Hemorrhagic stroke risk assessment method, system, storage medium and device

Abstract

The present invention relates to a hemorrhagic stroke risk assessment method, system, storage medium and device. Based on brain images and corresponding patient information and treatment plans, combined with prognosis data, a risk assessment is performed on the risk of hematoma expansion, edema changes and patient recovery status under the influence of different treatment plans. The method can make full use of imaging data, clinical information and the impact of treatment interventions on the volume changes of hematomas and edema involved in hemorrhagic stroke to perform risk assessment and analysis, which is more objective than relying on the subjective experience of doctors.

Description

Technical Field

[0001] The present invention relates to the technical field of image and text processing, and in particular to a method, system, storage medium and device for assessing the risk of hemorrhagic stroke. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Stroke is a disease that causes damage to brain tissue, including ischemic and hemorrhagic types. Hemorrhagic stroke has no history of trauma, but instead causes bleeding in the skull and brain parenchyma to form hematomas and the accompanying edema that damage the patient's brain nerves.

[0004] Existing technologies use some computer models to identify biomarker features and imaging features related to hemorrhagic stroke from brain image data, thereby determining the location, shape and size of hematoma and edema. Hematoma and edema will change dynamically due to the patient's individual constitution and the intervention of different treatment methods. Doctors use the changes in hematoma and edema obtained from brain image data, combined with diagnosis and treatment data and personal experience to judge the patient's possible recovery or corresponding risks in the future. Computer models are limited by training data sets, resulting in low recognition accuracy when processing data from different sources, different devices or different populations, making it difficult to determine the changing state of hematoma and edema. Accordingly, doctors rely too much on subjective experience when using the changing state of hematoma and edema to assess the risk of hemorrhagic stroke in patients. Summary of the invention

[0005] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a hemorrhagic stroke risk assessment method, system, storage medium and device, which perform risk assessment on the risk of hematoma expansion and edema changes under the influence of different treatment plans based on brain images and corresponding patient information and treatment plans, combined with prognosis data.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] The first aspect of the present invention provides a method for assessing the risk of hemorrhagic stroke, comprising the following steps:

[0008] Acquire brain image data and the diagnosis and treatment history data of the corresponding patient, extract the volume, location, shape and distribution information of hematoma and edema corresponding to the examination time point based on the brain image data, associate it with the diagnosis and treatment history data, form a data set and preprocess it;

[0009] The probability of hematoma expansion events in patients is predicted by using the increase in absolute or relative hematoma volume and the corresponding examination time points in the preprocessed data set;

[0010] The edema volume and the corresponding examination time points in the preprocessed data set were used to obtain the curve of edema volume change over time. After post-processing and fitting with the set treatment method code, the influence curve of different treatment methods on edema volume change was obtained.

[0011] The morphological characteristics of hematoma and edema in the preprocessed data set and the corresponding patient's diagnosis and treatment data are used to predict the risk level of patients affected by hematoma and edema in the future time period based on the set scoring rules.

[0012] A second aspect of the present invention provides a hemorrhagic stroke risk assessment system, comprising:

[0013] The data set construction module is configured to: obtain brain image data and diagnosis and treatment history data of the corresponding patient, extract the volume, position, shape and distribution information of hematoma and edema corresponding to the examination time point based on the brain image data, associate it with the diagnosis and treatment history data, form a data set and pre-process it;

[0014] The hematoma expansion risk prediction module is configured to: predict the probability of a hematoma expansion event occurring in a patient by using the increase in the absolute volume or relative volume of the hematoma and the corresponding examination time point in the preprocessed data set;

[0015] The edema volume change module is configured to: obtain a curve of edema volume change over time by using the edema volume and the corresponding examination time point in the preprocessed data set, and obtain the influence curve of different treatment methods on edema volume change by post-processing and fitting with the set treatment method code;

[0016] The prognosis risk prediction module is configured to: use the morphological characteristics of hematoma and edema in the preprocessed data set and the corresponding patient's diagnosis and treatment data, and based on the set scoring rules, predict the risk level of patients affected by hematoma and edema in the future time period.

[0017] Furthermore, the patient's medical history data is obtained by extracting text data from the patient's medical record to obtain the patient's personal information, medical history, and medical history.

[0018] Furthermore, the trained prediction model is used to predict the probability of a hematoma expansion event occurring in a patient according to the increase in the absolute volume or relative volume of the hematoma between set examination time points.

[0019] Furthermore, the edema volume and the corresponding examination time point in the preprocessed data set are used to obtain a curve of the edema volume changing with time. Specifically, the edema volume and the examination time point are used as coordinate axes to obtain a curve of the edema volume changing with time.

[0020] Furthermore, post-processing includes outlier detection, fitting processing, clustering and grouping.

[0021] Furthermore, the treatment method code corresponding to the patient is determined according to the diagnosis and treatment information in the data set, and the edema volume is sorted to obtain a curve of the edema volume changing over time under the influence of different treatment methods.

[0022] Furthermore, the Pearson correlation coefficient method was used to screen out features in the data set whose diagnosis and treatment information scores exceeded the set values, and the trained prediction model and the set scoring rules were used to predict the risk level of patients affected by hematoma and edema in future time periods.

[0023] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the above-mentioned hemorrhagic stroke risk assessment method.

[0024] The fourth aspect of the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps in the above-mentioned hemorrhagic stroke risk assessment method when executing the program.

[0025] Compared with the prior art, one or more of the above technical solutions have the following beneficial effects:

[0026] 1. Organically combine brain image data and medical history data, extract information such as the volume, location, shape, etc. of hematoma and edema, and associate it with medical history data to form a comprehensive data set for preprocessing, providing a more comprehensive and accurate basis for assessing the risk of hemorrhagic stroke.

[0027] 2. The modular construction of the system makes the entire assessment process more efficient and easier to manage. The modules are closely linked, which is conducive to data flow and information exchange, and helps the system's medical staff better understand the patient's risk situation and develop more personalized and targeted treatment plans.

[0028] 3. The use of data processing and model training technology can accurately predict the probability of hematoma expansion events in patients and the curve of edema volume changes over time. This quantitative prediction analysis provides doctors with more objective basis, helps to adjust treatment plans in time and improve treatment effects.

[0029] 4. Converting existing treatment interventions into mathematical codes and analyzing the effects of different treatments on the progression pattern of edema volume will help reveal the mechanism of action of different treatment strategies on disease progression and provide guidance for clinical selection of the best treatment plan.

[0030] 5. Combined with imaging results, edema, hematoma manifestations, and clinical and treatment data, scoring rules are used to predict the patient's risk level in the future, providing doctors with reference information on the patient's long-term prognosis, helping doctors to fully understand the patient's condition and make more scientific and reasonable clinical decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0032] Figure 1 is a schematic diagram of ROC curve comparison provided by one or more embodiments of the present invention;

[0033] Figure 2 is a schematic diagram for comparing fitting curves provided by one or more embodiments of the present invention;

[0034] Figure 3 is a schematic diagram for comparing the type 1 fitting curves provided by one or more embodiments of the present invention;

[0035] Figure 4 is a schematic diagram for comparing type 2 fitting curves provided by one or more embodiments of the present invention;

[0036] Figure 5 is a schematic diagram for comparing Class 3 fitting curves provided by one or more embodiments of the present invention;

[0037] Figure 6 is a schematic diagram for comparing Class 4 fitting curves provided by one or more embodiments of the present invention;

[0038] Figure 7 is a schematic diagram of edema volume trend of diagnosis and treatment combination 0000111 provided by one or more embodiments of the present invention;

[0039] Figure 8 It is a schematic diagram of the edema volume trend of the diagnosis and treatment combination 0001011 provided in one or more embodiments of the present invention. DETAILED DESCRIPTION

[0040] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0041] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0042] The following embodiments provide a method, system, storage medium and device for assessing the risk of hemorrhagic stroke, which predicts the risk of hematoma expansion based on brain images, corresponding patient information and treatment plans, combined with prognostic data.

[0043] Embodiment 1:

[0044] like Figure 1-Figure 8 As shown, the hemorrhagic stroke risk assessment system includes:

[0045] The data set construction module is configured to: obtain brain image data and diagnosis and treatment history data of the corresponding patient, extract the volume, position, shape and distribution information of hematoma and edema corresponding to the examination time point based on the brain image data, associate it with the diagnosis and treatment history data, form a data set and pre-process it;

[0046] The hematoma expansion risk prediction module is configured to: predict the probability of a hematoma expansion event occurring in a patient by using the increase in the absolute volume or relative volume of the hematoma and the corresponding examination time point in the preprocessed data set;

[0047] The edema volume change module is configured to: obtain a curve of edema volume change over time by using the edema volume and the corresponding examination time point in the preprocessed data set, and obtain the influence curve of different treatment methods on edema volume change by post-processing and fitting with the set treatment method code;

[0048] The prognosis risk prediction module is configured to: use the morphological characteristics of hematoma and edema in the preprocessed data set and the corresponding patient's diagnosis and treatment data, and based on the set scoring rules, predict the risk level of patients affected by hematoma and edema in the future time period.

[0049] This embodiment is used to solve the following problems:

[0050] 1) Factors related to the risk of hematoma expansion in patients with hemorrhagic stroke and the impact of therapeutic interventions on the condition.

[0051] 2) Changes in the patient's edema volume over time.

[0052] 3) The effects of different treatments on the progression pattern of edema volume and the relationship between hematoma volume, edema volume and treatment methods.

[0053] 4) To predict the 90-day mRS score in patients with hemorrhagic stroke.

[0054] 5) The relationship between the prognosis of hemorrhagic stroke patients and factors such as personal history, medical history, treatment methods and influencing characteristics.

[0055] The method provided in this embodiment comprises the following contents:

[0056] Hematoma expansion risk prediction model: Based on the patient's imaging data and clinical information, a prediction model is established to accurately predict whether the patient will experience hematoma expansion within 48 hours after onset, and to predict the probability of hematoma expansion in all patients. This model helps identify high-risk patients and implement timely and effective treatment interventions.

[0057] Edema volume change analysis method: A method based on data processing and mathematical modeling is proposed to construct a curve of the patient's edema volume over time. By calculating the difference between the true value and the fitting curve, and analyzing the differences in the progression of edema volume over time in different populations, a scientific basis is provided for understanding the edema progression pattern and formulating corresponding treatment strategies.

[0058] Analysis of the effect of treatment methods on progression patterns: Based on treatment methods and edema volume data, the effects of different treatment methods on edema volume progression patterns and how they affect changes in hematoma volume were analyzed. This analysis helps to reveal the mechanism of action of different treatment strategies on disease progression and provide guidance for clinical selection of the best treatment options.

[0059] 90-day mRS score prediction model for hemorrhagic stroke patients: Combined with imaging results, edema, hematoma manifestations, and clinical and treatment data, a regression prediction model was established to accurately predict the patient's 90-day mRS score. This model can provide doctors with reference information for patients' long-term prognosis and help develop more personalized treatment plans.

[0060] Comprehensive evaluation index and correlation analysis of prognostic factors: By analyzing the impact of personal history, medical history, treatment methods and other influencing characteristics on patient prognosis, a comprehensive evaluation index is established. This method helps doctors fully understand the patient's condition and make more scientific and reasonable clinical decisions.

[0061] In this embodiment, the known information is the relevant data of 160 patients with hemorrhagic stroke (sub001 to sub160), which includes personal history, medical history, onset and treatment-related information, as well as multiple repeated imaging examinations (CT plain scan) results and patient prognosis evaluation data.

[0062] According to the serial number of the first imaging examination upon admission, the time interval from onset to the first imaging examination, the serial number at each time point and the corresponding HM_volume, it was determined whether patients sub001 to sub100 had a hematoma expansion event within 48 hours after the onset of the disease. Based on the personal history, medical history, onset-related factors, imaging examination results and other variables of the first 100 patients (sub001 to sub100), a model was constructed to predict the probability of hematoma expansion in all patients (sub001 to sub160).

[0063] Note: A hematoma expansion event was considered to have occurred if the hematoma volume on subsequent examinations increased by ≥6 mL or the relative volume increased by ≥33% compared with the initial examination.

[0064] By observing the relevant data, it can be found that among patients sub001 to sub100, most of the follow-up data are concentrated before the follow-up 3 time point, so it is assumed that the follow-up after that has exceeded 48 hours and is not considered in this embodiment. According to the time interval from the onset of the patient to the first imaging examination plus the time interval obtained by subtracting the time of each follow-up from the time of the first examination, the time is uniformly converted to the hour system. It is observed that the first 100 patients, in their last follow-up, their time is more than 48 hours, and the patients who did not exceed 48 hours did not have a third follow-up. In summary, the assumption is confirmed, that is, this example only considers the data before the follow-up 3 time point.

[0065] In this embodiment, "Follow-up 3 time points" refers to the time points corresponding to the three follow-up visits conducted during the disease treatment or clinical research process. These three time points usually refer to different stages after the start of treatment or research, and are used to evaluate the development of the disease, the treatment effect, or the change of research results. The specific time points may vary depending on the specific disease, treatment plan, or research design.

[0066] According to the above analysis, the occurrence of a hematoma expansion event should meet the definition of hematoma expansion and the time requirements in the question. The definition of hematoma expansion given in the question is that if the increase in the absolute volume of the subsequent examination is greater than or equal to 6ml or the increase in the relative volume is greater than or equal to 33%, then it is determined that a hematoma expansion event has occurred. Based on this, the following logical model is established, which is shown in the following formula:

[0067]

[0068] If the patient data satisfies the above expression, it is set to true, that is, the value is 1, otherwise it is 0. i represents the hematoma volume at the ith follow-up, T i represents the time of the ith follow-up visit, V 0 、T 0It represents the hematoma volume and time when the patient is first admitted to the hospital for examination. It is particularly important to note that T represents the time interval from the onset of the patient to the first imaging examination. This data should also be included within 48 hours of onset.

[0069] Based on the constructed model, it can be obtained whether the hematoma expansion event occurs within 48 hours after the onset of patients sub001 to sub100 as shown in Table 1 below:

[0070] Table 1 Hematoma expansion table

[0071]

[0072]

[0073] Combined with the above known data, a prediction model is constructed to predict the probability of hematoma expansion in all patients. The target variable of the model is whether a hematoma expansion event occurs. The personal history, medical history, imaging examination results, onset and treatment-related characteristics of the first 100 patients are used as feature variables. Four methods, support vector machine (SVM), random forest (RF), K nearest neighbor (KNN) and logistic regression (LR), are used to establish prediction models for the influence of relevant features on hematoma expansion. 5-fold cross validation is used to verify the robustness of the model and predict the data in the test set respectively. Accuracy, ROC and AUC are used as evaluation indicators of the model to visualize the predictions of the four models respectively, as shown in Table 2 and Figure 1 By comparing and analyzing the advantages and disadvantages of the models, it is determined that the random forest algorithm model is the best regression prediction model in this case. Figure 1 The horizontal axis is the false positive rate (False Positive Rate), and the vertical axis is the true positive rate (True Positive Rate).

[0074] Table 2 Evaluation index results

[0075] algorithm AUC value Training set accuracy Test set accuracy RF 0.6793 0.725 0.7 SVM 0.6629 0.775 0.7 KNN 0.6209 0.775 0.7 LR 0.5812 0.775 0.55

[0076] The final probability of hematoma expansion in all patients (sub001 to sub160) is shown in Table 3.

[0077] Table 3 Prediction table of hematoma expansion probability

[0078]

[0079]

[0080] According to the edema volume (ED_volume) and repeated examination time points of the first 100 patients (sub001 to sub100), a time progression curve of edema volume for all patients was constructed (x-axis: time from onset to imaging examination, y-axis: edema volume, y=f(x)), and the residual between the true value of the first 100 patients (sub001 to sub100) and the fitted curve was calculated. The individual differences in the time progression pattern of edema volume in patients were explored, and the time progression curve of edema volume in different populations was constructed, and the residual between the true value of the first 100 patients (sub001 to sub100) and the curve was calculated. The effects of different treatments on the edema volume progression pattern were analyzed.

[0081] The relationship between hematoma volume, edema volume and treatment methods was analyzed.

[0082] The edema volume and corresponding time data of different time periods were extracted, and the extracted data were preprocessed first. After removing some data through box-type outlier detection, the data were fitted using Gaussian fitting and polynomial fitting with the time from onset to imaging examination (Time) as the X-axis and the edema volume (Vo lume) as the Y-axis. Finally, the final fitting curve was judged by four evaluation criteria: the sum of squared errors, mean square error, root mean square error, and goodness of fit. The obtained curve was used to calculate the residual between the true value of the first 100 patients and the fitted curve.

[0083] By comparison, it is found that the polynomial fitting has smaller values ​​in SSE, MSE, and RMSE, and the value of the polynomial fitting (ploynomial) in the fitting standard is closer to 1, so the polynomial fitting is finally selected as the fitting curve in this example. The following is a comparison table of fitting curve evaluation standards (Table 4) and a visualization diagram of the fitting curve ( Figure 2 ).

[0084] Table 4 Comparison of fitting curve evaluation standards

[0085]

[0086] The difference between the polynomial fitting curve and the true value is taken as the final fitting residual. The fitting results for the first 100 patients are shown in Table 5.

[0087] Table 5 Edema volume fitting residuals

[0088]

[0089]

[0090] The model established above already has a certain degree of predictiveness. If the patients are grouped and classified more carefully, it will obviously be more reasonable to explore the progression curve of edema over time. Considering that the change in edema volume is the main factor to be considered, K-mean clustering is used to reduce the dimension of the data. After clustering analysis of the data, it was found that setting the number of categories to 4 can reflect the most reasonable fitting effect. After classification, the data was subjected to box-type outlier detection and abnormal data was eliminated. Similarly, using the time from onset to imaging examination as the X-axis and the edema volume as the Y-axis, Gaussian fitting and polynomial fitting were used to fit the curves of different subclasses, and the same four error evaluation criteria were used to select the optimal fitting model. By comparison, it was found that polynomial fitting performed better in SSE, MSE, and RMS E, and polynomial fitting also performed well in the fitting standard, so polynomial fitting was finally selected as another fitting curve in this embodiment. The following is a comparison table of fitting curve evaluation criteria (Table 6, Table 7) and a visualization diagram of the fitting curve ( Figure 3 , Figure 4 , Figure 5 , Figure 6 ).

[0091] Table 6 Comparison of polynomial fitting curve evaluation standards

[0092]

[0093] Table 7 Gaussian fitting curve evaluation standard comparison table

[0094]

[0095] The polynomial fitting model was used, and the difference between the polynomial fitting curve and the true value was taken as the final fitting residual. (Note: subgroup number = class number - 1) The results of classification and fitting of the first 100 patients are shown in Table 8.

[0096] Table 8 Edema volume classification fitting residuals

[0097]

[0098]

[0099] Some patients in the data have problems such as missing or incomplete follow-up records, so the data of the first 100 patients are used uniformly. For different treatment methods, data are taken out, and the corresponding treatment methods are ventricular drainage, hemostasis treatment, intracranial pressure reduction treatment, antihypertensive treatment, sedation or analgesia treatment, antiemetic stomach protection, and nerve nutrition. They are integrated into a 7-bit 0, 1 sequence character type, similar to 0110111 data (hereinafter referred to as diagnosis and treatment code), and each 1 or 0 corresponds to whether the corresponding treatment method is used. Patients with the same treatment combination are placed in the same group. Through observation, it is found that some treatment combinations are only used by 1 or 2 patients. Because they are not universal, sparse data are not considered. The same treatment combination is regarded as a universal evaluation standard if at least 5 people use it, so 10 treatment combinations are obtained. It is sorted according to the edema volume, and the edema volume of different patients each time is added. A curve of edema volume change is fitted, which is used as the basis for the influence of this treatment combination on the edema volume progression pattern. Finally, the treatments that produced the same or different effects were classified, and the results of the effects of different treatments on the progression pattern of edema volume were obtained.

[0100] All were fitted with a cubic polynomial, and from the table below (Table 9) and the edema volume trend graphs for different treatment combinations attached below, e.g. Figure 7 The vertical axis is the edema volume, and the horizontal axis is the number of follow-up visits. Comprehensive analysis using trend charts shows that antihypertensive treatment will have a volume-reducing effect on edema, while sedation, analgesia, and hemostasis treatment may cause a certain increase in edema volume.

[0101] Table 9 Diagnosis and treatment combination and edema volume trend

[0102] Serial number Medical Coding Edema volume increase or decrease trend 1 0000111 Increase or decrease 2 0001011 reduce 3 0001111 Increase or decrease 4 0011011 reduce 5 0011111 Increase or decrease 6 0101111 Increase or decrease 7 0110011 reduce 8 0111011 Increase or decrease 9 0111111 Increase or decrease 10 1111111 Increase or decrease

[0103] Using the same method and the same evaluation criteria, ten different treatment combinations were divided. Patients with the same treatment were placed in the same group. The hematoma volumes of different patients in the same group were added up at each follow-up. Then, the curves of hematoma volume changes were fitted according to different treatment combinations, which were used as the basis for the effect of this treatment combination on the hematoma volume progression pattern. The changing trend of the effect of different treatment combinations on the hematoma volume progression pattern was analyzed. The same treatment combination was used as the primary key resume table to observe the changing trend of hematoma volume and edema volume under different treatment methods, and to analyze the relationship between hematoma volume, edema volume and treatment methods.

[0104] A comprehensive analysis was conducted from the following table (Table 10) and the hematoma volume trend chart of different treatment combinations, e.g. Figure 8, the vertical axis is the edema volume, and the horizontal axis is the number of follow-up visits. Among all treatment methods, the two treatment methods of antiemetic and gastric protection and nerve nutrition are mild treatment methods, and are not the main influencing factors for the changes in hematoma volume and edema volume. Antihypertensive therapy has a significant effect on the reduction of hematoma and edema volume, while hemostatic therapy may cause a certain increase in hematoma and edema volume. It can be seen from the embodiments that edema around the hematoma is a sign of secondary injury after cerebral hemorrhage. Further observation shows that when the hematoma volume gradually decreases, the edema volume around the hematoma will also decrease accordingly. It is worth noting that when the hematoma volume shows a rapid downward trend, the edema volume will directly decrease. When the hematoma volume decreases slowly or repeatedly, the edema volume will also have a slight increase trend first.

[0105] Table 10 Correspondence of diagnosis and treatment combinations

[0106]

[0107] A prediction model was constructed based on the personal history, disease history, disease-related and first imaging results of the first 100 patients (sub001 to sub100) to predict the 90-day mRS score of patients (sub001 to sub160). The prognosis (90-day mRS) of hemorrhagic stroke patients was analyzed in relation to personal history, disease history, treatment methods and imaging characteristics (including hematoma / edema volume, hematoma / edema location, signal intensity characteristics, shape characteristics), and other associations to provide recommendations for clinical decision-making. The mRS scoring standard table is shown in Table 11.

[0108] Table 11mRS scoring standard table

[0109]

[0110] Using the sorted data, the dependent variable label column was changed to the 90-day mRS score. In the data preprocessing stage, the Pearson correlation coefficient was used to score the characteristic variables related to the 90-day mRS score, and the top 20 features with the highest scores were selected. Random forest was used to reduce the dimension of these data, and the top 10 most important features were selected (original_shape_LeastAxis Length, HM_volume, original_shape_Maximum2DDiameterSlice, original_shape_MinorAxisLength, ED_volume, original_shape_Maximum3Ddiameter, age, NCCT_original_firstorder_Kurtosis, original_shape_SurfaceArea, original_shape_MajorAxisLength). See the table below (Table 12) for details.

[0111] Table 12 Pearson correlation coefficient scoring

[0112]

[0113]

[0114] First, the hyperparameters of the SVM (support vector machine), RF (random forest), KNN (K nearest neighbor) and LR (logistic regression) models are determined. The random search method is also used to search for the best hyperparameters within a certain range for the model. Four regression prediction models are established using the determined hyperparameters. The robustness of the model is verified by 5-fold cross validation and the data in the test set are predicted respectively. The accuracy is used as the evaluation indicator of the model to compare and analyze the advantages and disadvantages of the models.

[0115] Table 13 Comparison of accuracy of four models

[0116] algorithm Training set accuracy RF 0.5 SVM 0.45 KNN 0.3 LR 0.65

[0117] Finally, the LR model was determined as the best regression prediction model for this problem. The 90-day mRs scores of all 160 patients were predicted using this model, and the prediction results are as follows:

[0118] Table 14 90-day mRS score prediction table

[0119]

[0120]

[0121] The features were reprocessed, and the personal medical history, treatment history, size of hematoma edema at the first visit, the maximum difference of hematoma edema after follow-up, image signal intensity characteristics and shape characteristics of the sample patients were extracted for principal component analysis by Pearson correlation coefficient method. After obtaining the data, each feature was scored, and the contribution rate of the feature to the patient's prognosis was ranked, so as to effectively analyze the prognosis of each patient. For various types and data at different times, it was divided into 13 different feature variables, among which the 90-day mRS represents the patient's prognosis. The higher the value, the worse the recovery condition, and the maximum value of 6 represents death. The initial value of hematoma represents the volume of hematoma at the first examination of the patient. The shape of hematoma represents the sum of the shape values ​​of hematoma at the first examination of the patient. The intensity of hematoma represents the sum of the gray intensity values ​​of hematoma at the first examination of the patient. The difference of hematoma represents the difference between the maximum and minimum values ​​of hematoma volume in all follow-up records of the patient. The proportion of the maximum position of hematoma represents the maximum proportion of hematoma in 10 parts of the brain at the first examination of the patient. The same edema also has the above five characteristics. Treatment intensity represents the sum of the types of treatments a patient has received. Number of diseases represents the sum of the types of disease histories a patient has received.

[0122] Table 14 Correlation between characteristics and prognosis

[0123]

[0124]

[0125] According to Table 14, it can be observed that different characteristics have different impacts on patient prognosis, and it can also be observed that the correlations between different characteristics also vary in strength.

[0126] Through correlation analysis, it was found that the initial value of the hematoma had the greatest impact on the patient's prognosis, followed by the shape of the hematoma and the intensity of the hematoma. This means that the worse the hematoma volume is when the patient is first examined, the worse the patient's prognosis is. Edema-related features are also positively correlated with patient prognosis. The two features of the number of diseases and treatment intensity also have an impact on the patient's prognosis, indicating that if the patient has a partial history of the disease, it will also affect the future recovery, and the higher the treatment intensity, the better the recovery effect. It is worth noting that two of the features are negatively correlated with the patient's prognosis, namely the proportion of the maximum position of edema and the proportion of the maximum position of hematoma. This means that when the feature quantity increases, the patient's prognosis score decreases. It means that when the patient's hematoma or edema is concentrated in a certain area, it is easier to recover than when it is dispersed.

[0127] In addition to the relationship with prognosis, there are also obvious correlations between different characteristics. It can be observed that there is a strong positive correlation between the initial value of hematoma, hematoma shape and hematoma intensity, which means that when one variable increases, the other variables also show a significant increase trend, and this relationship is statistically significant. Some characteristics of edema also have this significant relationship.

[0128] Through correlation analysis, it was found that there are many factors that affect the patient's prognosis, some positively correlated and some negatively correlated. This finding is of great significance to clinical decision-making, which helps doctors better optimize strategies. It is recommended that we pay attention to the patient's first examination and distinguish different types of patients in order to better adopt treatment methods, and the more treatment methods are not necessarily better.

[0129] Correlation analysis can only reveal linear relationships between variables, but it does not necessarily indicate causality. If nonlinear relationships exist, these relationships may not be fully captured. Therefore, we need to consider other factors when making decisions. Other possibilities need to be carefully considered to avoid incorrect inferences. In the future, further research is needed on the causal relationship between various characteristics and patient prognosis to determine the best treatment plan for doctors for their patients.

[0130] By combining imaging data, clinical information and treatment methods, and establishing prediction models, mathematical modeling, analysis methods and regression prediction models, we have achieved a comprehensive and accurate prediction and analysis of the disease progression and prognosis of hemorrhagic stroke patients. These methods and models fill the gap in the current technical issues in this field, provide important theoretical basis for doctors, and are expected to play an important role in the clinical diagnosis, treatment and prognosis prediction of hemorrhagic stroke patients.

[0131] Embodiment 2:

[0132] The risk assessment method for hemorrhagic stroke includes the following steps:

[0133] Acquire brain image data and the diagnosis and treatment history data of the corresponding patient, extract the volume, location, shape and distribution information of hematoma and edema corresponding to the examination time point based on the brain image data, associate it with the diagnosis and treatment history data, form a data set and preprocess it;

[0134] The probability of hematoma expansion events in patients is predicted by using the increase in absolute or relative hematoma volume and the corresponding examination time points in the preprocessed data set;

[0135] The edema volume and the corresponding examination time points in the preprocessed data set were used to obtain the curve of edema volume change over time. After post-processing and fitting with the set treatment method code, the influence curve of different treatment methods on edema volume change was obtained.

[0136] The morphological characteristics of hematoma and edema in the preprocessed data set and the corresponding patient's diagnosis and treatment data are used to predict the risk level of patients affected by hematoma and edema in the future time period based on the set scoring rules.

[0137] Embodiment three:

[0138] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the hemorrhagic stroke risk assessment method described in the above-mentioned embodiment 1 are implemented.

[0139] Embodiment 4:

[0140] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the hemorrhagic stroke risk assessment method described in the first embodiment are implemented.

[0141] The steps involved in the above embodiments 2 to 4 correspond to those in embodiment 1. For the specific implementation, please refer to the relevant description of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0142] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for assessing the risk of hemorrhagic stroke, characterized in that: The following steps are involved: Acquire brain image data and the diagnosis and treatment history data of the corresponding patient, extract the volume, location, shape and distribution information of hematoma and edema corresponding to the examination time point based on the brain image data, associate it with the diagnosis and treatment history data, form a data set and preprocess it; The probability of hematoma expansion events in patients is predicted by using the increase in absolute or relative hematoma volume and the corresponding examination time points in the preprocessed data set; The edema volume and the corresponding examination time points in the preprocessed data set were used to obtain the curve of edema volume change over time. After post-processing and fitting with the set treatment method code, the influence curve of different treatment methods on edema volume change was obtained. Among them, the treatment combination and the edema volume were sorted accordingly, the edema volume of different patients was added each time, and the edema volume change curve was fitted by polynomial fitting, which was used as the basis for the influence of this treatment combination on the edema volume progression pattern; the existing treatment intervention means were converted into treatment method codes, and the influence of different treatment methods on the edema volume progression pattern was analyzed; Using the same method and the same evaluation criteria to divide different treatment combinations, patients with the same treatment methods were placed in the same group, the hematoma volumes of different patients in the same group at each follow-up were added up, and the curves of hematoma volume changes were fitted according to different treatment combinations as the basis for the effect of the treatment combination on the hematoma volume progression pattern; the changing trends of the effects of different treatment combinations on the hematoma volume progression pattern were analyzed; Using the morphological characteristics of hematoma and edema in the preprocessed data set and the corresponding patient's diagnosis and treatment data, based on the set scoring rules, the risk level of patients affected by hematoma and edema in the future time period is predicted; By using the Pearson correlation coefficient method to perform principal component analysis, each feature is scored after obtaining the data, and the contribution rate to the patient's prognosis is ranked, so as to effectively analyze the prognosis of each patient.

2. A hemorrhagic stroke risk assessment system, characterized in that: include: The data set construction module is configured to: obtain brain image data and diagnosis and treatment history data of the corresponding patient, extract the volume, position, shape and distribution information of the hematoma and edema corresponding to the examination time point based on the brain image data, associate it with the diagnosis and treatment history data, form a data set and pre-process it; The hematoma expansion risk prediction module is configured to: predict the probability of a hematoma expansion event occurring in a patient by using the increase in the absolute volume or relative volume of the hematoma and the corresponding examination time point in the preprocessed data set; The edema volume change module is configured as follows: using the edema volume and the corresponding examination time point in the pre-processed data set to obtain a curve of edema volume change over time, and after post-processing and fitting with the set treatment method code, obtain the influence curve of different treatment methods on edema volume change, wherein the treatment combination and the edema volume are sorted correspondingly, the edema volume of different patients each time is added, and a polynomial fitting is used to fit the edema volume change curve, which is used as the influence basis of the treatment combination on the edema volume progression pattern; converting the existing treatment intervention means into treatment method code, and analyzing the influence of different treatment methods on the edema volume progression pattern; Using the same method and the same evaluation criteria to divide different treatment combinations, patients with the same treatment methods were placed in the same group, the hematoma volumes of different patients in the same group at each follow-up were added up, and the curves of hematoma volume changes were fitted according to different treatment combinations as the basis for the effect of the treatment combination on the hematoma volume progression pattern; the changing trends of the effects of different treatment combinations on the hematoma volume progression pattern were analyzed; The prognosis risk prediction module is configured to: use the morphological characteristics of hematoma and edema and the diagnosis and treatment data of the corresponding patients in the preprocessed data set, and based on the set scoring rules, predict the risk level of the patient affected by hematoma and edema in the future time period; By using the Pearson correlation coefficient method to perform principal component analysis, each feature is scored after obtaining the data, and the contribution rate to the patient's prognosis is ranked, so as to effectively analyze the prognosis of each patient.

3. The hemorrhagic stroke risk assessment system according to claim 2, characterized in that: The patient's medical history data is obtained by extracting text data from the patient's medical record to obtain the patient's personal information, medical history, and medical history.

4. The hemorrhagic stroke risk assessment system according to claim 2, characterized in that: The trained prediction model is used to predict the probability of a patient experiencing a hematoma expansion event based on the increase in the absolute or relative volume of the hematoma between the set examination time points.

5. The hemorrhagic stroke risk assessment system according to claim 2, characterized in that: The post-processing includes outlier detection, fitting processing, clustering and grouping.

6. The hemorrhagic stroke risk assessment system according to claim 2, characterized in that: The Pearson correlation coefficient method is used to screen the features of the diagnosis and treatment information in the data set whose scores exceed the set values. The trained prediction model and the set scoring rules are used to predict the risk level of patients affected by hematoma and edema in the future time period.

7. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the program is executed by a processor, the steps in the hemorrhagic stroke risk assessment method as claimed in claim 1 are implemented.

8. A computer device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the hemorrhagic stroke risk assessment method as claimed in claim 1 are implemented.

Images