failure treatment plans, improving the targetedness and effectiveness of treatment plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0022] Figure 1 A flowchart of a method for recommending a treatment plan for respiratory failure based on artificial intelligence provided in an embodiment of the present application; Figure 2 A flowchart of another method for recommending a treatment plan for respiratory failure based on artificial intelligence provided in an embodiment of the present application; Figure 3 A flowchart of another method for recommending a treatment plan for respiratory failure based on artificial intelligence provided in an embodiment of the present application; Figure 4 A schematic diagram of the structure of an artificial intelligence-based respiratory failure treatment plan recommendation system provided in an embodiment of the present application.

[0023] The above drawings illustrate specific embodiments of the present application, which will be described in more detail below. These drawings and the textual description are not intended to limit the scope of the present application in any way, but rather to illustrate the concepts of the present application to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0024] To help those skilled in the art better understand the present invention, the following will provide a clear and complete description of the technical solutions in the embodiments of the present invention, in conjunction with the accompanying drawings. Obviously, the described embodiments are only a portion of the embodiments of the present invention, not all of them. All other embodiments derived by those skilled in the art based on the embodiments of the present invention without inventive effort are intended to fall within the scope of protection of the present invention.

[0025] The terms "first," "second," and so on, in the specification and claims of the present invention and the accompanying drawings are used to distinguish between different items, not to describe a specific order. Furthermore, the terms "including," "having," and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product, or end comprising a series of steps or elements is not limited to the listed steps or elements but may optionally include steps or elements not listed therein, or may optionally include other steps or elements inherent to such process, method, product, or end.

[0026] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute a separate or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0027] Figure 1This is a flowchart of a method for recommending a treatment plan for respiratory failure based on artificial intelligence provided in an embodiment of the present application. It should be understood that in other embodiments, the order of some steps in the method for recommending a treatment plan for respiratory failure based on artificial intelligence in this embodiment can be shared with each other according to actual needs, or some steps can be omitted or maintained. Figure 1 As shown, the method may include the following steps: Step S110: Acquire first multimodal medical acquisition data of the target object.

[0028] Among them, the first multimodal medical acquisition data includes text modality data and image modality data, and the text modality data includes physiological characteristic data, basic disease data, and blood gas analysis data.

[0029] In actual application scenarios, the acquisition of the first multimodal medical data of the target object involves multiple different data sources and acquisition methods. In terms of the acquisition of text modality data, the collection of physiological characteristic data relies on a variety of professional medical monitoring equipment. Taking heart rate data as an example, it can be monitored in real time through a wearable heart rate monitoring bracelet or an electrocardiogram monitor in a hospital ward. These devices will continuously record the changes in the target object's heart rate and transmit the data to the corresponding storage system. For blood pressure data, an electronic blood pressure monitor is usually used for measurement. After the measurement, the data will be automatically stored or manually entered into the medical information system. Respiratory rate data can be obtained through a respiratory sensor belt or camera-based respiratory monitoring technology, and the collected data will also be integrated into a unified data storage.

[0030] Basic disease data is primarily obtained from the target patient's medical history. The hospital's electronic medical record system is a crucial data source, detailing the various illnesses the target patient has suffered, including diagnosis time, treatment process, and recovery status. Furthermore, physicians' consultation notes are another crucial source of basic disease data. During their conversations with the target patient, doctors will inquire in detail about their medical history and record this information.

[0031] Blood gas analysis data is obtained by collecting a blood sample from a subject and then testing it using a specialized blood gas analyzer. When collecting blood samples, an appropriate sampling site, such as an artery, is typically selected to obtain accurate blood gas analysis results. After blood collection, the sample is promptly sent to the laboratory for testing by a professional using a blood gas analyzer. During the test, the instrument analyzes various blood components, such as oxygen partial pressure, carbon dioxide partial pressure, and pH. The test results are then digitized and stored in the hospital's laboratory information system.

[0032] Acquiring image modality data primarily involves the hospital's imaging equipment. Chest X-rays are obtained by scanning a patient's chest using an X-ray machine. During the imaging process, the patient must maintain the correct posture according to the doctor's instructions to ensure clear and accurate images. After the imaging is complete, the X-ray image data is transmitted to the hospital's Picture Archiving and Communication System (PACS) for storage and management.

[0033] CT scan images are acquired by performing multi-slice scans of the patient's chest. The CT scanner rotates around the patient, scanning the chest from different angles. The resulting series of slices are then reconstructed to create a three-dimensional CT scan image. This image data is also stored in a PACS system for subsequent review and analysis.

[0034] Step S120: performing data preprocessing on the first multimodal medical collected data to obtain second multimodal medical collected data.

[0035] The data preprocessing includes data cleaning, normalization and time window alignment.

[0036] After the first multimodal medical data is acquired, it is necessary to perform data preprocessing operations on the data because the data may have noise, missing values, outliers, different scales and time stamps, and other problems.

[0037] Data cleaning is the first step in preprocessing. For physiological characteristic data in text modal data, outliers must be detected and addressed first. For example, heart rate data normally fluctuates within a certain range. If the heart rate value recorded at a given moment deviates significantly from the normal heart rate range for the target subject, it may be an outlier. Outliers can be identified by setting upper and lower thresholds. The upper threshold can be determined based on medical knowledge and the individual condition of the target subject, and the lower threshold is similarly determined. For identified outliers, the following approaches can be used: If the outlier is caused by equipment failure or data entry error, the data record can be deleted directly. If the outlier is caused by a unique physiological condition of the target subject but still has some reference value, interpolation methods, such as linear interpolation, can be used to estimate a reasonable value based on the normal data points before and after the outlier.

[0038] Basic disease data may contain duplicate records or incorrect disease names. This can be achieved by establishing a standard dictionary of disease names, comparing collected disease names against the dictionary, and correcting any non-compliant names. Furthermore, data deduplication algorithms, such as hashing algorithms, can be used to deduplicate basic disease data to ensure data accuracy and uniqueness.

[0039] Cleaning blood gas analysis data also requires attention to outliers. Since each indicator in blood gas analysis data has a normal physiological range, such as blood oxygen partial pressure and carbon dioxide partial pressure, thresholds can be set based on these normal ranges, and data that exceeds these thresholds can be marked and processed. For missing blood gas analysis data, machine learning algorithms, such as regression analysis, can be used to predict missing values based on other relevant data about the target subject, such as physiological characteristics and underlying disease data.

[0040] Image modality data may suffer from image blur and noise. Image enhancement techniques can be used to improve image clarity and quality. For example, histogram equalization can be used to enhance image contrast and adjust the grayscale distribution to make details more visible. Image noise can be removed using filtering algorithms, such as Gaussian filtering, which smooths noise by performing a weighted average of each pixel in the image while preserving edge information.

[0041] Normalization is to unify data of different features into the same scale range for subsequent analysis and calculation. For physiological feature data in text modal data, such as continuous data such as heart rate and blood pressure, the minimum-maximum normalization method can be used. The specific calculation process of this method is: for each data point, first subtract the minimum value of the feature, and then divide it by the difference between the maximum and minimum values of the feature. In this way, the data can be mapped to the interval [0, 1]. For example, for heart rate data, let the heart rate value at a certain moment be H, the minimum value of the feature be H_min, and the maximum value be H_max, then the normalized heart rate value H_norm=(H-H_min) / (H_max-H_min).

[0042] Since basic disease data is discrete, one-hot encoding can be used to convert it into numerical data. Specifically, a binary vector is created for each disease, with a length equal to the number of disease types. If the target subject has a disease, the corresponding vector position is set to 1, otherwise it is set to 0. The one-hot encoded data is then normalized using a simple mean-standard deviation normalization method. This involves first calculating the mean and standard deviation of each feature, then subtracting the mean from each data point and dividing by the standard deviation.

[0043] For blood gas analysis data, you can also choose an appropriate normalization method based on the data type. If the data is continuous, you can use minimum-maximum normalization or mean-standard deviation normalization. If the data is discrete, you can use one-hot encoding before normalization.

[0044] The purpose of time window alignment is to ensure that data from different modalities are consistent in time. Since physiological characteristic data, blood gas analysis data, etc. may be collected at different time points, and the collection time of image modality data may also be different from that of text modality data, time window alignment is required. A fixed time window can be set, for example, in hours. For each time window, the physiological characteristic data, blood gas analysis data, etc. collected within the time window are integrated. If some data is missing in a time window, interpolation or missing value filling can be used to process it. For image modality data, it needs to be associated with the corresponding time window. If there is no corresponding image data in a time window, you can consider using the image data of the previous and next time windows for interpolation or using the last collected image data as a replacement.

[0045] Step S130: extracting a medical feature vector set of the target object from the second multimodal medical acquisition data.

[0046] After completing the data preprocessing, the second multimodal medical acquisition data is obtained. Next, it is necessary to extract the medical feature vector set of the target object from this data.

[0047] For feature extraction of text modal data, for physiological feature data, normalized features such as heart rate, blood pressure, and respiratory rate can be combined into a vector. For example, assuming that the physiological feature data includes heart rate value H_norm, systolic blood pressure SBP_norm, diastolic blood pressure DBP_norm, and respiratory rate RR_norm, the physiological feature vector can be: P=[H_norm, SBP_norm, DBP_norm, RR_norm].

[0048] For the underlying disease data, after one-hot encoding and normalization, the binary vectors corresponding to each disease are combined into a longer vector. Assuming there are n diseases, the resulting underlying disease vector B = [b_1, b_2, ..., b_n], where b_i represents the normalized binary value corresponding to the i-th disease.

[0049] For blood gas analysis data, normalized blood oxygen partial pressure, carbon dioxide partial pressure, and pH are combined into a vector. Let blood oxygen partial pressure be PO2_norm, carbon dioxide partial pressure be PCO2_norm, and pH be pH_norm. Then the blood gas analysis vector G = [PO2_norm, PCO2_norm, pH_norm].

[0050] Convolutional neural networks (CNNs) can be used to extract features from image modality data. First, the preprocessed image data is input into the CNN. CNNs typically consist of convolutional layers, pooling layers, and fully connected layers. In the convolutional layers, multiple different convolution kernels are used to slide across the image, performing convolution operations to extract local features of the image. Each convolution kernel generates a feature map, and multiple convolution kernels generate multiple feature maps. For example, assuming there are m convolution kernels, each of size k×k, then after passing through the convolution layer, m feature maps of size W×H are obtained (W and H are the width and height of the feature map).

[0051] The pooling layer downsamples the feature maps output by the convolutional layer, reducing the dimensionality of the data while retaining important feature information. Common pooling methods include max pooling and average pooling. Taking max pooling as an example, the maximum value within each pooling window is selected as the output of that window, resulting in a smaller feature map.

[0052] After multiple passes through the convolutional and pooling layers, the feature map of the final layer is expanded into a one-dimensional vector and then fed into a fully connected layer. The fully connected layer performs linear transformations and nonlinear activations on the input vector, ultimately outputting a fixed-length image feature vector I.

[0053] Finally, the physiological feature vector P, underlying disease vector B, blood gas analysis vector G extracted from the text modality data, and the image feature vector I extracted from the image modality data are concatenated to form the target object's medical feature vector set F = [P, B, G, I]. During the concatenation process, it is necessary to ensure that the dimensions of each vector match to ensure the consistency and usability of the medical feature vector set.

[0054] Step S140: performing classification processing on the target object according to the medical feature vector set to determine the object treatment category label of the target object.

[0055] Specifically, the classification process of the target object according to the medical feature vector set and the determination of the object treatment category label of the target object include two methods.

[0056] Method 1: Perform cluster analysis based on the medical feature vector set to obtain the cluster analysis results of the target object, and determine the object treatment category label of the target object based on the cluster analysis results and the mapping relationship between the cluster analysis results and the object treatment category label.

[0057] A possible implementation method, taking K-Means clustering as an example, performing cluster analysis based on the medical feature vector set to obtain the cluster analysis result of the target object includes the following steps: Step S1411: determining an initial cluster center according to the medical feature vectors included in the medical feature vector set.

[0058] Determining the initial cluster centers is a crucial first step in cluster analysis. A common approach is random selection, where several medical feature vectors are randomly selected from a set of medical feature vectors as initial cluster centers. In practice, each medical feature vector can be assigned a unique number. A random number generator is then used to generate a specified number of random numbers. These numbers are then used to select the corresponding medical feature vectors from the set of medical feature vectors as initial cluster centers.

[0059] Another more optimized method is the K-means++ algorithm. The specific steps of this algorithm are as follows: First, randomly select a medical feature vector as the first initial cluster center. Then, calculate the distance of other medical feature vectors to this initial cluster center, which can be calculated using Euclidean distance. For each medical feature vector x, its Euclidean distance to the initial cluster center c_1 is d(x, c_1) = √((x_1-c_1_1)^2+(x_2-c_1_2)^2+...+(x_n-c_1_n)^2), where x_i and c_1_i represent the values of the i-th dimension of the medical feature vector x and the initial cluster center c_1, respectively.

[0060] Next, select the medical feature vector farthest from the first initial cluster center as the second initial cluster center. Then, for the remaining medical feature vectors, calculate their shortest distances to all the selected initial cluster centers. For example, for medical feature vector x, calculate d_min(x) = min(d(x, c_1), d(x, c_2)). Then, based on these shortest distances, select the medical feature vector with the largest distance as the next initial cluster center. Repeat this process until the desired number of initial cluster centers have been selected.

[0061] Step S1412: Based on the treatment risk quantification index of the medical feature vector, obtain the medical weight coefficient corresponding to each medical feature vector.

[0062] In order to perform clustering analysis more accurately, it is necessary to consider the treatment risk quantification index of the medical feature vector. The evaluation of the treatment risk quantification index needs to comprehensively consider multiple factors, such as the severity of the target object's condition, the treatment difficulty, the likelihood of complications occurring, etc. The treatment risk of each medical feature vector can be scored by a group of experienced doctors through expert evaluation based on the information contained in the medical feature vector. For example, for a medical feature vector with a more severe condition, greater treatment difficulty, and a higher likelihood of complications, a higher score is given; for a medical feature vector with a less severe condition, smaller treatment difficulty, and a lower likelihood of complications, a lower score is given.

[0063] Then, according to these treatment risk quantification indexes, calculate the medical weight coefficients corresponding to each medical feature vector. The specific calculation method is to divide the treatment risk quantification index of each medical feature vector by the sum of the treatment risk quantification indexes of all medical feature vectors. Let the treatment risk quantification index of the medical feature vector x_i be r_i, and the sum of the treatment risk quantification indexes of all medical feature vectors be R = ∑r_j (j ranges from 1 to N, where N is the total number of medical feature vectors), then the medical weight coefficient w_i corresponding to the medical feature vector x_i is w_i = r_i / R.

[0064] Step S1413: According to each of the medical feature vectors and the medical weight coefficients corresponding to each of the medical feature vectors, determine the distances between each of the medical feature vectors and the initial clustering center, and generate an initial clustering result based on the distances between each of the medical feature vectors and the initial clustering center.

[0065] After determining the initial clustering center and the medical weight coefficients, it is necessary to calculate the distances between each medical feature vector and the initial clustering center. Here, the weighted Euclidean distance method is used for calculation. For the medical feature vector x and the initial clustering center c, its weighted Euclidean distance d_w(x, c) = √(w_1*(x_1 - c_1)^2 + w_2*(x_2 - c_2)^2 +... + w_n*(x_n - c_n)^2), where w_i is the medical weight coefficient corresponding to the i-th dimension of the medical feature vector x, and x_i and c_i are the values of the i-th dimension of the medical feature vector x and the initial clustering center c, respectively.

[0066] Based on the calculated weighted Euclidean distances, each medical feature vector is assigned to the cluster where the nearest initial clustering center is located. For example, for the medical feature vector x, if d_w(x, c_1) < d_w(x, c_2) <... < d_w(x, c_k) (k is the number of initial clustering centers), then the medical feature vector x is assigned to the cluster where the initial clustering center c_1 is located. In this way, an initial clustering result is generated.

[0067] Step S1414: Iteratively update the initial cluster center according to the initial clustering result, the medical feature vector of the target object, and the medical weight coefficient until a preset termination condition is met, thereby obtaining a cluster analysis result of the target object.

[0068] Based on the initial clustering results, the initial cluster centers need to be iteratively updated. For each cluster, the weighted average of all medical feature vectors in that cluster is calculated and used as the new cluster center. The specific calculation method is as follows: suppose there are m medical feature vectors x_1, x_2, ..., x_m in a cluster, and the corresponding medical weight coefficients are w_1, w_2, ..., w_m, respectively. Then, the value of the i-th dimension of the new cluster center c_new is c_new_i = (∑(w_j*x_j_i)) / (∑w_j) (where j ranges from 1 to m).

[0069] Then, the weighted Euclidean distance between each medical feature vector and the new cluster center is recalculated, and the medical feature vectors are redistributed to different clusters based on the distance to obtain a new clustering result. This iterative update process is repeated until a preset termination condition is met. The preset termination condition can include the cluster center no longer changing, that is, the difference between the new cluster center and the previous cluster center in each dimension is less than a minimum threshold; or the maximum number of iterations is reached to prevent the algorithm from falling into an infinite loop. When the preset termination condition is met, the cluster analysis result of the target object is obtained.

[0070] Another possible implementation method, taking hierarchical clustering as an example, performing cluster analysis based on the medical feature vector set to obtain the cluster analysis result of the target object includes the following steps: Step S1421: constructing an initial state of hierarchical clustering based on the distribution characteristics of each medical feature vector in the medical feature vector set, wherein each medical feature vector in the initial state is an independent initial cluster.

[0071] In this embodiment, when using a hierarchical clustering method for cluster analysis, it is first necessary to construct an initial state for the hierarchical clustering. Based on the distribution characteristics of each medical feature vector in the medical feature vector set, each medical feature vector is considered an independent initial cluster. In this way, the initial state contains the same number of independent initial clusters as the number of medical feature vectors.

[0072] Step S1422: Based on the treatment risk quantification index of the medical feature vector, obtain the medical weight coefficient corresponding to each medical feature vector.

[0073] This step can refer to the aforementioned step S1412 and will not be repeated here.

[0074] Step S1423: Calculating the weighted distances between the independent initial clusters based on the medical feature vectors and the medical weight coefficients corresponding to the medical feature vectors to generate a cluster distance matrix; After determining the medical weight coefficient, it is necessary to calculate the weighted distance between each independent initial cluster. For two independent initial clusters C_i and C_j, where C_i contains the medical feature vectors x_i1, x_i2, ..., x_im, and C_j contains the medical feature vectors x_j1, x_j2, ..., x_jn, the weighted distance between them is calculated by first calculating the weighted Euclidean distance between each medical feature vector in C_i and each medical feature vector in C_j. For medical feature vectors x_ik and x_jl, the weighted Euclidean distance is equal to the square root of the sum of (the first dimension value of the medical weight coefficient multiplied by (the first dimension value of x_ik minus the first dimension value of x_jl) squared plus the second dimension value of the medical weight coefficient multiplied by (the second dimension value of x_ik minus the second dimension value of x_jl), and so on, up to the p-th dimension value of the medical weight coefficient multiplied by (the p-th dimension value of x_ik minus the p-th dimension value of x_jl).

[0075] The weighted distance between the two independent initial clusters is calculated as the average of these weighted Euclidean distances. The specific calculation formula is: suppose there are two independent initial clusters C_i and C_j. The weighted distance between them is equal to (the sum of the weighted Euclidean distances between each medical feature vector in C_i and each medical feature vector in C_j) divided by (the number of medical feature vectors in C_i multiplied by the number of medical feature vectors in C_j).

[0076] After calculating the weighted distances between all independent initial clusters, a cluster distance matrix is generated. This matrix is a symmetric matrix, with rows and columns corresponding to each independent initial cluster, and the elements in the matrix represent the weighted distance between two independent initial clusters.

[0077] Step S1424: According to each weighted distance in the cluster distance matrix, iteratively merge the independent initial clusters whose weighted distance is less than or equal to the preset distance threshold until the number of clusters of the independent initial clusters reaches a preset number, thereby obtaining the cluster analysis result of the target object.

[0078] An iterative merge operation is performed based on the weighted distances in the cluster distance matrix. First, two independent initial clusters are found in the cluster distance matrix whose weighted distance is less than or equal to a preset distance threshold. The preset distance threshold is a pre-set parameter that controls the conditions for cluster merging. When two clusters C_i and C_j that meet the conditions are found, they are merged into a new cluster C_new.

[0079] The merged new cluster, C_new, contains all the medical feature vectors from the original two clusters. Next, we need to recalculate the weighted distances between the new cluster, C_new, and the remaining unmerged clusters. This calculation is similar to the weighted distance calculation between the two independent initial clusters. Specifically, we calculate the weighted Euclidean distance between each medical feature vector in C_new and each medical feature vector in the other clusters, and then take the average.

[0080] Update the cluster distance matrix, delete the rows and columns corresponding to the original clusters C_i and C_j, and add the rows and columns corresponding to the new cluster C_new. Repeat the above iterative merging process, constantly find clusters that meet the conditions for merging, and update the cluster distance matrix.

[0081] The iteration stops when the number of independent initial clusters reaches the preset number. The preset number is a parameter set according to actual needs, which determines the final number of clusters. At this point, the cluster analysis results of the target object are obtained.

[0082] Finally, the target patient's treatment category label is determined based on the cluster analysis results and a pre-established mapping between cluster analysis results and treatment category labels. This mapping can be established using historical data and expert experience, for example, by mapping a specific cluster to a specific treatment category label. If the target patient's medical feature vector is classified into a cluster, its corresponding treatment category label can be determined based on the mapping.

[0083] Method 2: Input the medical feature vector set into a pre-trained decision tree model to obtain the object treatment category label of the target object.

[0084] Before inputting the medical feature vector set into the decision tree model, the decision tree model needs to be trained. Figure 2 This is a flow chart of another method for recommending a treatment plan for respiratory failure based on artificial intelligence provided in an embodiment of the present application. Figure 2 As shown, before inputting the medical feature vector set into the decision tree model, the method may further include: Step S210: constructing an initial state of hierarchical clustering based on the distribution characteristics of each sample medical feature vector in the sample medical feature vector set of the sample object.

[0085] Wherein, each of the sample medical feature vectors in the initial state is an independent initial cluster.

[0086] To train the decision tree model, we first need to use the sample medical feature vectors of the sample subjects. Similar to the process of constructing the target subject's medical feature vectors, the sample medical feature vectors of the sample subjects are also obtained from the sample subject's multimodal medical data through preprocessing and feature extraction.

[0087] According to the distribution characteristics of each sample medical feature vector in the sample medical feature vector set, each sample medical feature vector is regarded as an independent initial cluster. In this way, in the initial state of hierarchical clustering, the number of independent initial clusters is equal to the number of sample medical feature vectors.

[0088] Step S220: Based on the treatment risk quantification index of the sample medical feature vector, obtain the medical weight coefficient corresponding to each of the sample medical feature vectors.

[0089] Similarly, to consider the treatment risk of each sample medical feature vector, it is necessary to obtain the corresponding medical weight coefficient for each sample medical feature vector. This can be achieved through expert evaluation, whereby a professional physician assigns a score to the treatment risk of each sample medical feature vector based on the information contained in the sample medical feature vector. This information includes the subject's physiological characteristics, underlying diseases, blood gas analysis results, and imaging features.

[0090] Then, divide the treatment risk quantification index of each sample medical feature vector by the sum of the treatment risk quantification indexes of all sample medical feature vectors to obtain the corresponding medical weight coefficient. Let the treatment risk quantification index of sample medical feature vector y_i be s_i, and the sum of the treatment risk quantification indexes of all sample medical feature vectors be S, which is equal to the sum of the treatment risk quantification indexes of each sample medical feature vector (i ranges from 1 to the total number of sample medical feature vectors). Then, the medical weight coefficient v_i corresponding to sample medical feature vector y_i is equal to s_i divided by S.

[0091] Step S230: Calculate the weighted distances between the independent initial clusters based on the sample medical feature vectors and the medical weight coefficients corresponding to the sample medical feature vectors to generate a sample cluster distance matrix.

[0092] After determining the medical weight coefficients, calculate the weighted distances between each independent initial cluster. For two independent initial clusters D_i and D_j, where D_i contains sample medical feature vectors y_i1, y_i2, ..., y_im, and D_j contains sample medical feature vectors y_j1, y_j2, ..., y_jn, first calculate the weighted Euclidean distance between each sample medical feature vector in D_i and each sample medical feature vector in D_j.

[0093] For the sample medical feature vectors y_ik and y_jl, their weighted Euclidean distance is equal to the square root of the sum of the squares of (the first dimension value of the medical weight coefficient multiplied by (the first dimension value of y_ik minus the first dimension value of y_jl) plus the second dimension value of the medical weight coefficient multiplied by (the second dimension value of y_ik minus the second dimension value of y_jl) and so on to the pth dimension value of the medical weight coefficient multiplied by (the pth dimension value of y_ik minus the pth dimension value of y_jl).

[0094] The weighted distance between the two independent initial clusters is then calculated as the average of these weighted Euclidean distances, which is calculated by adding the weighted Euclidean distances between each sample medical feature vector in D_i and each sample medical feature vector in D_j, and then dividing by (the number of sample medical feature vectors in D_i multiplied by the number of sample medical feature vectors in D_j).

[0095] After calculating the weighted distances between all independent initial clusters, a sample cluster distance matrix is generated. The rows and columns of the sample cluster distance matrix correspond to each independent initial cluster, and the elements in the sample cluster distance matrix represent the weighted distance between two independent initial clusters.

[0096] Step S240: according to each weighted distance in the sample cluster distance matrix, iteratively merge independent initial clusters whose weighted distance is less than or equal to a preset distance threshold until the number of clusters reaches a preset number, thereby obtaining an initial classification result of the sample object.

[0097] According to the weighted distance in the sample cluster distance matrix, an iterative merging operation is performed. Two independent initial clusters whose weighted distance is less than or equal to the preset distance threshold are found in the sample cluster distance matrix, and they are merged into a new cluster.

[0098] The merged new cluster contains all the sample medical feature vectors from the original two clusters. The weighted distance between the new cluster and other unmerged clusters is recalculated to update the sample cluster distance matrix.

[0099] Repeat the above iterative merging process, continuously search for clusters that meet the conditions for merging, and update the sample cluster distance matrix. When the number of independent initial clusters reaches the preset number, the iteration stops and the initial classification results of the sample objects are obtained.

[0100] Step S250: Obtaining the constraints of the initial decision tree model, wherein the constraints include a Gini coefficient constraint and a medical feature vector threshold constraint.

[0101] To train an appropriate decision tree model, it's necessary to obtain constraints for the initial decision tree model. The Gini coefficient constraint is used to measure the purity of a decision tree node, that is, the degree of consistency of the samples within the node. The smaller the Gini coefficient, the more consistent the categories of the samples within the node.

[0102] The medical feature vector threshold constraint is used to limit the value range of the medical feature vector used when splitting the decision tree. For example, for some medical feature vectors, their values may have a reasonable upper and lower limit range. When splitting the decision tree, the splitting threshold needs to be within this range.

[0103] Step S260: training the initial decision tree model according to the initial classification results and the constraint conditions to generate the decision tree model.

[0104] Specifically, this step may include the following sub-steps: Step S261: Obtain the classification label corresponding to each sample medical feature vector in the initial classification result, as well as the Gini coefficient constraint and the medical feature vector threshold constraint in the constraint conditions.

[0105] When training the initial decision tree model, the first step is to obtain the classification labels corresponding to each sample medical feature vector in the initial classification results. These classification labels represent the category to which the sample object belongs, such as different treatment categories.

[0106] At the same time, the Gini coefficient constraint and the medical feature vector threshold constraint are obtained from the constraints. The Gini coefficient constraint is a pre-set threshold used to control the purity of the decision tree nodes; the medical feature vector threshold constraint specifies the value range of the medical feature vector when the decision tree splits. This medical feature vector threshold constraint can be determined based on a large amount of clinical data. For example, when the target subject's PaO2 is less than 60 mmHg, the decision tree model can directly determine the target subject as a specific subject treatment category label (such as a high-risk label).

[0107] Step S262: Based on the classification label, traverse each medical feature vector in the sample medical feature vector set, calculate the Gini coefficient of the medical feature vector under different splitting thresholds, and select the splitting feature with the smallest Gini coefficient and the corresponding splitting threshold as the candidate splitting node.

[0108] Based on the classification label, each medical feature vector in the sample medical feature vector set is traversed. For each medical feature vector, different splitting thresholds are tried. For each splitting threshold, the sample medical feature vector set is divided into two subsets, and the Gini coefficient is calculated based on the sample distribution of these two subsets.

[0109] The Gini coefficient is calculated as follows: Assume a node contains N samples, of which n_c belong to category c. The Gini coefficient for this node is 1 minus the sum of the squares of the proportions of samples from each category c to the total number of samples. For each splitting threshold, the Gini coefficients of the two subsets after the split are calculated, and then a weighted average of the two subsets is taken based on the number of samples in each subset to obtain the Gini coefficient for that splitting threshold.

[0110] The split feature with the smallest Gini coefficient and the corresponding split threshold are selected and used as candidate split nodes. This process is to find the split method that can maximize the purity of the subset after the split.

[0111] Step S263: According to the medical feature vector threshold constraint, the splitting threshold in the candidate splitting node is normalized and checked. If the splitting threshold is within the numerical range of the normalized medical feature vector, the candidate splitting node is retained; otherwise, the candidate splitting node is eliminated.

[0112] Based on the medical feature vector threshold constraint, the splitting threshold in the candidate splitting node is normalized and verified. During the data preprocessing stage, the medical feature vector is normalized to keep its value range within a certain interval. The splitting threshold in the candidate splitting node is compared with the normalized value range of the medical feature vector.

[0113] If the splitting threshold is within this numerical range, the candidate splitting node is retained because the splitting threshold is reasonable; otherwise, the candidate splitting node is removed to ensure that the splitting threshold used when the decision tree splits is effective.

[0114] Step S264: Based on the retained candidate split nodes, generate a split priority queue by sorting them in ascending order of the Gini coefficient, select the candidate split node with the smallest Gini coefficient from the split priority queue as the current node splitting condition, and divide the sample medical feature vector set into two subsets.

[0115] Based on the retained candidate split nodes, they are sorted in ascending order of the Gini coefficient to generate a split priority queue. Candidate split nodes with smaller Gini coefficients are placed at the front of the queue, indicating that these split nodes can make the purity of the split subset higher.

[0116] A candidate split node with the smallest Gini coefficient is selected from the queue as the splitting condition for the current node. Based on this splitting condition, the set of sample medical feature vectors is divided into two subsets. For example, if the splitting condition is that a certain medical feature vector is greater than a certain splitting threshold, then the samples in the set of sample medical feature vectors with the medical feature vector greater than the splitting threshold are divided into one subset, and the samples with the medical feature vector less than or equal to the splitting threshold are divided into another subset.

[0117] Step S265: Recursively traverse the two subsets, repeat the steps of calculating the Gini coefficient, screening split features and split thresholds, checking the split thresholds, and generating a split priority queue until a preset stopping condition is met. The preset stopping condition includes that the Gini coefficient of the subset is zero, the split thresholds of all medical feature vectors do not meet the medical feature vector threshold constraint, or the maximum recursive depth is reached.

[0118] Recursively traverse the two subsets obtained by partitioning, and repeat the above steps of calculating the Gini coefficient, screening the splitting features and splitting thresholds, checking the splitting thresholds, and generating the splitting priority queue for each subset.

[0119] The recursion stops when the preset stopping conditions are met. These conditions include: the Gini coefficient of the subset is zero, indicating that the samples in the subset belong to the same category and no further splitting is required; the splitting thresholds of all medical feature vectors do not meet the medical feature vector threshold constraint, meaning that no suitable splitting method can be found; and the maximum recursion depth is reached to avoid overfitting caused by excessively deep decision trees.

[0120] Step S266: Generate a tree structure including classification rules and splitting thresholds based on all nodes after recursive splitting as the decision tree model.

[0121] After the recursive process is complete, a tree structure is generated based on all nodes after the recursive split. Each node contains a splitting feature, a splitting threshold, and a classification rule. The classification rule is determined based on the node's child nodes. For example, if a node splits into two child nodes, each corresponding to a different category, the node's classification rule can determine the category of the sample based on the value of the splitting feature.

[0122] These nodes are organized into a tree structure to form a decision tree model that includes classification rules and splitting thresholds. This decision tree model can be used to classify the medical feature vector set of the target object and determine its treatment category label.

[0123] Step S150: Determine a recommended treatment plan for the target subject based on the subject treatment category label.

[0124] After determining the target subject's treatment category label, a recommended treatment plan for the target subject is determined based on a pre-established mapping relationship between the target subject treatment category label and the recommended treatment plan. This mapping relationship can be established through expert experience, clinical research, and historical data.

[0125] For example, a specific treatment category label for a specific subject might correspond to a specific treatment plan, including medication and respiratory support. Medication treatment plans can select the appropriate medication type, dosage, and frequency of use based on the subject's condition and physical condition. Respiratory support treatment plans can include different methods, such as non-invasive and invasive ventilation.

[0126] For example, taking the object treatment category labels including low risk, medium risk, and high risk as an example, the recommended treatment plans corresponding to each object treatment category label can be: matching low-risk target objects with relatively mild recommended treatment plans such as non-invasive ventilation; matching medium-risk target objects with medium-intensity recommended treatment plans such as invasive mechanical ventilation; matching high-intensity recommended treatment plans such as extracorporeal membrane oxygenation (ECMO) for high-risk target objects.

[0127] In one possible implementation, after classifying the target object according to the medical feature vector set and determining the object treatment category label of the target object, the target object's condition can be monitored, and the object treatment category label of the target object can be dynamically adjusted according to the results of the condition monitoring to match a more suitable recommended treatment plan for the target object. Figure 3 This is a flow chart of another method for recommending a treatment plan for respiratory failure based on artificial intelligence provided in an embodiment of the present application. Figure 3 As shown, the method may further include: Step S310: Acquire quantitative disease monitoring data of the target object according to preset batch configuration parameters.

[0128] The quantitative data of disease monitoring include quantitative data of vital signs, quantitative data of blood gas analysis, and quantitative data of pathological trend.

[0129] The preset batch configuration parameters are pre-set parameters used to control the acquisition process of the disease monitoring quantitative data. According to these parameters, the disease monitoring quantitative data of the target object can be obtained from the relevant data source.

[0130] Quantitative vital sign data, such as heart rate, blood pressure, and respiratory rate, can be obtained from real-time monitoring equipment. This data reflects the subject's basic bodily function. Quantitative blood gas analysis data, obtained by testing a blood sample, includes indicators such as partial pressure of oxygen, partial pressure of carbon dioxide, and pH, directly reflecting the subject's respiratory function and acid-base balance.

[0131] Quantitative data on lesion trends can be obtained by analyzing multiple imaging examinations of the target subject. For example, by comparing chest CT scans taken at different times, changes in the size, number, and density of lung lesions can be analyzed to quantify the lesion's development trend.

[0132] Step S320: Obtaining the progress weights corresponding to the vital sign quantitative data, the blood gas analysis quantitative data, and the pathological trend quantitative data respectively.

[0133] Specifically, the specific process of this step is as follows: Step S321: Obtain a set of sample medical feature vectors of each sample patient from the pre-constructed disease progression history data of multiple sample patients, wherein each sample medical feature vector in the set of sample medical feature vectors includes vital sign features, blood gas analysis features, and pathological trend features, and has been normalized.

[0134] Pre-built historical disease progression data for multiple sample patients contains a large amount of information about the patient's condition at different times. From this data, a set of sample medical feature vectors is extracted for each sample patient. Each sample medical feature vector includes vital sign features, blood gas analysis features, and pathological trend features.

[0135] During the extraction process, these features are normalized to eliminate scale differences between different features. Normalization can use methods such as minimum-maximum normalization or mean-standard deviation normalization to ensure that the value range of all features is within a reasonable range.

[0136] Step S322: extracting the disease exacerbation index of the sample patient as a regression target variable, wherein the disease exacerbation index is quantitatively generated according to the occurrence time and severity of a preset exacerbation event.

[0137] The exacerbation index (ICI) is extracted from the patient's historical disease progression data as the regression target variable. Pre-defined exacerbation events can include worsening respiratory failure and the need for invasive ventilation. The ICI is quantified based on the timing and severity of these exacerbation events.

[0138] For example, different weights can be assigned to different exacerbation events, and a weighted score can be calculated based on the occurrence time and severity of the exacerbation event to serve as the disease exacerbation index.

[0139] Step S323: Using the vital sign characteristics, the blood gas analysis characteristics, and the pathological trend characteristics as independent variables, a linear regression model or a logistic regression model is fitted, and the regression coefficient corresponding to each characteristic is output.

[0140] Normalized vital sign characteristics, blood gas analysis characteristics, and lesion trend characteristics were used as independent variables, and the disease exacerbation index was used as the dependent variable to fit a linear regression model or a logistic regression model.

[0141] The general form of a linear regression model is that the dependent variable is equal to a constant term plus each independent variable multiplied by its corresponding regression coefficient, and then added together. A logistic regression model is used to handle binary classification problems, such as determining whether a disease is worsening. By fitting the model, the regression coefficient corresponding to each feature can be obtained.

[0142] Step S324: Calculate the importance index of each feature based on the absolute value of the regression coefficient of each feature.

[0143] The importance index of each feature is calculated based on the absolute value of the regression coefficient of each feature. The larger the absolute value of the regression coefficient, the greater the impact of the feature on the disease progression index, and the higher its importance index.

[0144] Assuming that the regression coefficients of vital signs characteristics, blood gas analysis characteristics, and pathological trend characteristics are the corresponding regression coefficient values, then their importance indicators are equal to the absolute values of their respective regression coefficients.

[0145] Step S325: normalizing the importance index so that the sum of the progress weights of the features is 1, and generating progress weights corresponding to the vital sign quantitative data, the blood gas analysis quantitative data, and the pathological trend quantitative data.

[0146] The importance index of each feature is normalized to convert them into progress weights. The specific method is to divide the importance index of each feature by the sum of the importance indexes of all features.

[0147] Assume that the progression weights corresponding to the quantitative vital sign data, quantitative blood gas analysis data, and quantitative lesion trend data are their corresponding weight values. Then, the progression weight corresponding to the quantitative vital sign data is equal to the importance index of the vital sign feature divided by (the importance index of the vital sign feature plus the importance index of the blood gas analysis feature plus the importance index of the lesion trend feature), the progression weight corresponding to the quantitative blood gas analysis data is equal to the importance index of the blood gas analysis feature divided by (the importance index of the vital sign feature plus the importance index of the blood gas analysis feature plus the importance index of the lesion trend feature), and the progression weight corresponding to the quantitative lesion trend data is equal to the importance index of the lesion trend feature divided by (the importance index of the vital sign feature plus the importance index of the blood gas analysis feature plus the importance index of the lesion trend feature). Through this calculation, the progression weights corresponding to the quantitative vital sign data, quantitative blood gas analysis data, and quantitative lesion trend data are generated. The sum of these weights is 1, which can reasonably reflect the importance of each piece of data in assessing disease progression.

[0148] Step S330: determining a quantitative index of the disease progression of the target subject based on an attention mechanism according to the quantitative data of vital signs, the quantitative data of blood gas analysis, the quantitative data of lesion trend, and the progression weight.

[0149] The attention mechanism allows the model to focus more on data that has a greater impact on disease progression. When determining the quantitative indicators of the target patient's disease progression, the first thing to consider is the relationship between each quantitative data and the corresponding progression weight.

[0150] For quantified vital sign data, each dimension is associated with a vital sign progression weight. For example, if the quantified vital sign data includes multiple dimensions such as heart rate, blood pressure, and respiratory rate, the data for each dimension is multiplied by the vital sign progression weight. Assuming the quantified vital sign data is a vector containing multiple dimensional values, each dimension in the vector is multiplied by the vital sign progression weight to obtain the weighted quantified vital sign data vector.

[0151] Similarly, for the blood gas analysis quantitative data, it is also a vector containing multiple dimensional values. Each dimensional value in the vector is multiplied by the blood gas analysis progress weight to obtain the blood gas analysis quantitative data vector after the weighting.

[0152] For the lesion trend quantitative data, which is also a vector containing multiple dimensional values, each dimensional value in the vector is multiplied by the lesion trend progression weight to obtain the lesion trend quantitative data vector after the weight is applied.

[0153] These weighted data are then integrated using an attention mechanism. The attention mechanism dynamically adjusts the level of attention given to different data based on their importance to disease progression in different situations. An attention function is used to calculate the attention coefficient for each data dimension. The attention function calculates the corresponding attention coefficient for each dimension in the weighted vector of quantified vital sign data. The attention function also calculates the corresponding attention coefficient for each dimension in the weighted vector of quantified blood gas analysis data and quantified lesion trend data.

[0154] Finally, the attention-adjusted data is aggregated to derive a quantitative indicator of disease progression. Specifically, the weighted and attention-adjusted values of the vital sign quantitative data are summed up to obtain a summary value. The weighted and attention-adjusted values of the blood gas analysis quantitative data are summed up to obtain a summary value. Finally, the weighted and attention-adjusted values of the lesion trend quantitative data are summed up to obtain a summary value. These three summed values are then added together to obtain a quantitative indicator of disease progression for the target patient.

[0155] Step S340: determining whether the treatment category label of the target object needs to be adjusted based on the disease progression quantification indicator, the current treatment category label of the target object, and the level conversion threshold corresponding to the disease progression quantification indicator.

[0156] Specifically, the specific process of this step can be shown as follows: Step S341: obtaining a preset level conversion threshold range associated with the current object treatment category label, wherein the preset level conversion threshold range includes an upper limit of a deterioration threshold and a lower limit of an improvement threshold.

[0157] Different treatment category labels correspond to different preset threshold ranges for transitioning to a different level. These threshold ranges are pre-set based on extensive clinical data and expert experience. A mapping table can be used to store the correspondence between treatment category labels and preset threshold ranges for transitioning to a different level. Based on the target subject's current treatment category label, the associated upper and lower deterioration thresholds are retrieved from this mapping table.

[0158] Step S342: performing a first comparison between the disease progression quantification indicator and the upper deterioration threshold, and generating a category increase instruction if the disease progression quantification indicator is greater than or equal to the upper deterioration threshold.

[0159] The calculated quantitative indicator of disease progression is compared with the upper threshold for deterioration. If the quantitative indicator is greater than or equal to the upper threshold for deterioration, this indicates that the target subject's condition may be deteriorating and requires more aggressive treatment. In this case, a category upgrade instruction can be generated, indicating that the target subject's treatment category label should be adjusted to a higher treatment category.

[0160] Step S343: performing a second comparison between the disease progression quantitative index and the improvement threshold lower limit, and generating a category downgrade instruction if the disease progression quantitative index is less than or equal to the improvement threshold lower limit.

[0161] Next, the quantitative indicator of disease progression is compared to the lower threshold for improvement. If the quantitative indicator is less than or equal to the lower threshold, it indicates that the target patient's condition has significantly improved, and the original aggressive treatment measures may not be necessary. In this case, a downgrade instruction can be generated, indicating that the target patient's treatment category label should be adjusted to a lower treatment category.

[0162] Step S344: determining the adjusted object treatment category label of the target object according to the category-up instruction or the category-down instruction, and verifying whether the adjusted object treatment category label is within a preset label level range.

[0163] Based on the generated category adjustment instruction or category adjustment instruction, the target subject's treatment category label is adjusted. If it is a category adjustment instruction, the treatment category label of the target subject is adjusted to a higher treatment category; if it is a category adjustment instruction, the treatment category label of the target subject is adjusted to a lower treatment category.

[0164] After the adjustment is complete, you need to verify that the adjusted treatment category label falls within the preset label range. The preset label range specifies the reasonable range of values for the treatment category label and helps prevent unreasonable label adjustments. You can use a preset label range list to store the preset label ranges and check whether the adjusted treatment category label falls within this list.

[0165] Step S345: If the adjusted object treatment category label exceeds the preset label level range, the current object treatment category label is maintained unchanged; otherwise, the adjusted object treatment category label is used as the updated object treatment category label.

[0166] If the adjusted treatment category label exceeds the preset label range, this indicates that the adjustment is unreasonable. In this case, the current treatment category label is maintained to ensure the stability and rationality of the treatment plan. If the adjusted treatment category label is within the preset label range, it is used as the updated treatment category label for subsequent treatment plan recommendations.

[0167] Step S350: If the subject treatment category label of the target subject needs to be adjusted, a recommended treatment plan for the target subject is determined based on the adjusted subject treatment category label.

[0168] If the previous steps determine that the target subject's treatment category label needs to be adjusted, the recommended treatment plan for the target subject is re-determined based on the adjusted treatment category label. Again, based on the pre-established mapping relationship between treatment category labels and recommended treatment plans, the recommended treatment plan corresponding to the adjusted treatment category label is found from this mapping relationship.

[0169] For example, if the adjusted treatment category label corresponds to a higher-level treatment plan, the medication dosage might be increased, a more effective medication might be switched, or more aggressive respiratory support might be employed. If the adjusted treatment category label corresponds to a lower-level treatment plan, medication use might be reduced, the intensity of respiratory support might be lowered, and so on. In this way, the recommended treatment plan can be dynamically adjusted based on the target patient's disease progression, thereby improving the effectiveness and targeted nature of treatment.

[0170] Throughout the entire data collection process, various privacy protection and anti-leakage technologies can be employed for potentially privacy-sensitive data, such as patients' personal identity information and detailed medical records. During the data collection phase, data can be encrypted using symmetric or asymmetric encryption algorithms to convert the data into ciphertext, which can only be decrypted and viewed by authorized personnel. For data storage, a secure storage system should be employed with strict access control to ensure that only authorized personnel have access. Furthermore, data should be backed up regularly to prevent loss or corruption. During data transmission, secure network protocols, such as SSL / TLS, should be used to ensure data security during transmission. These technologies can effectively protect patients' privacy-sensitive data and prevent data leakage and misuse.

[0171] The above examples describe in detail the various steps and implementation details of the AI-based respiratory failure treatment recommendation method, enabling those skilled in the art to implement the present invention's solution based on these descriptions. This process begins with acquiring multimodal medical data from the target subject, then proceeds through data preprocessing, feature extraction, and classification to determine the subject's treatment category label. This process then dynamically adjusts the label and determines the recommended treatment plan based on the progression of the disease. This complete closed-loop process provides more accurate and personalized treatment recommendations for patients with respiratory failure.

[0172] Figure 4 This is a schematic diagram of the structure of an artificial intelligence-based respiratory failure treatment plan recommendation system 100 provided in an embodiment of the present application. Figure 4 As shown, the processor 120 can be used in the artificial intelligence-based respiratory failure treatment plan recommendation system 100 and used to perform the functions of the present invention.

[0173] The AI-based respiratory failure treatment plan recommendation system 100 can be a general-purpose server or a special-purpose server, both of which can be used to implement the AI-based respiratory failure treatment plan recommendation method of the present invention. Although only one server is shown in the present invention, for convenience, the functions described in the present invention can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0174] For example, the artificial intelligence-based respiratory failure treatment plan recommendation system 100 may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and storage media 140 in different forms, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the artificial intelligence-based respiratory failure treatment plan recommendation system 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present invention can be implemented according to these program instructions. The artificial intelligence-based respiratory failure treatment plan recommendation system 100 also includes an input / output (I / O) interface 150 between the computer and other input and output devices.

[0175] For ease of explanation, only one processor is described in the artificial intelligence-based respiratory failure treatment plan recommendation system 100. However, it should be noted that the artificial intelligence-based respiratory failure treatment plan recommendation system 100 in the present invention may also include multiple processors, so the steps performed by one processor described in the present invention may also be performed jointly or individually by multiple processors. For example, if the processor of the artificial intelligence-based respiratory failure treatment plan recommendation system 100 executes step A and step B, it should be understood that step A and step B may also be performed jointly by two different processors or individually in one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor execute steps A and B together.

[0176] An embodiment of the present invention discloses a computer-readable storage medium storing a computer program for electronic data exchange, wherein the computer program enables a computer to execute the steps of the method for recommending a respiratory failure treatment plan based on artificial intelligence described in the aforementioned embodiment.

[0177] An embodiment of the present invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to enable a computer to execute the steps of the artificial intelligence-based respiratory failure treatment plan recommendation method described in the aforementioned embodiment.

[0178] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, i.e., they may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0179] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus the necessary general hardware platform, or of course, by means of hardware. Based on this understanding, the above technical solution, in essence, or the portion that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, including a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM), or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium capable of storing or storing data.

[0180] Finally, it should be noted that what is disclosed above is only a preferred embodiment of the present invention, which is only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for recommending treatment options for respiratory failure based on artificial intelligence, characterized in that: include: Acquire first multimodal medical collected data of the target object, where the first multimodal medical collected data includes text modality data and image modality data, and the text modality data includes physiological characteristic data, basic disease data, and blood gas analysis data; Performing data preprocessing on the first multimodal medical collected data to obtain second multimodal medical collected data, wherein the data preprocessing includes data cleaning, normalization, and time window alignment; extracting a medical feature vector set of the target object from the second multimodal medical acquisition data; performing classification processing on the target object according to the medical feature vector set to determine an object treatment category label of the target object; Determine a recommended treatment plan for the target subject based on the subject's treatment category label.

2. The method for recommending a treatment plan for respiratory failure based on artificial intelligence according to claim 1, characterized in that: The classifying process of the target object according to the medical feature vector set to determine the treatment category label of the target object includes: Performing a cluster analysis based on the medical feature vector set to obtain a cluster analysis result of the target object, and determining an object treatment category label of the target object based on the cluster analysis result and a mapping relationship between the cluster analysis result and the object treatment category label; Alternatively, the medical feature vector set is input into a pre-trained decision tree model to obtain the object treatment category label of the target object.

3. The method for recommending a treatment plan for respiratory failure based on artificial intelligence according to claim 2, characterized in that: The performing cluster analysis based on the medical feature vector set to obtain the cluster analysis result of the target object includes: determining an initial cluster center according to the medical feature vectors included in the medical feature vector set; Based on the treatment risk quantification index of the medical feature vector, obtaining the medical weight coefficient corresponding to each medical feature vector; Determining the distance between each medical feature vector and the initial cluster center according to each medical feature vector and the medical weight coefficient corresponding to each medical feature vector, and generating an initial clustering result based on the distance between each medical feature vector and the initial cluster center; According to the initial clustering result, the medical feature vector of the target object and the medical weight coefficient, the initial cluster center is iteratively updated until a preset termination condition is met, thereby obtaining a cluster analysis result of the target object.

4. The method for recommending a treatment plan for respiratory failure based on artificial intelligence according to claim 2, characterized in that: The performing cluster analysis based on the medical feature vector set to obtain the cluster analysis result of the target object includes: Based on the distribution characteristics of each medical feature vector in the medical feature vector set, constructing an initial state of hierarchical clustering, wherein each medical feature vector in the initial state is an independent initial cluster; Based on the treatment risk quantification index of the medical feature vector, obtaining the medical weight coefficient corresponding to each medical feature vector; Calculating the weighted distances between the independent initial clusters based on the medical feature vectors and the medical weight coefficients corresponding to the medical feature vectors to generate a cluster distance matrix; According to each weighted distance in the cluster distance matrix, independent initial clusters whose weighted distance is less than or equal to a preset distance threshold are iteratively merged until the number of clusters of the independent initial clusters reaches a preset number, thereby obtaining a cluster analysis result of the target object.

5. The method for recommending a treatment plan for respiratory failure based on artificial intelligence according to claim 2, characterized in that: Before performing classification processing on the target object according to the medical feature vector set and determining the treatment category label of the target object, the method further includes a training method for the decision tree model, specifically including: constructing an initial state of hierarchical clustering based on the distribution characteristics of each sample medical feature vector in the sample medical feature vector set of the sample object, wherein each sample medical feature vector in the initial state is an independent initial cluster; Based on the treatment risk quantification index of the sample medical feature vector, obtaining the medical weight coefficient corresponding to each of the sample medical feature vectors; Calculating the weighted distances between the independent initial clusters based on the sample medical feature vectors and the medical weight coefficients corresponding to the sample medical feature vectors to generate a sample cluster distance matrix; According to each weighted distance in the sample cluster distance matrix, iteratively merge independent initial clusters whose weighted distance is less than or equal to a preset distance threshold until the number of clusters reaches a preset number, thereby obtaining an initial classification result of the sample object; Obtaining constraints of an initial decision tree model, wherein the constraints include a Gini coefficient constraint and a medical feature vector threshold constraint; The initial decision tree model is trained according to the initial classification results and the constraint conditions to generate the decision tree model.

6. The method for recommending a treatment plan for respiratory failure based on artificial intelligence according to claim 5, characterized in that: The step of training the initial decision tree model according to the initial classification results and the constraint conditions to generate the decision tree model includes: Obtaining the classification label corresponding to each sample medical feature vector in the initial classification result, as well as the Gini coefficient constraint and the medical feature vector threshold constraint in the constraint conditions; Based on the classification label, traverse each medical feature vector in the sample medical feature vector set, calculate the Gini coefficient of the medical feature vector under different splitting thresholds, and select the splitting feature with the smallest Gini coefficient and the corresponding splitting threshold as a candidate splitting node; According to the medical feature vector threshold constraint, a normalization check is performed on the splitting threshold in the candidate splitting node; if the splitting threshold is within the numerical range of the normalized medical feature vector, the candidate splitting node is retained; otherwise, the candidate splitting node is eliminated; Based on the retained candidate split nodes, sorting them in ascending order of the Gini coefficient to generate a split priority queue, selecting the candidate split node with the smallest Gini coefficient from the split priority queue as the current node split condition, and dividing the sample medical feature vector set into two subsets; Recursively traversing the two subsets, repeatedly calculating the Gini coefficient, screening the splitting features and the splitting threshold, verifying the splitting threshold, and generating the splitting priority queue until a preset stopping condition is met, wherein the preset stopping condition includes that the Gini coefficient of the subset is zero, the splitting thresholds of all medical feature vectors do not meet the medical feature vector threshold constraint, or the maximum recursion depth is reached; A tree structure including classification rules and splitting thresholds is generated according to all nodes after recursive splitting as the decision tree model.

7. The method for recommending a treatment plan for respiratory failure based on artificial intelligence according to claim 1, characterized in that: After classifying the target object according to the medical feature vector set and determining the treatment category label of the target object, the method further includes: Acquire the target subject's disease monitoring quantitative data according to preset batch configuration parameters, wherein the disease monitoring quantitative data includes vital sign quantitative data, blood gas analysis quantitative data, and pathological trend quantitative data; Obtaining progression weights corresponding to the vital sign quantitative data, the blood gas analysis quantitative data, and the pathological trend quantitative data respectively; Determining a quantitative indicator of the target subject's disease progression based on an attention mechanism according to the quantitative vital sign data, the quantitative blood gas analysis data, the quantitative lesion trend data, and the progression weight; determining whether it is necessary to adjust the subject treatment category label of the target subject based on the disease progression quantification indicator, the current subject treatment category label of the target subject, and a level conversion threshold corresponding to the disease progression quantification indicator; If the subject treatment category label of the target subject needs to be adjusted, a recommended treatment regimen for the target subject is determined based on the adjusted subject treatment category label.

8. The method for recommending a treatment plan for respiratory failure based on artificial intelligence according to claim 7, characterized in that: The obtaining of the progress weights corresponding to the vital sign quantitative data, the blood gas analysis quantitative data, and the pathological trend quantitative data respectively includes: Obtaining a sample medical feature vector set of each sample patient from pre-constructed disease progression history data of multiple sample patients, wherein each sample medical feature vector in the sample medical feature vector set includes vital sign features, blood gas analysis features, and pathological trend features, and has been normalized; Extracting the disease exacerbation index of the sample patient as a regression target variable, wherein the disease exacerbation index is quantitatively generated according to the occurrence time and severity of a preset exacerbation event; Using the vital sign characteristics, the blood gas analysis characteristics, and the pathological trend characteristics as independent variables, fitting a linear regression model or a logistic regression model, and outputting the regression coefficient corresponding to each characteristic; Calculate the importance index of each feature based on the absolute value of the regression coefficient of each feature; The importance index is normalized so that the sum of the progress weights of the features is 1, and progress weights corresponding to the vital sign quantitative data, the blood gas analysis quantitative data, and the pathological trend quantitative data are generated.

9. The method for recommending a treatment plan for respiratory failure based on artificial intelligence according to claim 7, characterized in that: The determining whether it is necessary to adjust the subject treatment category label of the target subject based on the disease progression quantification indicator, the current subject treatment category label of the target subject, and the level conversion threshold corresponding to the disease progression quantification indicator includes: Obtaining a preset level conversion threshold range associated with the current subject treatment category label, wherein the preset level conversion threshold range includes an upper deterioration threshold limit and a lower improvement threshold limit; performing a first comparison between the disease progression quantification indicator and the upper deterioration threshold, and generating a category upward adjustment instruction if the disease progression quantification indicator is greater than or equal to the upper deterioration threshold; Performing a second comparison between the disease progression quantification index and the improvement threshold lower limit, and generating a category downgrade instruction if the disease progression quantification index is less than or equal to the improvement threshold lower limit; determining an adjusted subject treatment category label for the target subject according to the category-up instruction or the category-down instruction, and verifying whether the adjusted subject treatment category label is within a preset label hierarchy range; If the adjusted object treatment category label exceeds the preset label level range, the current object treatment category label is maintained unchanged; otherwise, the adjusted object treatment category label is used as the updated object treatment category label.

10. An artificial intelligence-based respiratory failure treatment plan recommendation system, characterized in that: It includes a processor and a computer-readable storage medium, wherein the computer-readable storage medium stores machine-executable instructions, and when the machine-executable instructions are executed by a computer, the method for recommending a respiratory failure treatment plan based on artificial intelligence according to any one of claims 1 to 9 is implemented.

Images