A method, device, equipment and medium for predicting the risk of hemorrhagic transformation

Abstract

The present invention relates to the technical field of intelligent diagnosis of acute ischemic stroke, and specifically to a method, device, equipment and medium for predicting the risk of hemorrhagic transformation. The method includes: obtaining magnetic resonance imaging data of the brain tissue of a target object, and clinical baseline data of the target object; determining imaging features of the brain tissue of the target object according to the magnetic resonance imaging data; determining clinical baseline features of the target object according to the clinical baseline data; and determining a prediction result of the risk of hemorrhagic transformation of the target object based on the imaging features and the clinical baseline features. The method for predicting the risk of hemorrhagic transformation of the present invention can improve the accuracy of HT risk prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent diagnosis of acute ischemic stroke, and particularly to a method, device, equipment and medium for predicting the risk of hemorrhagic transformation. Background Art

[0002] Acute Ischemic Stroke (AIS) is one of the main causes of disability, having a wide impact on individuals, families and society. One of the main complications of AIS patients is Hemorrhagic Transformation (HT), and its occurrence will seriously affect the prognosis of patients.

[0003] The occurrence of HT is affected by various factors, such as the disruption of the blood-brain barrier, neuroinflammation, immune system activation, reperfusion therapy (such as intravenous thrombolysis with tissue plasminogen activator or mechanical thrombectomy, etc.) and other baseline characteristics (such as age, hypertension, atrial fibrillation and diabetes, etc.). Research shows that among patients receiving AIS treatment, there is a high probability of HT occurring, and it is accompanied by a higher mortality rate. Therefore, reliably predicting the HT risk of AIS patients is crucial for guiding treatment decisions and improving patient prognosis.

[0004] Currently, Magnetic Resonance Imaging (MRI) technology has been widely used in the evaluation of AIS patients. In particular, Diffusion Weighted Imaging (DWI) and Perfusion Weighted Imaging (PWI) have been proven to have extremely high sensitivity in diagnosing cerebral infarction, but the research on predicting HT risk based on these images is still insufficient. Currently, the existing HT prediction methods based on images still have certain limitations and the prediction accuracy is relatively low, making it difficult to meet the clinical needs. Summary of the Invention

[0005] Aiming at the above problems of the prior art, the purpose of the present invention is to provide a method, device, equipment and medium for predicting the risk of hemorrhagic transformation, which can improve the accuracy of HT risk prediction.

[0006] To solve the above problems, the present invention provides a method for predicting the risk of hemorrhagic transformation, the method comprising:

[0007] Obtaining magnetic resonance imaging data of the brain tissue of a target object, and clinical baseline data of the target object;

[0008] Determining the imaging features of the brain tissue of the target object according to the magnetic resonance imaging data;

[0009] Determine the clinical baseline characteristics of the target object according to the clinical baseline data;

[0010] Based on the imaging characteristics and clinical baseline characteristics, determine the prediction result of the hemorrhagic transformation risk of the target object.

[0011] Optionally, the magnetic resonance imaging data includes perfusion weighted imaging data, and the clinical baseline data includes NIHSS score;

[0012] Determine the imaging characteristics of the brain tissue of the target object according to the magnetic resonance imaging data, including:

[0013] Determine the maximum time image of the residual function of the brain tissue of the target object according to the perfusion weighted imaging data;

[0014] Determine the imaging characteristics of the brain tissue of the target object according to the maximum time image of the residual function.

[0015] Furthermore, determine the imaging characteristics of the brain tissue of the target object according to the maximum time image of the residual function, including:

[0016] Determine the volume of the hypoperfused area in the brain tissue of the target object and the corresponding hypoperfusion intensity ratio of the brain tissue of the target object according to the maximum time image of the residual function.

[0017] Optionally, the magnetic resonance imaging data includes perfusion weighted imaging data and diffusion weighted imaging data;

[0018] Determine the imaging characteristics of the brain tissue of the target object according to the magnetic resonance imaging data, including:

[0019] Determine the maximum time image of the residual function of the brain tissue of the target object according to the perfusion weighted imaging data;

[0020] Determine the apparent diffusion coefficient image of the brain tissue of the target object according to the diffusion weighted imaging data;

[0021] Determine the imaging characteristics of the brain tissue of the target object based on at least the maximum time image of the residual function and the apparent diffusion coefficient image.

[0022] Furthermore, determine the imaging characteristics of the brain tissue of the target object based on at least the maximum time image of the residual function and the apparent diffusion coefficient image, including:

[0023] Determine the volume of the infarcted area in the brain tissue of the target object according to the apparent diffusion coefficient image;

[0024] Based on at least the maximum time image of the residual function, respectively determine the volume of the brain tissue of the target object where the TMAX value is greater than each of at least one preset threshold, the volume of the perfusion-diffusion mismatch region, and the corresponding hypoperfusion intensity ratio of the brain tissue of the target object.

[0025] Further, the clinical baseline data includes NIHSS score, attribute data, and disease history data;

[0026] Determine the clinical baseline characteristics of the target object according to the clinical baseline data, including:

[0027] Encode and / or standardize the attribute data and the disease history data respectively to obtain the attribute characteristics and the disease history characteristics of the target object.

[0028] Further, based on the imaging characteristics and the clinical baseline characteristics, determine the prediction result of the hemorrhagic transformation risk of the target object, including:

[0029] Obtain the low perfusion intensity ratio corresponding to the brain tissue of the target object;

[0030] Determine the hemorrhagic transformation risk level of the target object according to the low perfusion intensity ratio;

[0031] Use the hemorrhagic transformation risk prediction model corresponding to the hemorrhagic transformation risk level to process the imaging characteristics and the clinical baseline characteristics to obtain the prediction result of the hemorrhagic transformation risk of the target object.

[0032] On the other hand, the present invention provides a device for predicting the risk of hemorrhagic transformation, and the device includes:

[0033] A data acquisition module, configured to acquire magnetic resonance imaging data of the brain tissue of the target object and the clinical baseline data of the target object;

[0034] A first feature determination module, configured to determine the imaging characteristics of the brain tissue of the target object according to the magnetic resonance imaging data;

[0035] A second feature determination module, configured to determine the clinical baseline characteristics of the target object according to the clinical baseline data;

[0036] A risk prediction module, configured to determine the prediction result of the hemorrhagic transformation risk of the target object based on the imaging characteristics and the clinical baseline characteristics.

[0037] On the other hand, the present invention provides an electronic device, including a processor and a memory, and at least one instruction or at least one program segment is stored in the memory, and the at least one instruction or the at least one program segment is loaded and executed by the processor to implement the method for predicting the risk of hemorrhagic transformation as described above.

[0038] On the other hand, the present invention provides a computer-readable storage medium, and at least one instruction or at least one program segment is stored in the computer-readable storage medium, and the at least one instruction or the at least one program segment is loaded and executed by the processor to implement the method for predicting the risk of hemorrhagic transformation as described above.

[0039] On the other hand, the present invention provides a computer program product, including a computer program / instructions, which, when executed by a processor, implements the hemorrhagic transformation risk prediction method as described above.

[0040] Due to the above technical solutions, the present invention has the following beneficial effects:

[0041] According to the hemorrhagic transformation risk prediction method of the embodiments of the present invention, by predicting the HT risk of a target object based on the imaging features of the brain tissue of the target object and the clinical baseline features of the target object, it is possible to quickly and accurately predict the risk of hemorrhagic transformation during the treatment of AIS patients, further improving the accuracy of HT risk prediction, thereby providing reliable decision-making support for doctors, improving the diagnostic efficiency, and improving the treatment effect of AIS patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings based on these drawings without creative efforts.

[0043] Figure 1 is a schematic diagram of an HT risk prediction system provided by an embodiment of the present invention;

[0044] Figure 2 is a flowchart of a hemorrhagic transformation risk prediction method provided by an embodiment of the present invention;

[0045] Figure 3 is a schematic diagram of the structure of a fusion model based on prior knowledge provided by an embodiment of the present invention;

[0046] Figure 4 is a flowchart of a hemorrhagic transformation risk prediction method provided by another embodiment of the present invention;

[0047] Figure 5 is a schematic diagram of the structure of a hemorrhagic transformation risk prediction device provided by an embodiment of the present invention;

[0048] Figure 6 is a block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] To enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0050] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned accompanying drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0051] The method provided by the embodiments of the present invention can be applied to the scenario of predicting the HT risk of AIS patients. Currently, the HT risk of AIS patients is mainly predicted by using a prediction model based on MRI data (such as PWI images and DWI images). Since a large number of imaging markers related to HT have not been fully studied, the prediction model based on existing imaging means has certain limitations.

[0052] Existing prediction models based on a single variable (such as the clinical baseline characteristics or MRI parameters of patients) show low sensitivity and specificity in predicting HT, and the prediction accuracy of the models is difficult to meet the clinical needs. Since the occurrence of HT is the result of the combined action of multiple factors, the prediction model based on a single variable cannot comprehensively consider the complex interaction between different factors.

[0053] In addition, although machine learning techniques have been introduced into the medical field in recent years to explore HT risk prediction, existing models often rely only on a single algorithm and lack the integration and comprehensive utilization of multiple algorithms, resulting in instability and deviation of the models in practical applications.

[0054] An embodiment of the present invention provides a method for predicting the risk of hemorrhagic transformation. By integrating multiple machine learning algorithms and combining the clinical characteristics of patients and various imaging parameters, it can accurately predict the HT risk of AIS patients, overcome the deficiencies in the prior art, improve the accuracy and reliability in clinical applications, thereby providing a more intelligent and efficient decision-making support tool for doctors, helping to timely evaluate the HT risk of patients, optimize treatment strategies, and improve the prognosis of patients.

[0055] In order to make the objectives, technical solutions, and advantages of the embodiments disclosed in the present invention clearer and more understandable, the following further details the embodiments of the present invention in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the embodiments of the present invention and are not used to limit the embodiments of the present invention.

[0056] Refer to the attached Figure 1 figures, which show a schematic diagram of an HT risk prediction system provided by an embodiment of the present invention. As Figure 1 shown, the HT risk prediction system may include at least one medical scanning device 110 and a computer device 120. The computer device 120 and each medical scanning device 110 may be directly or indirectly connected through wired or wireless communication methods, and the embodiments of the present invention do not limit this.

[0057] The medical scanning device 110 may be, but is not limited to, a magnetic resonance imaging device, etc. The computer device 120 may be, but is not limited to, various servers, personal computers, laptop computers, smart phones, tablet computers, and portable wearable devices. The server may be an independent server or a server cluster or distributed system composed of multiple servers, and may also be a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, Content Delivery Network (CDN), and big data and artificial intelligence platforms.

[0058] The computer device 120 may obtain medical imaging data (such as perfusion-weighted imaging data, diffusion-weighted imaging data, etc.) of the brain tissue of the target object scanned by the medical scanning device 110, as well as the clinical baseline data of the target object, and determine the hemorrhagic transformation risk prediction result of the target object through the hemorrhagic transformation risk prediction method provided by the embodiment of the present invention.

[0059] In practical applications, the computer device 120 can provide accurate HT risk prediction results within a short time after MRI scanning, helping doctors make clinical decisions within a critical time window. The computer device 120 may be integrated into the hospital's imaging diagnosis platform as a real-time decision-making support tool to improve the diagnosis efficiency and the treatment effect of patients.

[0060] It should be noted that Figure 1 this is merely an example. Those skilled in the art can understand that although Figure 1 only one medical scanning device 110 is shown in [[ ]], it does not limit the embodiments of the present invention, and there may be more or fewer medical scanning devices 110 than shown in the figure.

[0061] Referring to the attached drawings of the specification Figure 2 which shows the flow of a method for predicting the risk of hemorrhagic transformation provided by an embodiment of the present invention. This method can be applied to Figure 1 the computer device 120 in [[ ]]. Specifically, as Figure 2 shown, the method may include the following steps:

[0062] S210: Obtain magnetic resonance imaging data of the brain tissue of the target object, as well as the clinical baseline data of the target object.

[0063] In the embodiments of the present invention, magnetic resonance imaging (MRI) technology can be used to collect images of the target brain tissue to obtain MRI data. Among them, the target brain tissue may be a patient who may have acute ischemic stroke (AIS).

[0064] In the embodiments of the present invention, the clinical baseline data of the target object can also be collected. The clinical baseline data may include, but is not limited to, one or more of the National Institutes of Health Stroke Scale (NIHSS) score of the target object at the time of admission, the attribute data of the target object, and the disease history data (such as hypertension, diabetes, atrial fibrillation, etc.).

[0065] It should be noted that the sources of the MRI data and the clinical baseline data can be either directly imported relevant data, or obtained by real-time configuration connection from other resource libraries, or obtained by searching according to information such as the patient's name from a stored image database. The embodiments of the present invention do not limit this.

[0066] In some embodiments, the MRI data may include, but is not limited to, perfusion weighted imaging (PWI) data, and the clinical baseline data may include, but is not limited to, the NIHSS score.

[0067] In some embodiments, the MRI data may include, but is not limited to, PWI data and diffusion weighted imaging (DWI) data, and the clinical baseline data may include, but is not limited to, the NIHSS score, the attribute data, and the disease history data.

[0068] S220: Determine the imaging features of the brain tissue of the target object according to the magnetic resonance imaging data.

[0069] In the embodiments of the present invention, the acquired MRI data can be processed to obtain the imaging features of the brain tissue of the target object.

[0070] Specifically, when different MRI data are acquired, different imaging features can be processed.

[0071] In some embodiments, when the MRI data only includes PWI data, determining the imaging features of the brain tissue of the target object according to the magnetic resonance imaging data may include: determining the Time to Maximum Enhancement (TMAX) image of the brain tissue of the target object according to the perfusion weighted imaging data; determining the imaging features of the brain tissue of the target object according to the TMAX image. Specifically, the imaging features of the brain tissue of the target object may include, but are not limited to, the volume of the hypoperfused area and the Hypoperfusion Intensity Ratio (HIR).

[0072] In the embodiments of the present invention, existing processing methods can be used to process the PWI data to obtain the TMAX image. For example, the RAPID software can be used to process the PWI data to obtain the TMAX image, which will not be elaborated herein. Among them, the TMAX image may include the time when the blood storage function of each voxel in the target brain tissue reaches the maximum value.

[0073] In the embodiments of the present invention, determining the imaging features of the brain tissue of the target object according to the TMAX image may include: determining the volume of the hypoperfused area in the brain tissue of the target object according to the TMAX image, and the hypoperfusion intensity ratio corresponding to the brain tissue of the target object.

[0074] Specifically, the brain tissue area with a TMAX value > 6s can be determined as the hypoperfused area according to the TMAX image, and the volume of this area (i.e., the volume of the hypoperfused area) can be determined. It is also possible to determine the brain tissue area with a TMAX value > 10s according to the TMAX image, and calculate the ratio of the volume of the brain tissue area with a TMAX value > 10s to the volume of the hypoperfused area as the HIR.

[0075] In some embodiments, when the MRI data includes PWI data and DWI data, determining the imaging features of the brain tissue of the target object according to the magnetic resonance imaging data may include: determining the Time to Maximum Enhancement (TMAX) image of the brain tissue of the target object according to the perfusion weighted imaging data; determining the Apparent Diffusion Coefficient (ADC) image of the brain tissue of the target object according to the diffusion weighted imaging data; determining the imaging features of the brain tissue of the target object based on at least the TMAX image and the ADC image.

[0076] Specifically, the imaging features of the brain tissue of the target object may include, but are not limited to, the volume of the infarct area, the volume of the area where the TMAX value is greater than each of at least one preset threshold, the volume of the perfusion-diffusion mismatch area, and various of the low perfusion intensity ratios. Among them, at least one preset threshold can be set in advance according to the actual situation. For example, it can be set to 4s, 6s, 8s, 10s, etc., and the embodiments of the present invention are not limited thereto.

[0077] In the embodiments of the present invention, the method for determining the TMAX image is the same as the case where the MRI data only includes PWI data, and the embodiments of the present invention will not be elaborated herein.

[0078] In the embodiments of the present invention, existing processing methods can be used to process the DWI data to obtain the ADC image. For example, the RAPID software can be used to process the DWI data to obtain the ADC image, and the embodiments of the present invention will not be elaborated herein. Among them, the ADC image may include the diffusion velocity of water molecules in each voxel of the target brain tissue.

[0079] In the embodiments of the present invention, at least based on the residual functional maximum time image and the apparent diffusion coefficient image, the imaging features of the brain tissue of the target object are determined, which may include: determining the volume of the infarct area in the brain tissue of the target object according to the apparent diffusion coefficient image; at least based on the residual functional maximum time image, respectively determining the volume of the area where the TMAX value in the brain tissue of the target object is greater than each of at least one preset threshold, the volume of the perfusion-diffusion mismatch area, and the low perfusion intensity ratio corresponding to the brain tissue of the target object.

[0080] Specifically, according to the ADC image, the brain tissue area where ADC < 620×10 -6 mm 2 / s can be determined as the infarct area, and the volume of this area (i.e., the volume of the infarct area) can be determined. According to the TMAX image, the brain tissue areas where the TMAX value is greater than each of at least one preset threshold can be determined, and the volumes of these areas can be determined. For example, when at least one preset threshold is set to 4s, 6s, 8s, and 10s, the brain tissue areas where the TMAX value > 4s, the TMAX value > 6s, the TMAX value > 8s, and the TMAX value > 10s can be determined respectively, and the volumes of these areas can be determined.

[0081] Specifically, the difference between the volume of the brain tissue area where the TMAX value > 6s (i.e., the volume of the hypoperfused area) and the volume of the infarct area can also be calculated as the volume of the perfusion-diffusion mismatch area. The ratio of the volume of the brain tissue area where the TMAX value > 10s to the volume of the brain tissue area where the TMAX value > 6s can be calculated as the HIR.

[0082] S230: Determine the clinical baseline characteristics of the target object based on the clinical baseline data.

[0083] In the embodiments of the present invention, the obtained clinical baseline data can be processed to obtain the clinical baseline characteristics of the target object.

[0084] Specifically, when different clinical baseline data are obtained, different clinical baseline characteristics can be processed. Among them, the clinical baseline characteristics can include but are not limited to NIHSS scores, attribute characteristics, disease history characteristics, etc.

[0085] In some embodiments, when the obtained clinical baseline data only include NIHSS scores, the NIHSS scores can be directly used as the clinical baseline characteristics of the target object.

[0086] In some embodiments, when the obtained clinical baseline data include not only NIHSS scores but also attribute data and disease history data, determining the clinical baseline characteristics of the target object based on the clinical baseline data can include: respectively encoding and / or normalizing the attribute data and disease history data to obtain the attribute characteristics and disease history characteristics of the target object.

[0087] Specifically, the attribute data of the target object can include the gender of the target object, and the disease history data can include whether the target object has diseases such as hypertension, diabetes, atrial fibrillation, etc. For the gender of the target object, binary processing can be performed, encoding females as 0 and males as 1. For the disease history of the target object, binary processing can also be performed, encoding no disease history as 0 and having a disease history as 1. The NIHSS score can be directly used as the clinical baseline characteristics of the target object.

[0088] In some embodiments, the attribute data and disease history data can also be normalized to eliminate the differences in data dimensions and magnitudes, so that different characteristics have the same scale. The specific normalization method can refer to the prior art, and the embodiments of the present invention will not elaborate here.

[0089] It can be understood that through encoding and other processing of the obtained clinical baseline data in the embodiments of the present invention, discrete attributes can be converted into numerical representations available for the model, facilitating subsequent use of machine learning algorithms for HT risk prediction.

[0090] S240: Based on the imaging characteristics and clinical baseline characteristics, determine the prediction result of the hemorrhagic transformation risk of the target object.

[0091] In the embodiments of the present invention, imaging features and clinical baseline features can be combined, and a machine learning algorithm can be used to predict the HI risk of the target brain tissue. Specifically, the obtained imaging features and clinical baseline features can be directly input into a pre-trained hemorrhagic transformation risk prediction model, which can process the input imaging features and clinical baseline features to obtain and output the hemorrhagic transformation risk prediction result of the target object.

[0092] It can be understood that according to the hemorrhagic transformation risk prediction method of the embodiments of the present invention, by predicting the HT risk of the target object based on the imaging features of the target object's brain tissue and the clinical baseline features of the target object, the risk of hemorrhagic transformation can be quickly and accurately predicted during the treatment of AIS patients, further improving the accuracy of HT risk prediction, so as to provide reliable decision-making support for doctors, improve the diagnosis efficiency, and improve the treatment effect of AIS patients.

[0093] In some embodiments, when different MRI data and clinical baseline data are obtained, different hemorrhagic transformation risk prediction models can be used. For example, when the obtained MRI data only includes PWI data, and / or the obtained clinical baseline data only includes the NIHSS score, the hemorrhagic transformation risk prediction model can be a Logistic regression model. When the obtained MRI data includes PWI data and DWI data, and the obtained clinical baseline data includes the NIHSS score, attribute data, and disease history data, the hemorrhagic transformation risk prediction model can be a fusion model based on prior knowledge.

[0094] In some embodiments, when the obtained MRI data only includes PWI data, and / or the obtained clinical baseline data only includes the NIHSS score, the obtained imaging features (including the volume of the hypoperfused area and HIR) and clinical baseline features (including the NIHSS score) can be input into a pre-trained Logistic regression model to obtain the hemorrhagic transformation risk prediction result output by the model.

[0095] In practical applications, due to the advantages of simplicity and efficiency of the Logistic regression model in clinical operations, and the simple model structure which is suitable for rapid deployment and application in the clinical environment, the Logistic regression model can be selected for HT risk prediction. Specifically, the Logistic regression model only uses three features, namely the volume of the hypoperfused area (the volume of the brain tissue area with TMAX value > 6s), HIR, and NIHSS score. That is, the input data of the Logistic regression model can include the volume of the hypoperfused area, HIR, and NIHSS score. The output data of the model can be the HT risk score, and the score range is from 0 to 1. When the score output by the Logistic regression model is greater than or equal to the preset score, it indicates a higher HT risk, and when it is less than the preset score, it indicates a lower HT risk. Among them, the preset score can be set in advance according to the actual situation. For example, it can be set to 0.5, and the embodiments of the present invention do not limit this.

[0096] Specifically, the Logistic regression model can be obtained by using the PWI sample data determined from the MRI data of multiple patients with AIS and the corresponding NIHSS score samples as the training sample data, training through the maximum likelihood estimation method, and optimizing the model parameters using the 5-fold cross-validation technique. Or it can comprehensively use the PWI sample data determined from the MRI data of multiple patients with AIS and the corresponding NIHSS score samples, and the PWI sample data determined from the MRI data of multiple normal brain tissues (i.e., the brain tissues of healthy people) and the corresponding NIHSS score samples as the training sample data, training through the maximum likelihood estimation method, and optimizing the model parameters using the 5-fold cross-validation technique. The specific model training method can refer to the prior art, and the embodiments of the present invention will not elaborate here.

[0097] It can be understood that by adopting the Logistic regression model in the embodiments of the present invention, only a small number of key features (such as the volume of the hypoperfused area, HIR, and NIHSS score) are needed to achieve high prediction accuracy, avoiding the dependence on excessive data, reducing the complexity of clinical data collection and processing, so that while ensuring high prediction accuracy, the operation process is simplified, the real-time performance and usability of the overall system are improved, and it is convenient for rapid clinical deployment and application. Doctors can quickly evaluate the HT risk situation of patients in the clinical scenario, thereby optimizing the treatment decision-making process, reducing waiting time, and improving medical efficiency.

[0098] In some embodiments, the fusion model based on prior knowledge can be composed of two sub-models and a truncated feature. The truncated feature can be an imaging feature and has a corresponding truncation threshold. The fusion model based on prior knowledge can divide the object to be measured into different hemorrhagic transformation risk levels based on the truncated feature, and use different models to predict the hemorrhagic transformation risk respectively.

[0099] It can be understood that in the embodiments of the present invention, by developing a fusion model based on prior knowledge, classifying based on the imaging features of patients, and then applying different sub-models for HT risk prediction respectively, the advantages of multiple machine learning algorithms can be combined. Compared with a single model, the accuracy, sensitivity, and specificity of HT risk prediction can be further improved, the prediction ability of the overall system in complex situations can be improved, the bias and misjudgment in single-model prediction can be significantly reduced, the robustness of the model can be enhanced, and it is suitable for deployment and application in complex clinical scenarios.

[0100] Specifically, with reference to the attached Figure 3 , the truncation feature of the fusion model based on prior knowledge can be the HIR value, and the two sub-models can be the Light Gradient Boosting Machine (LightGBM) model and the Extreme Gradient Boosting (XGBoost) model. That is to say, the fusion model based on prior knowledge can divide the object to be measured into different hemorrhagic transformation risk levels based on the HIR value, and use the LightGBM model and the XGBoost model respectively for hemorrhagic transformation risk prediction.

[0101] Specifically, with reference to the attached Figure 4 , when the obtained MRI data includes PWI data and DWI data, and the obtained clinical baseline data includes NIHSS score, attribute data, and disease history data, based on the imaging features and clinical baseline features, determining the hemorrhagic transformation risk prediction result of the target object may include:

[0102] S241: Obtain the hypoperfusion intensity ratio corresponding to the brain tissue of the target object.

[0103] Specifically, the HIR corresponding to the brain tissue of the target object calculated in the above step S220 can be directly obtained.

[0104] S242: Determine the hemorrhagic transformation risk level of the target object according to the hypoperfusion intensity ratio.

[0105] Specifically, after obtaining the HIR corresponding to the brain tissue of the target object, it can be judged whether the HIR is greater than the truncation threshold. When the HIR is greater than or equal to the truncation threshold, it is determined that the hemorrhagic transformation risk level of the target object is high risk. When the HIR is less than the truncation threshold, it is determined that the hemorrhagic transformation risk level of the target object is low risk. Among them, the truncation threshold can be preset according to the actual situation of the truncation feature. For example, when the truncation feature is the HIR value, it can be set to 0.5. The embodiments of the present invention do not limit this.

[0106] S243: Using a hemorrhagic transformation risk prediction model corresponding to the hemorrhagic transformation risk level, process the imaging features and clinical baseline features to obtain the hemorrhagic transformation risk prediction result of the target object.

[0107] Specifically, when the hemorrhagic transformation risk level of the target object is high risk, the XGBoost model can be used for hemorrhagic transformation risk prediction. When the hemorrhagic transformation risk level of the target object is low risk, the LightGBM model can be used for hemorrhagic transformation risk prediction.

[0108] Specifically, the obtained imaging features (including the infarct volume, the volume of the TMAX value greater than each preset threshold in at least one preset threshold, the perfusion-diffusion mismatch region volume, and the HIR) and clinical baseline features (including the NIHSS score, the attribute features, and the disease history features) can be input into the pre-trained XGBoost model or LightGBM model to obtain the hemorrhagic transformation risk prediction result output by the model.

[0109] Specifically, the XGBoost model and the LightGBM model can be trained using the PWI sample data and DWI sample data determined from the MRI data of multiple AIS patients, as well as the corresponding NIHSS score samples, attribute sample data, and disease history sample data as training sample data, and trained using the 5-fold cross-validation technique, and the hyperparameters are adjusted using random grid search to ensure the optimal performance of the model. Or the PWI sample data and DWI sample data determined from the MRI data of multiple AIS patients, as well as the corresponding NIHSS score samples, attribute sample data, and disease history sample data, and the PWI sample data and DWI sample data determined from the MRI data of multiple normal brain tissues (i.e., the brain tissues of healthy people), as well as the corresponding NIHSS score samples, attribute sample data, and disease history sample data can be integrated as training sample data, and trained using the 5-fold cross-validation technique, and the hyperparameters are adjusted using random grid search to ensure the optimal performance of the model. The specific model training method can refer to the prior art, and the embodiments of the present invention will not be elaborated here.

[0110] According to experiments, the fusion model based on prior knowledge significantly improves the accuracy of HT risk prediction for AIS patients by combining two efficient machine learning algorithms, LightGBM and XGBoost, and fusing HIR as the classification criterion. The area under the curve (AUC) value of this model on the test set reached 0.9161, which is better than the prediction effect of a single model. This model helps doctors better identify high-risk patients, so as to take timely treatment measures.

[0111] It can be understood that the fusion model based on prior knowledge classifies patients a priori based on the HIR index, and flexibly selects an appropriate sub-model (LightGBM model or XGBoost model) for HT risk prediction. This design not only improves the prediction flexibility of the model, enables the model to better handle different types of AIS patients, thereby obtaining more personalized prediction results, further improving the accuracy of HT risk prediction, but also greatly enhances the adaptability of the model in actual clinical applications.

[0112] In addition, through the multi-model fusion technology, the prediction errors and biases that may be generated by a single model are reduced, further improving the stability and robustness of the model, and ensuring that a high prediction performance can still be maintained in complex clinical scenarios.

[0113] It should be noted that when the obtained MRI data includes PWI data and DWI data, and the obtained clinical baseline data includes NIHSS score, attribute data, and disease history data, a Logistic regression model can also be used for HT risk prediction according to the user's choice. The embodiments of the present invention are not limited thereto.

[0114] In practical applications, in cases where high prediction real-time performance is required, the Logistic regression model can be selected for HT risk prediction. In cases where high prediction accuracy is required, the fusion model based on prior knowledge can be selected for HT risk prediction. By combining these two models, the embodiments of the present invention can not only provide a fast and simple clinical prediction tool (based on the Logistic regression model), but also provide a higher-accuracy prediction result (based on the fusion model) in complex scenarios, effectively improving the efficiency and accuracy of HT risk prediction for AIS patients.

[0115] In some embodiments, the obtained hemorrhagic transformation risk prediction results can also be displayed through a visualization interface. For example, the HT risk score of the target object and the importance of key features can be displayed. The key features may include, but are not limited to, features such as TMAX volume, NIHSS score, HIR, etc., so that doctors can intuitively understand the HT risk situation of each patient, thereby assisting clinical decision-making.

[0116] In some embodiments, the hemorrhagic transformation risk prediction results of the target object can also be updated in real time. Once the MRI data and clinical baseline data of the target object's brain tissue are updated, the above risk prediction process can be automatically executed to generate and display the latest prediction results, thereby helping doctors quickly judge the HT risk of the patient.

[0117] It can be understood that the embodiments of the present invention combine efficient machine learning algorithms with clinical operability design. By integrating multiple machine learning algorithms (such as Logistic regression, XGBoost, and LightGBM), the sensitivity and specificity of the hemorrhagic transformation risk prediction model are greatly improved, and the problem of unstable performance of existing single models is solved. This not only helps to improve the treatment prognosis of AIS patients but also can be more widely promoted in clinical applications.

[0118] Refer to the appended Figure 5 which shows the structure of a hemorrhagic transformation risk prediction device 500 provided by an embodiment of the present invention. As Figure 5 shown, the device 500 may include:

[0119] A data acquisition module 510, configured to acquire magnetic resonance imaging data of the brain tissue of the target object and the clinical baseline data of the target object;

[0120] A first feature determination module 520, configured to determine the imaging features of the brain tissue of the target object according to the magnetic resonance imaging data;

[0121] A second feature determination module 530, configured to determine the clinical baseline features of the target object according to the clinical baseline data;

[0122] A risk prediction module 540, configured to determine the hemorrhagic transformation risk prediction result of the target object based on the imaging features and the clinical baseline features.

[0123] In some embodiments, the risk prediction module 540 may include:

[0124] An HIR acquisition unit, configured to acquire the low perfusion intensity ratio corresponding to the brain tissue of the target object;

[0125] A risk level determination unit, configured to determine the hemorrhagic transformation risk level of the target object according to the low perfusion intensity ratio;

[0126] A risk prediction unit, configured to process the imaging features and the clinical baseline features by using the hemorrhagic transformation risk prediction model corresponding to the hemorrhagic transformation risk level to obtain the hemorrhagic transformation risk prediction result of the target object.

[0127] It should be noted that for the device provided in the above embodiments, when implementing its functions, only the above-mentioned division of each functional module is used as an example. In actual applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device provided in the above embodiments and the corresponding method embodiments belong to the same concept, and the specific implementation process can be seen in the corresponding method embodiments, which will not be elaborated here.

[0128] An embodiment of the present invention further provides an electronic device, which includes a processor and a memory. At least one instruction or at least one program segment is stored in the memory, and the at least one instruction or the at least one program segment is loaded and executed by the processor to implement the hemorrhagic transformation risk prediction method provided in the above method embodiment.

[0129] The memory can be used to store software programs and modules. The processor runs the software programs and modules stored in the memory to execute various functional applications and data processing. The memory mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for functions, etc.; the data storage area can store data created according to the use of the device, etc. In addition, the memory can include high-speed random access memory, and can also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage devices. Correspondingly, the memory can also include a memory controller to provide the processor with access to the memory.

[0130] With reference to the accompanying specification Figure 6 FIG. shows a block diagram of an electronic device 600 according to an embodiment of the present invention. The electronic device 600 may include one or more processors 602, a system control logic 608 connected to at least one of the processors 602, a system memory 604 connected to the system control logic 608, a non-volatile memory (NVM) 606 connected to the system control logic 608, and a network interface 610 connected to the system control logic 608.

[0131] The processor 602 may include one or more single-core or multi-core processors. The processor 602 may include any combination of general-purpose processors and dedicated processors (such as graphics processors, application processors, baseband processors, etc.). In the embodiments herein, the processor 602 may be configured to execute one or more embodiments according to various embodiments as Figures 2 to 4 shown.

[0132] In some embodiments, the system control logic 608 may include any suitable interface controller to provide any suitable interface to at least one of the processors 602 and / or any suitable device or component communicating with the system control logic 608.

[0133] In some embodiments, the system control logic 608 may include one or more memory controllers to provide an interface connected to the system memory 604. The system memory 604 may be used to load and store data and / or instructions. In some embodiments, the memory 604 of the device 600 may include any suitable volatile memory, such as a suitable dynamic random access memory (DRAM).

[0134] The NVM / memory 606 may include one or more tangible, non-transitory computer-readable media for storing data and / or instructions. In some embodiments, the NVM / memory 606 may include any suitable non-volatile memory such as flash memory and / or any suitable non-volatile storage device, such as at least one of a HDD (Hard Disk Drive), a CD (Compact Disc) drive, and a DVD (Digital Versatile Disc) drive.

[0135] The NVM / memory 606 may include a portion of the storage resources installed on the device 600, or it may be accessible by the device but not necessarily part of the device. For example, the NVM / storage 606 may be accessed via the network interface 610 over a network.

[0136] Specifically, the system memory 604 and the NVM / memory 606 may respectively include a temporary copy and a permanent copy of the instructions 620. The instructions 620 may include instructions that, when executed by at least one of the processors 602, cause the device 600 to implement Figures 2 to 4 the hemorrhagic transformation risk prediction method as shown. In some embodiments, the instructions 620, hardware, firmware, and / or its software components may alternatively be disposed in the system control logic 608, the network interface 610, and / or the processor 602.

[0137] The network interface 610 may include a transceiver for providing a radio interface for the device 600 to communicate with any other suitable device (such as a front-end module, an antenna, etc.) over one or more networks. In some embodiments, the network interface 610 may be integrated with other components of the device 600. For example, the network interface 610 may be integrated with at least one of the communication module of the processor 602, the system memory 604, the NVM / memory 606, and a firmware device (not shown) having instructions. When at least one of the processors 602 executes the instructions, the device 600 implements Figures 2 to 4 one or more embodiments of the various embodiments as shown.

[0138] The network interface 610 may further include any suitable hardware and / or firmware to provide a multiple-input multiple-output radio interface. For example, the network interface 610 may be a network adapter, a wireless network adapter, a telephone modem, and / or a wireless modem.

[0139] In one embodiment, at least one of the processors 602 may be logically encapsulated with one or more controllers for the system control logic 608 to form a System in Package (SiP). In one embodiment, at least one of the processors 602 may be integrated with the logic of one or more controllers for the system control logic 608 on the same die to form a System on Chip (SoC).

[0140] Device 600 may further include: an Input / Output (I / O) device 612. The I / O device 612 may include a user interface that enables a user to interact with the device 600; the design of the peripheral component interface enables peripheral components to also interact with the device 600. In some embodiments, the device 600 further includes sensors for determining at least one of environmental conditions and location information related to the device 600.

[0141] In some embodiments, the user interface may include, but is not limited to, a display (e.g., a liquid crystal display, a touch screen display, etc.), a speaker, a microphone, one or more cameras (e.g., a still image camera and / or a video camera), a flashlight (e.g., a light emitting diode flash) and a keyboard.

[0142] In some embodiments, the peripheral component interface may include, but is not limited to, a non-volatile memory port, an audio jack and a power interface.

[0143] In some embodiments, the sensors may include, but are not limited to, a gyroscope sensor, an accelerometer, a proximity sensor, an ambient light sensor and a positioning unit. The positioning unit may also be part of or interact with the network interface 610 to communicate with components of a positioning network (e.g., Global Positioning System (GPS) satellites).

[0144] It can be understood that the structure illustrated in the embodiments of the present invention does not constitute a specific limitation on the electronic device 600. In other embodiments of the present invention, the electronic device 600 may include more or fewer components than shown, or combine certain components, or split certain components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0145] An embodiment of the present invention also provides a computer-readable storage medium, which may be disposed in the electronic device to store at least one instruction or at least one segment of program related to implementing a method for predicting the risk of hemorrhagic transformation. The at least one instruction or the at least one segment of program is loaded and executed by the processor to implement the method for predicting the risk of hemorrhagic transformation provided in the above method embodiments.

[0146] Optionally, in the embodiments of the present invention, the above storage medium may include but is not limited to: various media such as USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical discs that can store program codes.

[0147] An embodiment of the present invention further provides a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the hemorrhagic transformation risk prediction method provided in the above various optional implementation examples.

[0148] It should be noted that: the above sequence of the embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. And the above specific embodiments of this specification have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be executed in a different order from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0149] Each embodiment in this specification is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0150] Those of ordinary skill in the art can understand that all or part of the steps for implementing the above embodiments can be completed by hardware, or can be completed by a program instructing the relevant hardware. The program can be stored in a computer-readable storage medium, and the above-mentioned storage medium can be a read-only memory, a magnetic disk, or an optical disc, etc.

[0151] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for predicting the risk of hemorrhagic transformation, characterized in that Comprising: Obtaining magnetic resonance imaging data of the brain tissue of a target object, as well as the clinical baseline data of the target object; Determining the imaging features of the brain tissue of the target object according to the magnetic resonance imaging data; Determining the clinical baseline features of the target object according to the clinical baseline data; Inputting the imaging features and the clinical baseline features into a pre-trained hemorrhagic transformation risk prediction model to determine the hemorrhagic transformation risk prediction result of the target object. The hemorrhagic transformation risk prediction model is a fusion model based on prior knowledge. The fusion model based on prior knowledge consists of two sub-models and a truncated feature. The truncated feature is the HIR value. The two sub-models are a light gradient boosting machine model and an extreme gradient boosting model respectively; Wherein, the magnetic resonance imaging data includes perfusion weighted imaging data and diffusion weighted imaging data, The determining the imaging features of the brain tissue of the target object according to the magnetic resonance imaging data includes: Determining the maximum residual function time image of the brain tissue of the target object according to the perfusion weighted imaging data; Determining the apparent diffusion coefficient image of the brain tissue of the target object according to the diffusion weighted imaging data; Determining the imaging features of the brain tissue of the target object based on at least the maximum residual function time image and the apparent diffusion coefficient image.

2. The method according to claim 1, wherein The magnetic resonance imaging data includes perfusion weighted imaging data, and the clinical baseline data includes the NIHSS score; The determining the imaging features of the brain tissue of the target object according to the magnetic resonance imaging data includes: Determining the maximum residual function time image of the brain tissue of the target object according to the perfusion weighted imaging data; Determining the imaging features of the brain tissue of the target object according to the maximum residual function time image.

3. The method according to claim 2, wherein The determining the imaging features of the brain tissue of the target object according to the maximum residual function time image includes: Determining the volume of the hypoperfused area in the brain tissue of the target object and the hypoperfusion intensity ratio corresponding to the brain tissue of the target object according to the maximum residual function time image.

4. The method according to claim 1, wherein The determining the imaging features of the brain tissue of the target object based on at least the maximum residual function time image and the apparent diffusion coefficient image includes: Determining the volume of the infarcted area in the brain tissue of the target object according to the apparent diffusion coefficient image; Based on at least the maximum residual function time image, respectively determining the volume of the brain tissue of the target object where the TMAX value is greater than each preset threshold among at least one preset threshold, the perfusion-diffusion mismatch area volume, and the hypoperfusion intensity ratio corresponding to the brain tissue of the target object.

5. The method according to claim 1, characterized in that, The clinical baseline data includes the NIHSS score, attribute data, and disease history data; The determining the clinical baseline features of the target object according to the clinical baseline data includes: Encoding and / or standardizing the attribute data and the disease history data respectively to obtain the attribute features and disease history features of the target object.

6. The method according to claim 1, characterized in that Determining the hemorrhagic transformation risk prediction result of the target object based on the imaging features and the clinical baseline features includes: Obtaining the low perfusion intensity ratio corresponding to the brain tissue of the target object; Determining the hemorrhagic transformation risk level of the target object according to the low perfusion intensity ratio; Using a hemorrhagic transformation risk prediction model corresponding to the hemorrhagic transformation risk level to process the imaging features and the clinical baseline features to obtain the hemorrhagic transformation risk prediction result of the target object.

7. A device for predicting the risk of hemorrhagic transformation, characterized in that, Including: A data acquisition module for acquiring magnetic resonance imaging data of the brain tissue of the target object and the clinical baseline data of the target object, where the magnetic resonance imaging data includes perfusion weighted imaging data and diffusion weighted imaging data; A first feature determination module for determining the imaging features of the brain tissue of the target object according to the magnetic resonance imaging data. Determining the imaging features of the brain tissue of the target object according to the magnetic resonance imaging data includes: determining the maximum residual function time image of the brain tissue of the target object according to the perfusion weighted imaging data; determining the apparent diffusion coefficient image of the brain tissue of the target object according to the diffusion weighted imaging data; determining the imaging features of the brain tissue of the target object based on at least the maximum residual function time image and the apparent diffusion coefficient image; A second feature determination module for determining the clinical baseline features of the target object according to the clinical baseline data; A risk prediction module for inputting the imaging features and the clinical baseline features into a pre-trained hemorrhagic transformation risk prediction model to determine the hemorrhagic transformation risk prediction result of the target object. The hemorrhagic transformation risk prediction model is a fusion model based on prior knowledge. The fusion model based on prior knowledge consists of two sub-models and a truncated feature. The truncated feature is the HIR value. The two sub-models are a light gradient boosting machine model and an extreme gradient boosting model respectively.

8. An electronic device, characterized in that, The electronic device includes a processor and a memory. At least one instruction or at least one program segment is stored in the memory. The at least one instruction or the at least one program segment is loaded and executed by the processor to implement the hemorrhagic transformation risk prediction method according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, At least one instruction or at least one program segment is stored in the computer-readable storage medium. The at least one instruction or at least one program segment is loaded and executed by the processor to implement the hemorrhagic transformation risk prediction method according to any one of claims 1-6.

Images