A post-stroke depression prediction model, a construction method and application thereof

Abstract

The application discloses a prediction model for post-stroke depression and a construction method and application thereof. It is found that gender, combined diseases, NIHSS and MLR are independent influencing factors for the occurrence of post-stroke depression (PSD), and a nomogram model for predicting PSD is constructed by taking gender, combined diseases, NIHSS and MLR as key indexes. It is proved by a training set and an independent verification set that the prediction model provided by the application has good prediction performance, discrimination and calibration degree, and can stably distinguish between high-risk and low-risk patients with PSD. The model provides an efficient and practical quantitative tool for early identification of high-risk patients with PSD, and has wide application prospect and conversion value.

Description

Technical Field

[0001] This invention belongs to the field of bioinformatics, specifically relating to a predictive model for post-stroke depression, its construction method, and its application. Background Technology

[0002] Poststroke depression (PSD) is an affective disorder syndrome that occurs after a stroke. Its clinical manifestations meet the diagnostic criteria for depression, and it has a clear causal relationship with stroke. PSD has a high incidence rate, which not only seriously reduces patients' quality of life, but also increases the recurrence rate and mortality rate of stroke by hindering the recovery of neurological function, thus affecting the overall prognosis of patients and imposing a heavy medical and socioeconomic burden on patients, their families, and society.

[0003] The pathogenesis of PSD exhibits high heterogeneity, encompassing a multi-layered theoretical framework, including hypothalamic-pituitary-adrenal (HPA) axis dysfunction hypothesis, immune cell hypothesis, neurotrophic factor hypothesis, neurotransmitter hypothesis, and the interaction of oxidative stress and free radical accumulation. These hypotheses do not act independently but are intricately interconnected, collectively contributing to the development of PSD. Therefore, to a certain extent, PSD may be a brain pathological change caused by stroke itself.

[0004] Currently, the diagnosis of PSD mainly relies on subjective assessments, such as the Hamilton Depression Rating Scale (HDRS) and the Patient Health Questionnaire-9 (PHQ-9). However, their objectivity and accuracy have certain limitations in clinical application. As mentioned earlier, the pathogenesis of PSD involves the dynamic interaction of multiple systems, including neuroinflammatory activation, HPA axis dysregulation, and oxidative stress. This case complexity may prevent a single biomarker from accurately predicting the risk of stroke one month after the stroke. Therefore, finding reliable combinations of biomarkers to construct predictive models has become an urgent technical problem to be solved in this field. Summary of the Invention

[0005] To overcome the shortcomings of the prior art, the present invention provides a method for predicting the risk of post-stroke depression, and also provides a model, system, device, computer-readable storage medium, and computer program product for predicting the risk of post-stroke depression.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: The first aspect of this invention provides a method for predicting the risk of post-stroke depression, the method being performed by a computer, the method comprising: Data Acquisition: Acquire predictive indicator data from the subject sample, including gender, comorbidities, NIHSS, and MLR.

[0007] Data processing: The predictive index data is input into the constructed post-stroke depression risk prediction model, which predicts whether the subject has a risk of depression based on the predictive index data.

[0008] Output results.

[0009] Furthermore, the steps for constructing the post-stroke depression risk prediction model are as follows: obtaining prediction indicator data, including gender, comorbidities, NIHSS, and MLR; and inputting the prediction indicator data into a machine learning algorithm to construct the post-stroke depression risk prediction model.

[0010] Furthermore, the construction steps of the post-stroke depression risk prediction model also include validating the model's effectiveness. The methods for validating the model's effectiveness include ROC curves, calibration curves, decision analysis curves, and SHAP analysis.

[0011] Furthermore, the post-stroke depression risk prediction model is a nomogram model.

[0012] Furthermore, the post-stroke depression risk prediction model obtains results using the following criteria: when the predicted score calculated after substituting the predicted index data into the prediction model is higher than a threshold, the subject is classified as having a risk of depression; if the predicted score calculated after substituting the predicted index into the prediction model is lower than a threshold, the subject is classified as not having a risk of depression.

[0013] Furthermore, the subjects were stroke patients.

[0014] Furthermore, the machine learning algorithm includes algorithmic models developed using various development tools; The development tools include TensorFlow, Scikit-Learn, PyTorch, OpenNN, RapidMiner, Azure Machine Learning, Apache Mahout, Shogun, KNIME, Vertex AI, H2Oai, Anaconda, Keras, Tableau, Fast.ai, Catalyst, Amazon ML, MLJAR, and Spell. The algorithm models include linear regression, logistic regression, Lasso regression, Ridge regression, linear discriminant analysis, nearest neighbor, decision tree, multilayer perceptron, neural network, support vector machine, naive Bayes, AdaBoost, GBDT, XGBoost, LightGBM, CatBoost, and random forest.

[0015] Furthermore, the algorithm model is a regression analysis model.

[0016] A second aspect of the present invention provides a prediction model for post-stroke depression risk, wherein the prediction indicators of the prediction model include gender, comorbidities, NIHSS, and MLR.

[0017] Furthermore, the prediction model is a nomogram model.

[0018] In this invention, the term "nomogram model," also known as a nomogram model, is a clinical predictive model based on multivariate regression analysis. It integrates multiple predictive indicators and then plots them on a plane using scaled line segments at a specific ratio. This expresses the relationships between the variables in the predictive model and assesses disease risk. Nomograms transform complex regression equations into visual graphs, making the predictive model results more readable and facilitating subject assessment. They are widely used in medical research and clinical practice.

[0019] In a specific embodiment of the present invention, the nomogram includes seven straight lines arranged from top to bottom and parallel to each other. Each straight line represents a scale with graduations. Specifically, the first row is a score scale, ranging from 0 to 100; the second row is a gender scale, with 2 for females and 1 for males; the third row is a comorbidity scale, ranging from 0 to 3; the fourth row is a NHISS scale, ranging from 0 to 22; the fifth row is an MLR scale, ranging from 0.1 to 1; the sixth row is a total score scale, ranging from 0 to 200; and the seventh row is a risk scale for PSD, ranging from 0.1 to 0.9.

[0020] The third aspect of this invention provides the application of the predictive indicators described in the first aspect of this invention in constructing a predictive model for post-stroke depression risk.

[0021] A fourth aspect of the present invention provides a system for predicting the risk of post-stroke depression, the system comprising: Data acquisition unit: used to acquire predictive indicator data of subjects, including gender, comorbidities, NIHSS, and MLR.

[0022] Data classification unit: used to classify and predict the post-stroke depression risk prediction model obtained by the data acquisition unit through the construction method described in the first aspect of the present invention, and to obtain the classification result of whether the subject has a risk of depression.

[0023] Result output unit: Used to output classification results.

[0024] Furthermore, the subjects were stroke patients.

[0025] The fifth aspect of the present invention provides a computer device, the device comprising: a memory and a processor, the memory being used to store program instructions; the processor being used to invoke the program instructions, and when the program instructions are executed, to implement the steps of the method for predicting post-stroke depression risk described in the first aspect of the present invention.

[0026] The sixth aspect of the present invention provides a computer-readable storage medium or computer program product, wherein the computer-readable storage medium stores a computer program, and the computer program product includes a computer program that, when executed by a processor, implements the steps of the method for predicting post-stroke depression risk as described in the first aspect of the present invention.

[0027] Advantages and beneficial effects of the present invention: This invention identifies gender, comorbidities, NIHSS, and MLR as independent influencing factors for the occurrence of post-stroke depression (PSD). A nomogram model for predicting PSD is constructed using gender, comorbidities, NIHSS, and MLR as key indicators. Training and independent validation sets demonstrate that the prediction model provided by this invention has good predictive efficacy, discrimination, and calibration, and can robustly distinguish between high-risk and low-risk PSD patients. This model provides an efficient and practical quantitative tool for the early identification of high-risk PSD patients, and has broad application prospects and translational value. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the method for predicting the risk of post-stroke depression provided by the present invention.

[0029] Figure 2 A flowchart for case screening and model building.

[0030] Figure 3 The independent predictive value of the predictive indicators of this invention is given by: A is the ROC curve of MLR, B is the ROC curve of NHISS, C is the ROC curve of gender, and D is the ROC curve of comorbidities.

[0031] Figure 4 This is the nomogram prediction model of the present invention.

[0032] Figure 5 For the performance verification of the prediction model of this invention, A is the ROC curve of the training set and B is the ROC curve of the validation set.

[0033] Figure 6 Here is the calibration curve of the prediction model of this invention, where A is the training set and B is the validation set.

[0034] Figure 7 This is the decision analysis (DCA) curve of the prediction model of this invention, where A is the training set and B is the validation set.

[0035] Figure 8 This is a SHAP analysis diagram of the prediction indicators of the prediction model of this invention.

[0036] Figure 9 The waterfall plot shows the predictive interpretation of a representative case for the predictive model of this invention.

[0037] Figure 10 This is a diagram showing the specific relationship between the prediction indicators of the prediction model of this invention and the SHAP value. Detailed Implementation

[0038] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0039] In some of the processes described in the specification, claims, and accompanying drawings of this invention, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] Figure 1 This is a flowchart illustrating the method for predicting post-stroke depression risk provided by the present invention. Specifically, the method includes: 101: Data Acquisition: Acquire predictive indicator data from the subject sample, including gender, comorbidities, NIHSS, and MLR.

[0042] In this invention, the prediction refers to assessing the risk of depression in stroke patients. In some embodiments, the risk should be assessed relative to the average risk of the target population. The prediction of depression risk herein means, through the indicators of this invention, whether the subject is assigned to a group at high risk of depression or a group at low risk of depression.

[0043] In some implementations, the subject may be human or non-human and may include, for example, animal strains or species used as a "model system" for research purposes. Similarly, subjects may include adults or adolescents (e.g., children). Furthermore, subjects may refer to mammals (e.g., humans or non-humans). Examples of mammals include, but are not limited to, any member of the mammalian class: humans, non-human primates (e.g., chimpanzees) and other apes and monkeys; livestock such as cattle, horses, sheep, goats, and pigs; domestic animals such as rabbits, dogs, and cats; and laboratory animals including rodents such as rats, mice, and guinea pigs. Examples of non-mammals include, but are not limited to, birds, fish, etc.

[0044] In a specific embodiment of the present invention, the case screening and model building process is as follows: Figure 2 As shown. The subjects were 320 patients with acute ischemic stroke hospitalized in the Department of Neurology, Shihezi People's Hospital, between January 2020 and January 2025. They were randomly assigned to the training set (n=314) and the test set (n=96) in a 7:3 ratio. Inclusion criteria included: age ≥18 years; first-time acute ischemic stroke (onset time ≤7 days), diagnosis conforming to the Chinese Guidelines for the Diagnosis and Treatment of Acute Ischemic Stroke 2023; neurological deficit symptoms lasting ≥24 hours; and new infarct lesions confirmed by cranial MRI / DWI. Exclusion criteria included: history of depressive disorder; severe aphasia or cognitive impairment preventing completion of assessment; comorbid severe infection, tumor, or immune system disease; and use of immunosuppressants or anti-inflammatory drugs prior to admission. The diagnostic criteria for PSD are as follows: at the 3-month follow-up after stroke, a score of ≥7 on the Hamilton Depression Scale (HAMD-17) is defined as PSD. The assessment method is as follows: the assessment is performed by two trained psychiatrists in a blinded manner, and if there is a discrepancy, a third senior physician makes the judgment.

[0045] Patient clinical data collected included baseline information: age, sex, race, marital status, education level, past medication history, smoking history, alcohol consumption history, hypertension, diabetes, coronary artery disease, etc.; stroke characteristics: NIHSS score (neurological deficit), infarct location (anterior circulation / posterior circulation / lacunar); laboratory indicators: CBC indicators: fasting venous blood was collected within 24 hours of admission and tested using a fully automated hematology analyzer. CBC-derived inflammatory markers were calculated: neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), systemic immune inflammatory index (SII), monocyte-to-lymphocyte ratio (MLR), and systemic inflammatory response index (SIRI).

[0046] Based on the 3-month follow-up results after stroke, patients were divided into those with post-stroke seizure disorder (PSD) and those without. Statistical analysis was performed on various indicators as variables. For continuous variables that conformed to a normal distribution, the mean ± standard deviation (χ ± S) was used, and the differences between the two groups were compared using an independent samples t-test. For continuous variables that did not conform to a normal distribution, the median (M) and quartiles (Q1, Q3) were used for description, and the Mann-Whitney U test was used for comparison.

[0047] We used univariate and multivariate logistic regression models to analyze the influencing factors of PSD occurrence. Univariate analysis compared clinical characteristics and inflammatory markers between the PSD group and the non-PSD group (t-test / chi-square test). Multicollinearity was tested using the variance inflation factor (VIF), and variables with VIF > 5 were removed. To optimize the fit of the multivariate logistic regression model, we used the Akaike Information Criterion (AIC) for backward stepwise selection and incorporated the selected variables into a nomogram to predict the probability of PSD occurrence.

[0048] The model's discriminative ability is visualized and quantified by plotting the Receiver Operating Characteristic Curve (ROC) and calculating the area under the curve (AUC). Discriminative power assessment: ROC-AUC and 95% confidence intervals are used to calculate sensitivity, specificity, and accuracy. Calibration assessment: Hosmer-Lemeshow goodness-of-fit test is used to plot the calibration curve. Clinical applicability assessment: Decision Curve Analysis (DCA) is used to assess net benefit at different thresholds.

[0049] Data analysis was performed using R 4.1.2 and Python 3.9. Continuous variables were expressed as mean ± standard deviation. Mean values ​​are expressed as χ²(S) or median (interquartile range), and categorical variables are expressed as frequencies (percentages). All statistical tests are two-tailed, and p < 0.05 is considered statistically significant.

[0050] In a specific embodiment of the present invention, a total of 320 patients with acute ischemic stroke were included, of whom 97 had PSD and 223 did not, with a PSD incidence rate of 30.31%. Comparative analysis of demographic data is shown in Table 1. Statistically significant differences were found between the PSD group and the non-PSD group in terms of gender (χ²=14.57, P<0.001), marital status (χ²=8.69, P=0.003), smoking status (χ²=13.35, P=0.001), and alcohol consumption status (χ²=4.66, P=0.031). There were no significant differences between the two groups in terms of race (P=0.756), education level (P=0.568), and age group (P=0.356).

[0051] Table 1 Comparison of demographic data between the two groups of patients

[0052] The results of the clinical data comparison are shown in Table 2. There were statistically significant differences between the two groups in NIHSS score, neutrophil count, MLR, history of long-term medication and number of comorbidities, while no significant differences were found in most blood routine indicators, inflammation ratio and BMI grouping.

[0053] Table 2 Comparison of Clinical Data

[0054] 102: Data Processing: Input the predicted index data into the constructed post-stroke depression risk prediction model, which predicts whether the subject has a risk of depression based on the predicted index data.

[0055] In a specific embodiment of the present invention, the clinical data in Table 2 were initially screened to obtain candidate variables related to the outcome variable (occurrence of PSD) (P<0.05). Univariate logistic regression analysis was then performed, and the specific variable assignments are detailed in Table 3. The results of the univariate logistic regression analysis showed that gender, marital status, smoking status, number of comorbidities, NIHSS score, and MLR were independent influencing factors for PSD (P<0.05), while race, education level, alcohol consumption, OCSP classification, history of long-term medication use, age group, BMI group, and most inflammation ratios (PLR, SII, SIRI, PIV, NLR) did not show independent predictive effects (P>0.05), as shown in Table 4.

[0056] Table 3 Variable assignment details

[0057] Table 4 Univariate Logistic Regression Analysis of PSD Risk Factors

[0058] Further multivariate analysis of variables with univariate p-values ​​< 0.05 (gender, marital status, smoking status, comorbidities, NIHSS, MLR) revealed that gender, comorbidities, NIHSS, and MLR were independent risk factors for PSD. Male sex was a protective factor against PSD (OR = 0.26, 95% CI: 0.11–0.59, P = 0.001 compared to female sex). Comorbidities: Multiple comorbidities were independent risk factors for PSD. Compared to those without comorbidities, patients with two comorbidities (OR = 5.82, 95% CI: 1.382–4.57, P = 0.017) or three comorbidities (OR = 5.90, 95% CI: 1.382–5.19, P = 0.017) had a significantly increased risk of developing PSD, while those with one comorbidity had no significant effect (P = 0.234). Neurological deficits: Higher NIHSS scores were associated with a higher risk of PSD (OR=1.28, 95%CI: 1.16–1.42, P<0.001). Inflammatory markers: MLR (monocyte / lymphocyte ratio) was a strong predictor of PSD (OR=1556.73, 95%CI: 118.06–20526.74, P<0.001), see Table 5. Multivariate analysis showed that the influencing factors such as gender, comorbidities, NIHSS score, and MLR all had VIF values ​​<5 and were close to 1, indicating independence among the variables and excluding collinearity interference, see Table 6.

[0059] Table 5. Multivariate Logistic Regression Analysis of PSD Risk Factors

[0060] Table 6 Collinearity Analysis

[0061] In a specific embodiment of the present invention, the predictive efficacy of gender, comorbidities, NIHSS, and MLR was validated using ROC curves, and the results are shown in […]. Figure 3 The AUC (95% CI) for MLR was 0.830 (0.784-0.877), the AUC (95% CI) for NHISS was 0.724 (0.664-0.784), the AUC (95% CI) for sex was 0.616 (0.559-673), and the AUC (95% CI) for comorbidities was 0.686 (0.627-0.745).

[0062] Nonographs were constructed using factors such as gender, comorbidities, NIHSS score, and MLR. The nonograph model allows for a visual observation of the scores associated with each independent risk factor. In the nonograph, the scale value of each predictor corresponds to a score on the scoring scale. All predictor variables are summed to obtain a total score, which reflects the probability of a poor prognosis. For example, a female patient with acute ischemic stroke (20 points), comorbidities (25 points), NIHSS score of 11 (40 points), and MLR score of 0.55 (50 points) has a total score of 20+25+40+50=135, with a PSD risk of 0.9. (See...) Figure 4 .

[0063] After modeling, the AUC and C-index of the training set were 0.90 (95% CI: 0.86–0.94), while the AUC and C-index of the validation set were 0.84 (95% CI: 0.75–0.93). This indicates that the model has good discriminative ability. Figure 5 Table 7.

[0064] Table 7. ROC details for the prediction model

[0065] The Hosmer-Lemeshow test p-value for the model is >0.05, indicating a good match between the predicted and observed probabilities in the two calibration curves, demonstrating the model's excellent calibration capability. (See...) Figure 6 The DCA curve also showed that the model curve was significantly higher than both the full intervention line and the full non-intervention line, indicating the model's clinical applicability. The model demonstrated significant net clinical benefits in both the modeling and validation groups. (See attached diagram.) Figure 7 .

[0066] Based on SHAP value analysis, this study evaluated the relative importance of each feature in the final prediction model for predicting PSD. Measured by the mean absolute SHAP value (mean(|SHAP value|)), the results showed that the importance of each feature was ranked as follows: MLR > NIHSS score > gender > number of comorbidities (see [link to relevant documentation]). Figure 8 MLR (mean(|SHAP value|)=1.00) was the most important factor, making it the most critical determinant in the model; followed by the NIHSS score (mean(|SHAP value|)=0.75), which reflects the severity of neurological deficits; gender (mean(|SHAP value|)=0.50) and the number of comorbidities (mean(|SHAP value|)=0.25) also contributed to the model's predictions, but their impact was relatively small, see [link to relevant documentation]. Figure 8 .

[0067] In a specific embodiment of the present invention, a predictive explanation waterfall plot was drawn using a representative case. Figure 9 The baseline predictive value for this case was -1.28. The four features included in the model contributed to the final predicted value (f(x) = -0.0402) in different directions and to varying degrees: Features increasing the risk of PSD: A higher MLR value (MLR=0.575) was the largest risk driver, increasing the predicted value by +1.59; comorbidities (three diseases) also increased the predicted value by +0.699. Features decreasing the risk of PSD: Female (sex=2) decreased the predicted value by -0.708; a lower NIHSS score (NIHSS=5) decreased the predicted value by -0.335. The sum of these feature contributions yielded the final predicted value (-0.0402) for this case, corresponding to a probability of approximately 49% for PSD, indicating that this case was at a borderline risk level for PSD. This analysis intuitively reveals the model's decision-making logic for specific individuals, namely, comprehensively weighing the level of inflammation, comorbidity burden, sex, and the severity of neurological deficits before arriving at a predictive conclusion.

[0068] Figure 10The specific correlation patterns between the values ​​of each feature (MLR, NIHSS score, gender, and number of comorbidities) in the final predictive model and their predictive contribution (SHAP value) were visualized. The results show that for continuous variables MLR and NIHSS score, there is a clear positive dose-response relationship between their feature values ​​and SAP values. Higher MLR values ​​correspond to larger SAP values ​​(changing from negative to positive), indicating that high MLR levels are the main factor driving the model to predict an individual's high risk of PSD. Similarly, higher NIHSS scores are associated with larger SAP values, suggesting that more severe neurological deficits continuously increase the predicted risk of PSD. For categorical variables gender and number of comorbidities, the relationship between feature category and SAP values ​​is clear. Samples with gender "2" (female) are concentrated in the negative SAP value region, while samples with gender "1" (male) are mostly located in the positive SAP value region, directly confirming the conclusion that males are a protective factor against PSD. The more comorbidities, the larger the corresponding SAP values ​​tend to be, indicating that an increased comorbidity burden is a clear risk factor.

[0069] 103: Output results.

[0070] It should be understood that the systems, devices, and methods described in this invention can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the couplings or direct couplings or communication connections shown or discussed may be indirect couplings or communication connections through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0071] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0072] Furthermore, the functional modules in this embodiment of the invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules can be implemented in hardware or as software functional modules.

[0073] The above are merely embodiments of this application and do not limit the scope of this patent application. Any equivalent structural or procedural changes made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of this application.

Claims

1. A method for predicting the risk of post-stroke depression, characterized in that, The method is performed by a computer, and the method includes: Data Acquisition: Acquire predictive indicator data from the subject sample, including gender, comorbidities, NIHSS, and MLR; Data processing: The predictive index data is input into the constructed post-stroke depression risk prediction model, which predicts whether the subject has a risk of depression based on the predictive index data; Output results.

2. The method according to claim 1, characterized in that, The steps for constructing the post-stroke depression risk prediction model are as follows: obtain prediction index data, including gender, comorbidities, NIHSS, and MLR; input the prediction index data into a machine learning algorithm to construct the post-stroke depression risk prediction model. Preferably, the construction step of the post-stroke depression risk prediction model further includes validating the model efficacy, and the methods for validating the model efficacy include ROC curves, calibration curves, decision analysis curves, and SHAP analysis. Preferably, the post-stroke depression risk prediction model is a nomogram model.

3. The method according to claim 1, characterized in that, The post-stroke depression risk prediction model obtains results using the following criteria: when the predicted score calculated after substituting the predicted index data into the prediction model is higher than a threshold, the subject is classified as having a risk of depression; if the predicted score calculated after substituting the predicted index data into the prediction model is lower than a threshold, the subject is classified as not having a risk of depression.

4. The method according to claim 1, characterized in that, The subjects were stroke patients.

5. The method according to claim 1, characterized in that, The machine learning algorithms include algorithm models developed using various development tools; The development tools include TensorFlow, Scikit-Learn, PyTorch, OpenNN, RapidMiner, Azure Machine Learning, Apache Mahout, Shogun, KNIME, Vertex AI, H2Oai, Anaconda, Keras, Tableau, Fast.ai, Catalyst, Amazon ML, MLJAR, and Spell. The algorithm models include linear regression, logistic regression, Lasso regression, Ridge regression, linear discriminant analysis, nearest neighbor, decision tree, multilayer perceptron, neural network, support vector machine, naive Bayes, AdaBoost, GBDT, XGBoost, LightGBM, CatBoost, and random forest. Preferably, the algorithm model is a regression analysis model.

6. A risk prediction model for post-stroke depression, characterized in that, The predictive indicators of the predictive model include gender, comorbidities, NIHSS, and MLR; Preferably, the prediction model is a nomogram model.

7. The application of the predictive index as described in claim 1 in constructing a post-stroke depression risk prediction model.

8. A system for predicting the risk of post-stroke depression, characterized in that, The system includes: Data acquisition unit: used to acquire predictive indicator data of subjects, including gender, comorbidities, NIHSS, and MLR; Data classification unit: used to classify and predict the post-stroke depression risk prediction model obtained by the data acquisition unit through the construction method described in claim 2, and obtain the classification result of whether the subject has a depression risk; Result output unit: Used to output classification results; Preferably, the subject is a stroke patient.

9. A computer device, characterized in that, The device includes: a memory and a processor, the memory being used to store program instructions; the processor being used to invoke the program instructions, which, when executed, implement the steps of the method for predicting post-stroke depression risk as described in any one of claims 1-5.

10. A computer-readable storage medium or computer program product, characterized in that, The computer-readable storage medium stores a computer program, the computer program product comprising a computer program that, when executed by a processor, implements the steps of the method for predicting post-stroke depression risk as described in any one of claims 1-5.

Images