Prediction of postoperative haemorrhage after cerebral tumour surgery using machine learning algorithms

Prediction of postoperative haemorrhage after cerebral tumour surgery using machine learning algorithms | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Prediction of postoperative haemorrhage after cerebral tumour surgery using machine learning algorithms Yasin Göktürk, Seyit Kağan Başarslan, Şule Göktürk, Hikmet Kocaman, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7012786/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 23 Oct, 2025 Read the published version in BMC Medical Informatics and Decision Making → Version 1 posted 14 You are reading this latest preprint version Abstract Background Postoperative intracranial hemorrhage is a critical complication following cerebral tumor surgery, often associated with increased morbidity and mortality. This study aimed to predict the risk of postoperative intracerebral hemorrhage in patients undergoing intracranial tumor surgery by employing machine learning (ML) algorithms for risk stratification and identifying key contributing factors. Methods A retrospective analysis was conducted on 118 patients who underwent intracranial tumor surgery and were monitored in the neurosurgical intensive care unit between January 2024 and January 2025. Patients with radiologically confirmed hematomas ≥ 5 cm³ on brain CT within 24–48 hours postoperatively were classified as "Positive" for bleeding, while others were labeled "Negative." Clinical and biochemical parameters were analyzed using SPSS and R. Multiple ML algorithms—including Bagging MARS, Boosting C5.0, SVM, and Random Forests—were developed and evaluated using performance metrics such as AUC, F-score, accuracy, and Brier score. Results The Bagging MARS model demonstrated superior predictive performance, with a test AUC of 0.8693, accuracy of 80.8%, Brier score of 0.1580, and F-score of 0.8649. Platelet count, serum sodium level, and glomerular filtration rate (GFR) emerged as the most influential predictors of hemorrhage. Model explainability was enhanced using SHAP and LIME analyses, offering both global and local interpretability of the predictions. Conclusion ML algorithms, particularly Bagging MARS, show high accuracy in predicting postoperative hemorrhage following brain tumor surgery. Biomarkers such as platelet count, sodium, and GFR offer clinically meaningful insights for early risk detection and intervention. Integration of these predictive models into clinical decision support systems may significantly improve postoperative monitoring and patient outcomes. Brain tumor Machine learning Hemorrhage Risk prediction Artificial intelligence Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Neurosurgical surgery means " craniotomy ". It includes basic techniques used to reach the surgical area in intracranial pathologies. It consists of a series of technical methods used in cerebral tumors. Management of bleeding-coagulation is difficult in intracerebral tumor surgery (ICTS) and subsequent intensive care follow-ups. In this struggle, avoiding intracranial bleeding is mandatory and has prognostic value. Perioperative bleeding can be associated with a number of factors including anticoagulant drugs and coagulation status. In addition, it is also related to the characteristic-morphological features of the intracranial mass and its intracerebral location [ 1 ]. After craniotomy performed in ICTS, infections and bleeding in the surgical cavity can be a common and potentially fatal complication in neurosurgical intensive care [ 2 ]. Regular neurological assessment in the intensive care unit or recovery room is extremely important. Unexpected onset or worsening of neurological signs and any deterioration in the patient's state of consciousness should prompt immediate neuroradiological investigation. Urgent computed tomography of the brain is the method of choice if intracranial complications are suspected after brain tumor removal [ 3 ]. Although it is controlled with computed tomography (CT) within the first 24 hours postoperatively [ 4 ], when bleeding occurs, it causes a decrease in the patient's Glasgow coma score and may require reoperation. It is vital to be able to foresee this situation and take precautions. Perioperative bleeding can be associated with a number of factors, including antihemostatic and anticoagulant medications [ 5 ]. On the other hand, the risk of venous thromboembolism is also increased after elective and emergency neurosurgery, and a hypercoagulable state has been described in a few patients, particularly after surgery for ICTS [ 6 ]. It is important to know what changes occur in coagulation in the perioperative period to balance the risk of thrombosis against the risk of bleeding. Routine coagulation tests include activated partial thromboplastin time (aPTT), prothrombin time (PT), platelet count (Plt) and fibrinogen levels. A study conducted to determine the incidence of postoperative hematoma and to identify risk factors for perioperative bleeding disorders during the 5-year period from 1989 to 1993 found that two-thirds of patients had risk factors for perioperative bleeding disorders. Administration of antiplatelet agents (aspirin and nonsteroidal antiinflammatory drugs) was the most frequently associated risk factor. At least 75% of these identified risk factors were potentially preventable or amenable to correction [ 7 ]. In patients who develop bleeding at the surgical site, blood transfusions and blood component therapy can be life-saving in life-threatening bleeding situations. In recent years, machine learning algorithms have gained prominence in healthcare for their ability to detect complex patterns in clinical data and support predictive decision-making. Particularly in neurosurgery, these methods offer promising potential for early identification of high-risk patients, enabling timely interventions and improved outcomes [ 8 – 10 ]. However, their application in predicting postoperative intracerebral hemorrhage remains limited, highlighting the need for targeted studies such as the present research.To our knowledge, there is no study on machine learning that can specifically predict postoperative hemorrhage in patients who have undergone intracerebral tumor surgery. Therefore, the aim of the current study was to predict postoperative bleeding using machine learning algorithms using clinical and laboratory results of patients after brain tumour surgery. Consent to participate was obtained from all patients included in the study. Material and Method Study design This study covers a retrospective examination of patients who underwent surgery with a diagnosis of intracerebral tumor between January 2024 and January 2025, with the approval of the Kayseri City Hospital Ethics Committee determined by the institutional or regional responsible committee ( Ethics Committee approval number: 346/25.02.2025 ) and in accordance with the principles stated in the 1975 Helsinki Declaration, revised in 1983 . Intracerebral tumor patients who underwent surgery in a single center were followed up in the Kayseri City Hospital Neurosurgery Clinic unit after surgery. In the postoperative clinical follow-up of the operated 118 patients, age, gender, systemic inflammatory response markers, intensive care unit stay, and biochemical parameters were recorded. Patients with radiologically confirmed hematoma of at least 5 cm³ on CT at 24–48 hours postoperatively were classified as 'Positive' for bleeding. Inclusion cirteria These are all patients who were operated on with the diagnosis of "intracranial mass" in the K****C**** H***** Neurosurgery Clinic for 1 year in 2024. Exclusion criteria Patients who were followed up in the Postoperative Neurosurgery Intensive Care Unit and had serious systemic diseases that could affect biochemical and hematological parameters, or systemic inflammatory diseases, bleeding diathesis, or using anticoagulants (except those using prophylactic low molecular weight anticoagulants) were excluded from the study. Volume measurement process with computerized tomography Width-length and height values ​​of the hematoma were measured in the postoperative 1st day brain tomography and its volume (cm 3 ) was calculated. Sodium, Calcium, PT, aPTT, Fibrinogen and Platelet values ​​will also be included in the study and were compared with other markers with the blood volume measurement calculated on the first postoperative day. Machine Learning Algorithms and Experimental Settings During the first stage of analysis process, statistical analysis was performed using R (version 4.4.2; R Foundation for Statistical Computing, Vienna, Austria) and IBM SPSS Statistics for Windows (version 26.0; IBM Corp., Armonk, N.Y., USA). A p-value less than 0.05 was considered statistically significant. Descriptive statistics were used to summarize patient characteristics and measurements; continuous variables were presented as mean ± standard deviation (sd), while categorical variables were presented as numbers (n) and percentages (%). For the comparison of baseline characteristics between the bleeding (Positive) and no bleeding (Negative) groups, independent samples t-tests were used for continuous variables such as Age and Length of ICU Stay. Categorical variables including Gender, Progression, Nutrition, Comorbidity, HT (Hypertension), DM (Diabetes Mellitus), and Cancer were analyzed using Pearson's Chi-square test of independence or Fisher's exact test, as appropriate, based on the expected cell counts. For the comparison of various laboratory measurements between the two groups, independent samples t-tests or t-tests with Welch correction (for unequal variances) were employed for variables such as White Blood Cell count (WBC), new WBC count (WBCNew), Neutrophil count, Lymphocyte count, Hemoglobin, Platelet count (PLT), Prothrombin Time (PT), activated Partial Thromboplastin Time (aPTT), International Normalized Ratio (INR), Fibrinogen, Aspartate Aminotransferase (AST), Alanine Aminotransferase (ALT), Blood Urea Nitrogen (BUN), Creatinine, Glomerular Filtration Rate (GFR), Albumin (ALB), Preoperative C-Reactive Protein (PreoperativeCRP), Postoperative C-Reactive Protein (PostoperativeCRP), Sodium (NA), Potassium (K), Chloride (CL), Magnesium (MG), and Calcium (CA). The specific test used for each variable is indicated in the table footnotes. As the main scope of this study, a comprehensive machine learning approach is employed to classify individuals based on their risk of brain hemorrhage and to identify key contributing factors. Similar to statistical analysis, all analyses were performed using R statistical software (version 4.4.2). The initial dataset underwent several preprocessing steps. An outlier analysis was conducted using Tukey's method, z-scores, and box plots; fifteen units commonly identified as outliers by these methods were excluded from the study. The primary outcome variable, "Bleeding Situation," was derived from the original bleeding indicator, categorizing individuals as "Negative" (no hemorrhage) or "Positive" (hemorrhage). The dataset was then partitioned into training (75%) and testing (25%) sets using stratified sampling based on the "Bleeding Situation" variable to ensure proportional representation in both sets. A comprehensive preprocessing pipeline was constructed using the recipes package. This involved median imputation for all numeric variables and mode imputation for nominal variables. Subsequently, zero-variance variables were removed, and numeric variables exhibiting high correlation (threshold of 0.85) were eliminated. Numeric variables were then centered and scaled. Finally, all nominal variables were converted into numerical format using one-hot encoding. Machine learning algorithms Model development and hyperparameter optimization were conducted within the tidymodels framework. A k-fold cross-validation strategy, specifically 10-fold cross-validation repeated 5 times, was applied to the training data, stratified by the "Bleeding Situation" variable. This resampling technique was used for robust hyperparameter tuning across a predefined grid of sixty hyperparameter combinations for each model. The primary metric for selecting the optimal hyperparameter set during the tuning process was the Area Under the Receiver Operating Characteristic Curve (ROC AUC). A wide array of classification algorithms was implemented, including Logistic Regression, K-Nearest Neighbors (KNN), Naive Bayes, Classification and Regression Trees (CART), Support Vector Machines (SVM) with linear, polynomial, and radial basis function kernels, Multilayer Perceptron (NNET), C5.0 decision trees and rule-based models, Multivariate Adaptive Regression Splines (MARS), Random Forests, Bayesian Additive Regression Trees (BART), Partial Least Squares (PLS), RuleFit, bagged versions of CART, MARS, MLP, and C5.0, and boosted models including XGBoost, boosted C5.0, and LightGBM. Regularized logistic regression models, namely Ridge, Lasso, and Elastic Net, were also evaluated. Workflows encapsulating the preprocessing recipe and model specifications were created for each algorithm, and parallel processing (doParallel package) was utilized to expedite the tuning process. Model performance was evaluated using a focused set of metrics, including ROC AUC, Brier score, accuracy, and F-measure, collected using the yardstick package. To identify the most influential variables, variable importance plots were generated. Furthermore, model explainability was enhanced through the creation of SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) plots. ROC curves were also plotted to visualize the trade-off between true positive rate and false positive rate. After identifying the optimal hyperparameters for each model based on the cross-validation results on the training set (primarily maximizing ROC AUC), the final models were fitted to the entire training dataset and subsequently evaluated on the unseen test set. The performance metrics on the test set were then compiled to compare the generalization ability of the models. Visualizations such as confusion matrices were used to assess classification performance on the test data. Key R packages pivotal to this analysis included tidyverse for general data manipulation, tidymodels, caret, ranger, glmnet, kernlab, xgboost, C50, nnet, earth, kknn, rpart, mixOmics, and lightgbm. The results, including training performance, test performance, and optimal hyperparameters for each model, were systematically organized and saved. Results The demographic and clinical characteristics of the patients are presented in Table 1 . No statistically significant difference was found between the groups in terms of age, gender, hypertension (HT), diabetes mellitus (DM), and dietary method (p > 0.05). In other parameters (duration of intensive care unit stay, progression, nutrition, comorbidity, presence of cancer), a statistically significant difference was observed between the groups (p < 0.05) (Table 1 ). Table 2 shows the comparison of laboratory parameters between individuals with and without cerebral haemorrhage. There was a statistically significant difference between the groups in parameters such as lymphocytes, PLT, BUN, ALB, postoperative CRP, Na, K, Cl, Mg (p 0.05) (Table 2 ). Table 1 The summary characteristics of patients considered in the study Bleeding p Yes (Positive) No (Negative) Age, mean ± sd 60.26 ± 17.65 56.81 ± 15.97 0.289 a Length of ICU Stay, mean ± sd 10.23 ± 10.58 5.1 ± 6.87 0.002 a Gender, n(%) Male 23 (58.97%) 50 (63.29%) 0.650 b Female 16 (41.03%) 29 (36.71%) Progression, n(%) Discharged 31 (79.49%) 77 (97.47%) 0.003 c Exitus 8 (20.51%) 2 (2.53%) Nutrition, n(%) Enteral 36 (92.31%) 78 (98.73%) 0.105 b TPN 3 (7.69%) 1 (1.27%) Comorbidity, n(%) Yes 26 (66.67%) 35 (44.3%) 0.022 b No 13 (33.33%) 44 (55.7%) HT, n(%) Yes 7 (17.95%) 12 (15.19%) 0.701 b No 32 (82.05%) 67 (84.81%) DM, n(%) Yes 7 (17.95%) 13 (16.88%) 0.886 b No 32 (82.05%) 64 (83.12%) Cancer, n(%) Yes 8 (20.51%) 5 (6.33%) 0.029 b No 31 (79.49%) 74 (93.67%) a : t-test, b : Fisher exact test, c :Chi-square test of independence Table 2 The comparison of various measurements based on bleeding situations Bleeding p Variable, mean ± sd Yes (Positive) No (Negative) WBC 14079.49 ± 6228.17 13559.37 ± 4908.82 0.650 d WBCNew 13.74 ± 5.91 13.58 ± 4.88 0.891 d Neutrophil 11.83 ± 5.6 11.59 ± 4.7 0.819 d Lymphocyte 0.93 ± 0.56 1.19 ± 0.73 0.033 d Hemoglobin 11.8 ± 1.96 12.38 ± 1.73 0.125 d PLT 209.18 ± 70.45 234.97 ± 68.86 0.025 e PT 12.74 ± 1.33 12.33 ± 1.25 0.114 d aPTT 22.2 ± 4.09 21.56 ± 3.28 0.400 d INR 1.11 ± 0.12 1.06 ± 0.13 0.068 d Fibrinogen 3817.65 ± 1260.86 3313.88 ± 1210.45 0.185 d AST 29.92 ± 44.85 20.05 ± 19.65 0.195 d ALT 22.1 ± 18.32 25.33 ± 28.92 0.463 d BUN 21.46 ± 10 16.88 ± 6.84 0.013 d Creatinine 0.74 ± 0.32 0.71 ± 0.18 0.606 d GFR 94.33 ± 20.89 99.16 ± 18.95 0.227 d ALB 32.79 ± 4.47 34.99 ± 3.8 0.011 d PreoperativeCRP 12.01 ± 25.64 5.35 ± 8.89 0.123 d PostoperativeCRP 64.08 ± 72.48 33.82 ± 38.29 0.018 d NA 140.79 ± 4.47 138.33 ± 3.47 0.004 d K 4.07 ± 0.53 4.26 ± 0.42 0.039 a CL 104.44 ± 4.02 102.73 ± 3.98 0.033 d MG 2.16 ± 0.28 2.03 ± 0.22 0.018 d CA 8.27 ± 0.64 8.35 ± 0.57 0.495 d BleedindWidth 22.38 ± 11 0 ± 0 - BleedingLength 17.4 ± 10.88 0 ± 0 - BleedingHeight 18.72 ± 8.7 0 ± 0 - BleedingVolume 12.41 ± 25.15 0 ± 0 - a : t-test, d : t-test with Welch correction This study undertook a comprehensive evaluation of multiple machine learning algorithms to accurately classify brain hemorrhage risk and to delineate the most significant predictive factors. The performance characteristics of these models, assessed on both the training dataset (via 5-repeated 10-fold cross-validation) and an independent test dataset, are systematically presented. Comparative Model Performance Table 3 provides a detailed comparison of all implemented machine learning models based on four key performance metrics: Accuracy, Brier Score (lower indicates better calibration), F-Score, and Area Under the Receiver Operating Characteristic Curve (AUC). On the crucial independent test dataset, several algorithms demonstrated notable predictive efficacy. The Bagging algorithm utilizing a Multivariate Adaptive Regression Splines (MARS) base learner (Bagging MARS) emerged as a particularly strong performer, achieving a test AUC of 0.8693, an accuracy of 0.8077 (21 out of 26 test cases correctly classified), a low Brier score of 0.1580, and a high F-Score of 0.8649. Other models exhibiting robust test performance included the Boosting algorithm with a C5.0 base learner (Boosting C5.0) , which recorded a test AUC of 0.8235 and an accuracy similar to Bagging MARS at 0.8073. The Support Vector Machine (SVM) with a Radial Basis Function (RBF) kernel also showed excellent discriminatory power with a test AUC of 0.8562, albeit with a slightly lower accuracy of 0.7538. Similarly, Random Forests yielded a high test AUC of 0.8431, though its accuracy was 0.7308. Given its leading AUC on the test set and a strong, balanced performance across accuracy and F-Score, the Bagging MARS model was selected for subsequent in-depth interpretability analyses. Table 3 Performance Metrics (Accuracy, Brier Score, F-Score, AUC) of Machine Learning Models on Training and Test Datasets for Brain Hemorrhage Classification. Model Split Accuracy Brier Score F Score AUC Bagging (CART learner) Train 0.7344 0.2290 0.7454 0.6828 Test 0.6923 0.1944 0.8000 0.7712 Bagging (MARS learner) Train 0.7878 0.2275 0.8790 0.7916 Test 0.8077 0.1580 0.8649 0.8693 Bagging (MLP learner) Train 0.7847 0.2068 0.7518 0.7253 Test 0.7692 0.1869 0.8333 0.7582 Bayesian Additive Trees Train 0.7741 0.2094 0.8023 0.7024 Test 0.6538 0.2020 0.7907 0.7353 Boosting (C5.0 learner) Train 0.8341 0.2467 0.7293 0.7787 Test 0.8073 0.1582 0.8571 0.8235 Boosting (LightGBM) Train 0.7646 0.2232 0.7979 0.6453 Test 0.7538 0.2264 0.7907 0.7686 Boosting (XGBoost) Train 0.7526 0.2173 0.7839 0.6033 Test 0.7923 0.1931 0.8000 0.7712 C5.0 Train 0.8025 0.2974 0.6910 0.6991 Test 0.8077 0.1576 0.8485 0.7876 C5.0 - Rules Train 0.7407 0.2263 0.7579 0.6933 Test 0.7692 0.1823 0.8235 0.7451 CART Train 0.6832 0.3975 0.6848 0.6592 Test 0.6923 0.2800 0.7143 0.7908 Elastic Net Train 0.8189 0.2588 0.7777 0.8087 Test 0.7692 0.2359 0.8235 0.7124 KNN Train 0.7667 0.3333 0.7235 0.6547 Test 0.6923 0.3077 0.7500 0.6863 Lasso Train 0.8185 0.2582 0.7789 0.8073 Test 0.7692 0.2365 0.8235 0.7059 Logistic Regression Train 0.8156 0.2838 0.7699 0.7291 Test 0.7692 0.2308 0.8235 0.7157 MARS Train 0.7443 0.2361 0.7723 0.6432 Test 0.7692 0.1674 0.8333 0.7418 Multilayer Perceptron Train 0.7854 0.2153 0.7457 0.7413 Test 0.7692 0.1891 0.8235 0.7680 Naive Bayes Train 0.7780 0.3071 0.7947 0.6367 Test 0.7692 0.2083 0.8421 0.7889 Partial Least Squares Train 0.8010 0.1947 0.7645 0.7262 Test 0.7308 0.1908 0.8000 0.6536 Random Forests Train 0.7912 0.2110 0.8080 0.6464 Test 0.7308 0.1741 0.8205 0.8431 Ridge Train 0.7741 0.2101 0.7550 0.7300 Test 0.7692 0.1841 0.8235 0.6863 RuleFit Train 0.7793 0.2855 0.7431 0.7869 Test 0.7310 0.2650 0.7880 0.7120 SVM (Linear kernel) Train 0.8076 0.1778 0.7643 0.7867 Test 0.7692 0.1604 0.8235 0.7124 SVM (Polynomial kernel) Train 0.7646 0.2128 0.7979 0.7353 Test 0.7538 0.1792 0.7907 0.8235 SVM (Radial kernel) Train 0.7646 0.2232 0.7979 0.7351 Test 0.7538 0.2264 0.7907 0.8562 In-depth Analysis of the Bagging MARS Model A detailed examination of the selected Bagging MARS model was conducted to understand its predictive mechanisms and identify the primary determinants of brain hemorrhage risk. · Variable Importance: Figure 1 illustrates the hierarchy of predictor importance as determined by the Bagging MARS model. Platelet count (PLT) was unequivocally identified as the most influential variable, with a relative importance score of 100.0. Following PLT, Sodium (Na) and Glomerular Filtration Rate (GFR) were found to be substantial contributors, with importance scores of 59.0 and 40.9, respectively. Other significant predictors included Postoperative C-Reactive Protein (CRP) (importance: 31.3) and Hemoglobin (importance: 30.8), underscoring their roles in the model's risk assessment. · Classification Accuracy and Discriminatory Power on Test Data: The confusion matrix ( Figure 2 ) for the Bagging MARS model on the test dataset (N=26) provides a granular view of its classification performance. The model correctly identified 16 out of 17 individuals without brain hemorrhage (True Negatives), resulting in a high specificity of 94.1%. For individuals who did experience brain hemorrhage, the model correctly classified 5 out of 9 cases (True Positives), yielding a sensitivity (recall) of 55.6%. The model made only 1 false positive prediction and 4 false negative predictions. The overall discriminatory capability of the model, as graphically represented by the Receiver Operating Characteristic (ROC) curve ( Figure 3 ), culminated in an Area Under the Curve (AUC) of 0.8693. This high AUC value signifies the model's robust ability to distinguish between patients at high and low risk of brain hemorrhage. · Global Model Interpretability (SHAP Analysis): The SHAP (SHapley Additive exPlanations) summary plot ( Figure 4 ) provides global insights into how feature values influence the model's prediction (on a log-odds scale). For Platelet count (PLT), higher values (represented by redder dots) are generally associated with an increased positive impact on the prediction of brain hemorrhage (dots shift to the right of the zero SHAP value line). Conversely, for a feature like Alanine Aminotransferase (ALT), higher values tend to have a negative impact on the log-odds of predicting hemorrhage (redder dots shift to the left). This visualization effectively captures the magnitude and direction of each predictor's influence across the dataset, for example, GFR shows that lower values (bluer dots) tend to increase the positive SHAP value, suggesting lower GFR is a risk factor. · Local Model Interpretability (LIME Analysis): LIME (Local Interpretable Model-agnostic Explanations) plots were employed to dissect individual predictions: o Figure 5a provides an explanation for a specific correct 'Positive' prediction (Test Set Index: 3; Actual: 'Positive', Predicted: 'Positive' with a probability of 0.64). For this instance, higher Age (contributing approximately +0.115 to the prediction weight), Na (+0.062), and BUN (+0.046) were the primary drivers pushing the prediction towards a positive hemorrhage outcome. o Figure 5b details a misclassification (Test Set Index: 8; Actual: 'Negative', Predicted: 'Positive' with a probability of 0.34). Here, features such as Cl (+0.032), AST (+0.026), and Hemoglobin (+0.026) incorrectly influenced the model to predict a positive outcome. Notably, a lower Age value for this patient provided a strong counter-signal (contribution: -0.115), but was overridden by other factors. These local explanations are critical for understanding model behavior on a case-by-case basis and for identifying potential reasons for specific prediction outcomes. Discussion Postoperative intracerebral hemorrhage remains a significant cause of morbidity and mortality following intracranial tumor surgery, despite advancements in surgical and perioperative care. Identifying patients at high risk for postoperative bleeding is crucial for optimizing intensive care management and reducing adverse outcomes. In this context, the present study demonstrates that machine learning algorithms—particularly the Bagging MARS model—can effectively predict hemorrhage risk using routinely collected clinical and laboratory data. The strong performance metrics of this model, including high accuracy and AUC values, underscore its potential as a clinical decision support tool for individualized risk stratification. In addition, Platelet count (PLT) was identified as the most influential variable with a relative importance score of 100.0. Following PLT, Sodium (Na) and Glomerular Filtration Rate (GFR) were found to be valuable contributors with importance scores of 59.0 and 40.9 respectively. Ziai et al stated in their study that platelet dysfunction is common in patients with spontaneous intracerebral hemorrhage and that low platelet count and platelet dysfunction are effective factors in the expansion of hematoma volüme [ 11 ]. Hyponatremia is a common electrolyte disorder in patients with neurological diseases; however, its predictive role for outcome in patients with supratentorial spontaneous intracerebral hemorrhage is controversial. In their study conducted in 2024, Qian et al. reported that hyponatremia occurring in the first 7 days after hemorrhage in patients with supratentorial spontaneous intracerebral hemorrhage was an independent predictor of 90-day mortality and adverse outcomes [ 12 ]. They reiterated the need for meticulous electrolyte monitoring in patients treated surgically. In our study, low sodium levels were also associated with poor prognosis. It is the second important parameter affecting the occurrence of postoperative hemorrhage. The impact of chronic kidney disease (CKD) on the severity and prognosis of spontaneous intracerebral hemorrhage (ICH) has rarely been studied. Li et al. found that lower baseline eGFR was associated with an increased risk of in-hospital mortality, nonroutine discharge, hemorrhagic stroke severity, and in-hospital complications such as pneumonia, hydrocephalus, and hematoma evacuation in patients with acute intracerebral hemorrhage [ 13 ]. Chan et al retrospectively reviewed a group of 35 patients who underwent cranial surgery and demonstrated perioperative thrombocytopenia (platelet count less than 150,000/microliter). Fourteen of 35 patients (40%) developed postoperative intracranial hematomas requiring reoperation, and seven (20%) died within 2 weeks after surgery. Analysis revealed that a perioperative platelet count of less than 100,000/microliter was associated with a higher risk of postoperative hematoma formation [ 14 ]. Postoperative bleeding rates after cranial surgery vary greatly in the literature, ranging from 0.77–50% [ 2 , 15 – 17 ]. However, the definition of postoperative bleeding varies. Series that include patients with radiographic evidence of bleeding have reported rates ranging from 10.8–50% [ 1 , 5 , 6 ]. In fact, the location of postoperative hematoma depends largely on preoperative pathology. Intraparenchymal hematomas have been frequently observed as locations after cranial operations. In routine intensive care follow-up, blood coagulation cascades in particular are routinely examined. In our study, Plt values, Sodium and GFR values ​​stood out as the three most important predictive parameters for postoperative bleeding in Machine Learning programs. We hope that MARS Bagging and the first three important parametric laboratory values ​​(Plt, Sodium and GFR) will contribute to making important decisions that will change patient prognosis. Gerlach et al retrospectively analyzed the data of 296 patients with and without hematoma who underwent surgery for meningioma to determine the risk factors associated with postoperative hematoma [ 18 ]. Meningioma surgery carries a higher risk for postoperative hematoma in the elderly. Thrombocytopenia and other hemostatic abnormalities are frequently associated with postoperative bleeding after meningioma surgery. Expansion of coagulation tests and specific replacement therapy may prevent hematoma formation and improve patient outcomes. Literature suggests that most hematomas are discovered within three days after the initial operation [ 18 ]. In the planning of our study, we focused on the patients' brain CT scans and laboratory values ​​taken at the 48th postoperative hour. Kalfas reported that 35% of ICH cases occurred within 12 hours [ 17 ], while Taylor reported that 44 out of 50 of 2,305 patients developed a hematoma within six hours after surgery [ 19 ]. Of course, it should not be forgotten that longer periods of observation in the intensive care unit may be required after intracerebral tumor surgery. Various risk factors for hematoma development have been analyzed, and hypertension is usually the most important comorbidity in most studies [ 2 , 17 ]. However, in our study, no statistically significant difference was found between the groups in terms of age, gender, hypertension (HT), diabetes mellitus (DM) and diet method. A statistically significant difference was observed between the groups in other parameters (length of stay in the intensive care unit, progression, nutrition, comorbidity, presence of cancer). Other risk factors include use of nonsteroidal anti-inflammatory drugs, use of anticoagulants, coagulopathies, thrombocytopenia, factor XIII deficiencies, and decreased fibrinogen levels. Our study undertook a comprehensive evaluation of multiple machine learning algorithms to accurately classify the risk of postoperative cerebral hemorrhage and identify the most important predictive factors. Of all the machine learning models implemented based on four key performance metrics, the Boosting Algorithm (Boosting C5.0) with C5.0-based learner recorded a test AUC of 0.8235 and an accuracy similar to Bagging MARS of 0.8073, demonstrating robust test performance. Similar to Bagging MARS, Random Forests produced a high test AUC of 0.8431, but its accuracy is 0.7308. A detailed review of the selected Bagging MARS model was performed to understand its predictive mechanisms and the primary determinants of cerebral hemorrhage risk were identified. In the Bagging MARS model; Platelet count (PLT) was determined as the undisputed most influential variable with a relative importance score of 100.0. Following PLT, Sodium (Na) and Glomerular Filtration Rate (GFR) were found to be significant contributors. Other significant predictors included Postoperative C-Reactive Protein (CRP) (importance: 31.3) and Hemoglobin (importance: 30.8), highlighting their role in risk assessment by the model. The model demonstrated its robust ability to distinguish between patients at high and low risk of brain hemorrhage. In a similar study, Yang et al. used radiomics models based on machine learning algorithms and effective computer-aided tools to predict ruptured intracranial aneurysms in 2023. The AdaBoost algorithm was found to be superior in the application of radiomics combined with machine learning algorithm to predict aneurysm ruptures [ 20 ]. Mia et al. In their study published in 2024, they show the high efficiency of the ADASYN_RF algorithm on the cerebral stroke prediction dataset. Also, the AUC channels help to determine which type of categorization is best. Following this procedure, cerebral stroke can be predicted more accurately using ADASYN_RF methods [ 21 ]. Limitations of Study This study was conducted to predict the factors that may affect the development of postoperative hemorrhage in operated brain tumors using the machine learning method. Only one patient underwent reoperation with the diagnosis of developing intracerebral hematoma. However, this study only includes one-year tumor cases from a single center. These factors constitute the limitations of the study and longer-term studies can be conducted to make generalizations. Conclusion In conclusion, the present study demonstrates that machine learning algorithms—particularly Bagging MARS—can effectively predict postoperative hemorrhage risk following cerebral tumor surgery. Key predictors such as platelet count, sodium, and GFR offer actionable insights for clinical management. While promising, further validation with larger, multicenter datasets is necessary before clinical implementation. These predictive tools may support clinicians in early risk identification and tailored intensive care. Declarations Contribution of Authors Detailing the work; project preparation, data collection and writing; Exp.Dr.Yasin Göktürk, Exp.Dr. Şule Göktürk and Assoc.Prof.Dr.S.Kağan Başarslan have contributions. Ph.Dr.Hikmet Kocaman and Ph.Dr. HasanYıldırım did the statistical analysis used the AI programmes fort he study. Assoc.Prof.Dr.S.Kağan Başarslan contributed by operating on the patients. Exp.Dr.Şule Gökürk and Assoc.Prof.Dr.S.Kağan Başarslan are the responsible physicians of Neurosurgery Intensive Care and they personally contributed to the postoperative management. Full approval of the manuscript by all authors should be explicitly stated by including the following statement. Institutional Review Board Statement :This study was conducted with the approval of the Kayseri City Hospital Ethics Committee determined by the institutional or regional responsible committee (Ethics Committee approval number: 346/25.02.2025) and in accordance with the principles stated in the 1975 Helsinki Declaration, revised in 1983. Consent for publication : Consent to participate was obtained from all patients participating in the study and the study was conducted after approval was obtained. Data Availability Statement: We encourage all authors of articles published in BMC journals to share their research data. Funding: None. Conflicts of interest: The authors declare no conflicts of interest. Presentation: none References Robba C, Bertuetti R, Rasulo F, Bertuccio A, Matta B. Coagulation management in patients undergoing neurosurgical procedures. Curr Opin Anaesthesiol. 2017;30(5):527–33. Seifman MA, Lewis PM, Rosenfeld JV, Hwang PYK. Postoperative intracranial haemorrhage: a review. Neurosurg Rev. 2011;34:393–407. Chernov MF, Ivanov PI. Urgent Reoperation for Major Regional Complications After Removal of Intracranial Tumors: Outcome and Prognostic Factors in 100 Consecutive Case. Neurol Med Chir (Tokyo). 2007;47:243–9. Zetterling M, Ronne-Engström E. High intraoperative blood loss may be a risk factor for postoperative hematoma. J Neurosurg Anesthesiol. 2004;16(2):151–5. Nittby HR, Maltese A, Stahl N. Early postoperative haematomas in neurosurgery. Acta Neurochir (Wien). 2016;158:837–46. Nilsson CU, Strandberg K, Engstro¨m M, Reinstrup P. Coagulation during elective neurosurgery with hydroxyethyl starch fluid therapy: an observational study with thromboelastometry, fibrinogen and factor XIII. Perioper Med (Lond). 2016;5:20. Palmer JD, Sparrow OC, Iannotti F. Postoperative hematoma: a 5-year survey and identification of avoidable risk factors. Neurosurgery. 1994;35(6):1061–4. discussion 1064-5. Fan W, Wu Z, Zhao W, Jia L, Li S, Wei W, Chen X. The performance of artificial intelligence in image-based prediction of hematoma enlargement: a systematic review and meta-analysis. Ann Med. 2025;57(1):2515473. Wagner MW, Namdar K, Biswas A, Monah S, Khalvati F, Ertl-Wagner BB. Radiomics, machine learning, and artificial intelligence—what the neuroradiologist needs to know. Neuroradiology. 2021;63(12):1957–67. Xu C, Yang J, Xiong H, Cui X, Zhang Y, Gao M, He L, Fang Q, Han C, Liu W, Wang Y, Zhang J, Yuan Y, Zeng Z, Xu R. Machine learning and multi-omics analysis reveal key regulators of proneural-mesenchymal transition in glioblastoma. Sci Rep. 2025;15(1):19731. Ziai WC, Torbey MT, Kickler TS, Oh S, Bhardwaj A, Wityk RJ. Platelet count and function in spontaneous intracerebral hemorrhage. J Stroke Cerebrovasc Dis 2003 Jul-Aug;12(4):201–6. Qian A, Zheng L, He Z, Zhou J, Tang S, Xing W. Predictive value of hyponatremia for short-term mortality in supratentorial spontaneous intracerebral hemorrhage: a single center study. Front Neurol. 2024;15:1301197. Li Z, Zixiao L, Zhou Q, Gu H, Wang Y, Zhao X. and Chinese Stroke Center Alliance investigators. Effects of estimated glomerular filtration rate on clinical outcomes in patients with intracerebral hemorrhage. BMC Neurol. 2022;22(1):19. Chan K, Mann K, Chan T. The significance of thrombocytopenia in the development of postoperative intracranial haematoma. J Neurosurg. 1989;71:38–41. 10.3171/jns.1989.71.1.0038 . Bullock R, Hannemann CO, Murray L, Teasdale GM. Recurrent hematomas following craniotomy for traumatic intracranial mass. J Neurosurg. 1990;72:9–14. Basali A, Mascha EJ, Kalfas I, Schubert A. Relation between perioperative hypertension and intracranial hemorrhage after craniotomy. Anesthesiology. 2000;93:48–54. Kalfas IH, Little JR. Postoperative hemorrhage: a survey of 4992 intracranial procedures. Neurosurgery. 1988;23:343–7. Gerlach R, Raabe A, Scharrer I, Meixensberger J, Seifert V. Post-operative hematoma after surgery for intracranial meningiomas: causes, avoidable risk factors and clinical outcome. Neurol Res. 2004;26:61–6. Taylor WAS, Thomas NWM, Wellings JA, Bell BA. Timing of postoperative intracranial hematoma development and implications for the best use of neurosurgical intensive care. J Neurosurg. 1995;82:48–50. Yang W, Li W, Wu X, Zhong W, Wang J, Zhou Y, Huang T, Zhou L, Zhou Z. Comparison of Ruptured Intracranial Aneurysms Identification Using Different Machine Learning Algorithms and Radiomics. Diagnostics. 2023;13:2627. Mia R, Khanam S, Mahjabeen A, Hoque Ovy N, Ghimire D, Park M, Ara Begum MI. Sanwar Hosen ASM. Exploring Machine Learning for Predicting Cerebral Stroke: AStudyin Discover. Electronics 2024, 13, 686. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 23 Oct, 2025 Read the published version in BMC Medical Informatics and Decision Making → Version 1 posted Editorial decision: Revision requested 13 Aug, 2025 Reviews received at journal 11 Aug, 2025 Reviews received at journal 06 Aug, 2025 Reviewers agreed at journal 23 Jul, 2025 Reviews received at journal 21 Jul, 2025 Reviews received at journal 20 Jul, 2025 Reviewers agreed at journal 18 Jul, 2025 Reviewers agreed at journal 18 Jul, 2025 Reviewers agreed at journal 17 Jul, 2025 Reviewers invited by journal 17 Jul, 2025 Editor assigned by journal 16 Jul, 2025 Editor invited by journal 15 Jul, 2025 Submission checks completed at journal 14 Jul, 2025 First submitted to journal 14 Jul, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7012786","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":488757189,"identity":"1b28fa83-88e6-48bb-b07a-0813bd295429","order_by":0,"name":"Yasin Göktürk","email":"","orcid":"","institution":"University of Health Sciences, Kayseri City Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yasin","middleName":"","lastName":"Göktürk","suffix":""},{"id":488757190,"identity":"b9ce0c5b-25be-482c-8f43-869c11d1d7ec","order_by":1,"name":"Seyit Kağan Başarslan","email":"","orcid":"","institution":"University of Health Sciences, Kayseri City Hospital","correspondingAuthor":false,"prefix":"","firstName":"Seyit","middleName":"Kağan","lastName":"Başarslan","suffix":""},{"id":488757191,"identity":"b34a0220-16f6-4ff9-9a00-0ea2bd6532dc","order_by":2,"name":"Şule Göktürk","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA7UlEQVRIie2RuwrCMBRAEwp1Sena4usXUvwhS0GXOgnSTaVwu+he/8LJuUXoFHSNOKj4A3UTdDDRTWytm0POFsiBc+9FSKH4S7QJugbr8fLAnKN4NhDSvyl4gmcsI5T7HkWIkkqKhoFJJbMqKe3YnZ4xcKEMILjAnZg16IrUVaFCuRt2MOSEHrb6bgGU2LMsEan7YsVyod58Kqa+N4RCeV+mFisiLLoZMmw3h2ElBXEXNOM5Psu0l9JLShXKTqEtl2zHvmfHm46cpZuWzdKOvDQXpySm5Tt5Pmq1xMac4zUoCftI8uN/hUKhULzxAG0dX+UUmUjbAAAAAElFTkSuQmCC","orcid":"","institution":"University of Health Sciences, Kayseri City Hospital","correspondingAuthor":true,"prefix":"","firstName":"Şule","middleName":"","lastName":"Göktürk","suffix":""},{"id":488757192,"identity":"13c59960-5c10-423a-a204-eec01fb8c028","order_by":3,"name":"Hikmet Kocaman","email":"","orcid":"","institution":"Karamanoğlu Mehmetbey University","correspondingAuthor":false,"prefix":"","firstName":"Hikmet","middleName":"","lastName":"Kocaman","suffix":""},{"id":488757193,"identity":"efe6a99c-2317-4aa8-9d26-6fed64af8d8c","order_by":4,"name":"Hasan Yıldırım","email":"","orcid":"","institution":"Karamanoglu Mehmetbey University","correspondingAuthor":false,"prefix":"","firstName":"Hasan","middleName":"","lastName":"Yıldırım","suffix":""}],"badges":[],"createdAt":"2025-06-30 16:38:03","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7012786/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7012786/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s12911-025-03245-8","type":"published","date":"2025-10-23T16:17:28+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":87345654,"identity":"c01793f8-f192-4243-bb42-0e41e0419637","added_by":"auto","created_at":"2025-07-23 02:16:35","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":347145,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eVariable importance ranking for the top 10 predictors in the Bagging MARS model for brain hemorrhage risk.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7012786/v1/a48773fbced0e56fca40de07.jpg"},{"id":87345653,"identity":"dd59c142-2c39-4cd5-bd73-93f27cde13ff","added_by":"auto","created_at":"2025-07-23 02:16:35","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":177983,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eConfusion matrix illustrating the classification performance of the Bagging MARS model on the test dataset.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7012786/v1/f165f92fac127276b8cc0fd7.jpg"},{"id":87345663,"identity":"7f24d188-f0af-45df-bb0e-79373c22dfdf","added_by":"auto","created_at":"2025-07-23 02:16:35","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":268267,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eReceiver Operating Characteristic (ROC) curve and Area Under the Curve (AUC) for the Bagging MARS model on the test dataset.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7012786/v1/a94509f6e5d3ebe96a773c0a.jpg"},{"id":87345655,"identity":"cdd1c668-f860-4848-85e7-543ff8d21f19","added_by":"auto","created_at":"2025-07-23 02:16:35","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":607242,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eSHAP summary plot illustrating the impact of individual predictors on the Bagging MARS model's output for brain hemorrhage risk.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7012786/v1/a6de70bf134af2f6372b8de0.jpg"},{"id":87345662,"identity":"8cb8956a-5669-4f58-9112-2b48b82716eb","added_by":"auto","created_at":"2025-07-23 02:16:35","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":815213,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003e\u003cstrong\u003e5a.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e LIME explanation detailing local feature importance for a specific 'Positive' prediction by the Bagging MARS model.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u003cstrong\u003e5b.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e LIME explanation detailing local feature importance for a specific 'Negative' actual observation predicted as 'Positive' by the Bagging MARS model.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure5a.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7012786/v1/3ec337b6e4dcac52c21f88e3.jpg"},{"id":94490087,"identity":"8a2e8603-ac85-4f65-9ba3-e70f9a8a9633","added_by":"auto","created_at":"2025-10-27 17:07:26","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3450223,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7012786/v1/5b4407b7-353c-4c55-a732-1fe7acfe2082.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Prediction of postoperative haemorrhage after cerebral tumour surgery using machine learning algorithms","fulltext":[{"header":"Introduction","content":"\u003cp\u003eNeurosurgical surgery means \"\u003cem\u003ecraniotomy\u003c/em\u003e\". It includes basic techniques used to reach the surgical area in intracranial pathologies. It consists of a series of technical methods used in cerebral tumors. Management of bleeding-coagulation is difficult in intracerebral tumor surgery (ICTS) and subsequent intensive care follow-ups. In this struggle, avoiding intracranial bleeding is mandatory and has prognostic value. Perioperative bleeding can be associated with a number of factors including anticoagulant drugs and coagulation status. In addition, it is also related to the characteristic-morphological features of the intracranial mass and its intracerebral location [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eAfter craniotomy performed in ICTS, infections and bleeding in the surgical cavity can be a common and potentially fatal complication in neurosurgical intensive care [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Regular neurological assessment in the intensive care unit or recovery room is extremely important. Unexpected onset or worsening of neurological signs and any deterioration in the patient's state of consciousness should prompt immediate neuroradiological investigation. Urgent computed tomography of the brain is the method of choice if intracranial complications are suspected after brain tumor removal [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Although it is controlled with computed tomography (CT) within the first 24 hours postoperatively [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e], when bleeding occurs, it causes a decrease in the patient's Glasgow coma score and may require reoperation. It is vital to be able to foresee this situation and take precautions.\u003c/p\u003e\u003cp\u003ePerioperative bleeding can be associated with a number of factors, including antihemostatic and anticoagulant medications [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. On the other hand, the risk of venous thromboembolism is also increased after elective and emergency neurosurgery, and a hypercoagulable state has been described in a few patients, particularly after surgery for ICTS [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. It is important to know what changes occur in coagulation in the perioperative period to balance the risk of thrombosis against the risk of bleeding. Routine coagulation tests include activated partial thromboplastin time (aPTT), prothrombin time (PT), platelet count (Plt) and fibrinogen levels. A study conducted to determine the incidence of postoperative hematoma and to identify risk factors for perioperative bleeding disorders during the 5-year period from 1989 to 1993 found that two-thirds of patients had risk factors for perioperative bleeding disorders. Administration of antiplatelet agents (aspirin and nonsteroidal antiinflammatory drugs) was the most frequently associated risk factor. At least 75% of these identified risk factors were potentially preventable or amenable to correction [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. In patients who develop bleeding at the surgical site, blood transfusions and blood component therapy can be life-saving in life-threatening bleeding situations.\u003c/p\u003e\u003cp\u003eIn recent years, machine learning algorithms have gained prominence in healthcare for their ability to detect complex patterns in clinical data and support predictive decision-making. Particularly in neurosurgery, these methods offer promising potential for early identification of high-risk patients, enabling timely interventions and improved outcomes [\u003cspan additionalcitationids=\"CR9\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e–\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. However, their application in predicting postoperative intracerebral hemorrhage remains limited, highlighting the need for targeted studies such as the present research.To our knowledge, there is no study on machine learning that can specifically predict postoperative hemorrhage in patients who have undergone intracerebral tumor surgery. Therefore, the aim of the current study was to predict postoperative bleeding using machine learning algorithms using clinical and laboratory results of patients after brain tumour surgery.\u003c/p\u003e\u003cp\u003eConsent to participate was obtained from all patients included in the study.\u003c/em\u003e\u003c/p\u003e"},{"header":"Material and Method","content":"\u003cp\u003e\u003cb\u003eStudy design\u003c/b\u003e\u003c/p\u003e\u003cp\u003eThis study covers a retrospective examination of patients who underwent surgery with a diagnosis of intracerebral tumor between January 2024 and January 2025, with the approval of the \u003cem\u003eKayseri City Hospital Ethics Committee determined by the institutional or regional responsible committee\u003c/em\u003e (\u003cem\u003eEthics Committee approval number: 346/25.02.2025\u003c/em\u003e) and in accordance with the \u003cem\u003eprinciples stated in the 1975 Helsinki Declaration, revised in 1983\u003c/em\u003e. Intracerebral tumor patients who underwent surgery in a single center were followed up in the Kayseri City Hospital Neurosurgery Clinic unit after surgery. In the postoperative clinical follow-up of the operated 118 patients, age, gender, systemic inflammatory response markers, intensive care unit stay, and biochemical parameters were recorded. Patients with radiologically confirmed hematoma of at least 5 cm³ on CT at 24–48 hours postoperatively were classified as 'Positive' for bleeding.\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eInclusion cirteria\u003c/strong\u003e\u003c/p\u003e\u003cp\u003eThese are all patients who were operated on with the diagnosis of \"intracranial mass\" in the K****C**** H***** Neurosurgery Clinic for 1 year in 2024.\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eExclusion criteria\u003c/strong\u003e\u003c/p\u003e\u003cp\u003ePatients who were followed up in the Postoperative Neurosurgery Intensive Care Unit and had serious systemic diseases that could affect biochemical and hematological parameters, or systemic inflammatory diseases, bleeding diathesis, or using anticoagulants (except those using prophylactic low molecular weight anticoagulants) were excluded from the study.\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eVolume measurement process with computerized tomography\u003c/strong\u003e\u003c/p\u003e\u003cp\u003eWidth-length and height values ​​of the hematoma were measured in the postoperative 1st day brain tomography and its volume (cm\u003csup\u003e3\u003c/sup\u003e) was calculated. Sodium, Calcium, PT, aPTT, Fibrinogen and Platelet values ​​will also be included in the study and were compared with other markers with the blood volume measurement calculated on the first postoperative day.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eMachine Learning Algorithms and Experimental Settings\u003c/b\u003e\u003c/p\u003e\u003cp\u003eDuring the first stage of analysis process, statistical analysis was performed using R (version 4.4.2; R Foundation for Statistical Computing, Vienna, Austria) and IBM SPSS Statistics for Windows (version 26.0; IBM Corp., Armonk, N.Y., USA). A p-value less than 0.05 was considered statistically significant. Descriptive statistics were used to summarize patient characteristics and measurements; continuous variables were presented as mean ± standard deviation (sd), while categorical variables were presented as numbers (n) and percentages (%). For the comparison of baseline characteristics between the bleeding (Positive) and no bleeding (Negative) groups, independent samples t-tests were used for continuous variables such as Age and Length of ICU Stay. Categorical variables including Gender, Progression, Nutrition, Comorbidity, HT (Hypertension), DM (Diabetes Mellitus), and Cancer were analyzed using Pearson's Chi-square test of independence or Fisher's exact test, as appropriate, based on the expected cell counts. For the comparison of various laboratory measurements between the two groups, independent samples t-tests or t-tests with Welch correction (for unequal variances) were employed for variables such as White Blood Cell count (WBC), new WBC count (WBCNew), Neutrophil count, Lymphocyte count, Hemoglobin, Platelet count (PLT), Prothrombin Time (PT), activated Partial Thromboplastin Time (aPTT), International Normalized Ratio (INR), Fibrinogen, Aspartate Aminotransferase (AST), Alanine Aminotransferase (ALT), Blood Urea Nitrogen (BUN), Creatinine, Glomerular Filtration Rate (GFR), Albumin (ALB), Preoperative C-Reactive Protein (PreoperativeCRP), Postoperative C-Reactive Protein (PostoperativeCRP), Sodium (NA), Potassium (K), Chloride (CL), Magnesium (MG), and Calcium (CA). The specific test used for each variable is indicated in the table footnotes.\u003c/p\u003e\u003cp\u003eAs the main scope of this study, a comprehensive machine learning approach is employed to classify individuals based on their risk of brain hemorrhage and to identify key contributing factors. Similar to statistical analysis, all analyses were performed using R statistical software (version 4.4.2). The initial dataset underwent several preprocessing steps. An outlier analysis was conducted using Tukey's method, z-scores, and box plots; fifteen units commonly identified as outliers by these methods were excluded from the study. The primary outcome variable, \"Bleeding Situation,\" was derived from the original bleeding indicator, categorizing individuals as \"Negative\" (no hemorrhage) or \"Positive\" (hemorrhage). The dataset was then partitioned into training (75%) and testing (25%) sets using stratified sampling based on the \"Bleeding Situation\" variable to ensure proportional representation in both sets. A comprehensive preprocessing pipeline was constructed using the recipes package. This involved median imputation for all numeric variables and mode imputation for nominal variables. Subsequently, zero-variance variables were removed, and numeric variables exhibiting high correlation (threshold of 0.85) were eliminated. Numeric variables were then centered and scaled. Finally, all nominal variables were converted into numerical format using one-hot encoding.\u003c/p\u003e\u003cp\u003e\u003cb\u003eMachine learning algorithms\u003c/b\u003e\u003c/p\u003e\u003cp\u003eModel development and hyperparameter optimization were conducted within the tidymodels framework. A k-fold cross-validation strategy, specifically 10-fold cross-validation repeated 5 times, was applied to the training data, stratified by the \"Bleeding Situation\" variable. This resampling technique was used for robust hyperparameter tuning across a predefined grid of sixty hyperparameter combinations for each model. The primary metric for selecting the optimal hyperparameter set during the tuning process was the Area Under the Receiver Operating Characteristic Curve (ROC AUC). A wide array of classification algorithms was implemented, including Logistic Regression, K-Nearest Neighbors (KNN), Naive Bayes, Classification and Regression Trees (CART), Support Vector Machines (SVM) with linear, polynomial, and radial basis function kernels, Multilayer Perceptron (NNET), C5.0 decision trees and rule-based models, Multivariate Adaptive Regression Splines (MARS), Random Forests, Bayesian Additive Regression Trees (BART), Partial Least Squares (PLS), RuleFit, bagged versions of CART, MARS, MLP, and C5.0, and boosted models including XGBoost, boosted C5.0, and LightGBM. Regularized logistic regression models, namely Ridge, Lasso, and Elastic Net, were also evaluated. Workflows encapsulating the preprocessing recipe and model specifications were created for each algorithm, and parallel processing (doParallel package) was utilized to expedite the tuning process.\u003c/p\u003e\u003cp\u003eModel performance was evaluated using a focused set of metrics, including ROC AUC, Brier score, accuracy, and F-measure, collected using the yardstick package. To identify the most influential variables, variable importance plots were generated. Furthermore, model explainability was enhanced through the creation of SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) plots. ROC curves were also plotted to visualize the trade-off between true positive rate and false positive rate. After identifying the optimal hyperparameters for each model based on the cross-validation results on the training set (primarily maximizing ROC AUC), the final models were fitted to the entire training dataset and subsequently evaluated on the unseen test set. The performance metrics on the test set were then compiled to compare the generalization ability of the models. Visualizations such as confusion matrices were used to assess classification performance on the test data. Key R packages pivotal to this analysis included tidyverse for general data manipulation, tidymodels, caret, ranger, glmnet, kernlab, xgboost, C50, nnet, earth, kknn, rpart, mixOmics, and lightgbm. The results, including training performance, test performance, and optimal hyperparameters for each model, were systematically organized and saved.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eThe demographic and clinical characteristics of the patients are presented in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. No statistically significant difference was found between the groups in terms of age, gender, hypertension (HT), diabetes mellitus (DM), and dietary method (p\u0026thinsp;\u0026gt;\u0026thinsp;0.05). In other parameters (duration of intensive care unit stay, progression, nutrition, comorbidity, presence of cancer), a statistically significant difference was observed between the groups (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e shows the comparison of laboratory parameters between individuals with and without cerebral haemorrhage. There was a statistically significant difference between the groups in parameters such as lymphocytes, PLT, BUN, ALB, postoperative CRP, Na, K, Cl, Mg (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05), while there was no significant difference in other parameters (p\u0026thinsp;\u0026gt;\u0026thinsp;0.05) (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eThe summary characteristics of patients considered in the study\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e\u003cp\u003eBleeding\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003ep\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003ctr\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eYes (Positive)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNo (Negative)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eAge, mean\u0026thinsp;\u0026plusmn;\u0026thinsp;sd\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e60.26\u0026thinsp;\u0026plusmn;\u0026thinsp;17.65\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e56.81\u0026thinsp;\u0026plusmn;\u0026thinsp;15.97\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.289\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eLength of ICU Stay, mean\u0026thinsp;\u0026plusmn;\u0026thinsp;sd\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e10.23\u0026thinsp;\u0026plusmn;\u0026thinsp;10.58\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e5.1\u0026thinsp;\u0026plusmn;\u0026thinsp;6.87\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cb\u003e0.002\u003c/b\u003e\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eGender, n(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e23 (58.97%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e50 (63.29%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e0.650\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFemale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e16 (41.03%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e29 (36.71%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eProgression, n(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDischarged\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e31 (79.49%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e77 (97.47%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e0.003\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eExitus\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8 (20.51%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e2 (2.53%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eNutrition, n(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEnteral\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e36 (92.31%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e78 (98.73%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e0.105\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTPN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3 (7.69%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1 (1.27%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eComorbidity, n(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e26 (66.67%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e35 (44.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e\u003cb\u003e0.022\u003c/b\u003e\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e13 (33.33%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e44 (55.7%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eHT, n(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7 (17.95%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e12 (15.19%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e0.701\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e32 (82.05%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e67 (84.81%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eDM, n(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7 (17.95%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e13 (16.88%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e0.886\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e32 (82.05%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e64 (83.12%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eCancer, n(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8 (20.51%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e5 (6.33%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e\u003cb\u003e0.029\u003c/b\u003e\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e31 (79.49%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e74 (93.67%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e\u003cp\u003e\u003csup\u003ea\u003c/sup\u003e: t-test, \u003csup\u003eb\u003c/sup\u003e: Fisher exact test, \u003csup\u003ec\u003c/sup\u003e:Chi-square test of independence\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eThe comparison of various measurements based on bleeding situations\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003eBleeding\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003ep\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eVariable, mean\u0026thinsp;\u0026plusmn;\u0026thinsp;sd\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eYes (Positive)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eNo (Negative)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWBC\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e14079.49\u0026thinsp;\u0026plusmn;\u0026thinsp;6228.17\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e13559.37\u0026thinsp;\u0026plusmn;\u0026thinsp;4908.82\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.650\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWBCNew\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e13.74\u0026thinsp;\u0026plusmn;\u0026thinsp;5.91\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e13.58\u0026thinsp;\u0026plusmn;\u0026thinsp;4.88\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.891\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNeutrophil\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e11.83\u0026thinsp;\u0026plusmn;\u0026thinsp;5.6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e11.59\u0026thinsp;\u0026plusmn;\u0026thinsp;4.7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.819\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLymphocyte\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.93\u0026thinsp;\u0026plusmn;\u0026thinsp;0.56\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.19\u0026thinsp;\u0026plusmn;\u0026thinsp;0.73\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e0.033\u003c/b\u003e\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHemoglobin\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e11.8\u0026thinsp;\u0026plusmn;\u0026thinsp;1.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e12.38\u0026thinsp;\u0026plusmn;\u0026thinsp;1.73\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.125\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePLT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e209.18\u0026thinsp;\u0026plusmn;\u0026thinsp;70.45\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e234.97\u0026thinsp;\u0026plusmn;\u0026thinsp;68.86\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e0.025\u003c/b\u003e\u003csup\u003ee\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e12.74\u0026thinsp;\u0026plusmn;\u0026thinsp;1.33\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e12.33\u0026thinsp;\u0026plusmn;\u0026thinsp;1.25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.114\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eaPTT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e22.2\u0026thinsp;\u0026plusmn;\u0026thinsp;4.09\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e21.56\u0026thinsp;\u0026plusmn;\u0026thinsp;3.28\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.400\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eINR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.11\u0026thinsp;\u0026plusmn;\u0026thinsp;0.12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.06\u0026thinsp;\u0026plusmn;\u0026thinsp;0.13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.068\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFibrinogen\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3817.65\u0026thinsp;\u0026plusmn;\u0026thinsp;1260.86\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3313.88\u0026thinsp;\u0026plusmn;\u0026thinsp;1210.45\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.185\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAST\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e29.92\u0026thinsp;\u0026plusmn;\u0026thinsp;44.85\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e20.05\u0026thinsp;\u0026plusmn;\u0026thinsp;19.65\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.195\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eALT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e22.1\u0026thinsp;\u0026plusmn;\u0026thinsp;18.32\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e25.33\u0026thinsp;\u0026plusmn;\u0026thinsp;28.92\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.463\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBUN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e21.46\u0026thinsp;\u0026plusmn;\u0026thinsp;10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e16.88\u0026thinsp;\u0026plusmn;\u0026thinsp;6.84\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e0.013\u003c/b\u003e\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCreatinine\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.74\u0026thinsp;\u0026plusmn;\u0026thinsp;0.32\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.71\u0026thinsp;\u0026plusmn;\u0026thinsp;0.18\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.606\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGFR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e94.33\u0026thinsp;\u0026plusmn;\u0026thinsp;20.89\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e99.16\u0026thinsp;\u0026plusmn;\u0026thinsp;18.95\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.227\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eALB\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e32.79\u0026thinsp;\u0026plusmn;\u0026thinsp;4.47\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e34.99\u0026thinsp;\u0026plusmn;\u0026thinsp;3.8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e0.011\u003c/b\u003e\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePreoperativeCRP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e12.01\u0026thinsp;\u0026plusmn;\u0026thinsp;25.64\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5.35\u0026thinsp;\u0026plusmn;\u0026thinsp;8.89\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.123\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePostoperativeCRP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e64.08\u0026thinsp;\u0026plusmn;\u0026thinsp;72.48\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e33.82\u0026thinsp;\u0026plusmn;\u0026thinsp;38.29\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e0.018\u003c/b\u003e\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e140.79\u0026thinsp;\u0026plusmn;\u0026thinsp;4.47\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e138.33\u0026thinsp;\u0026plusmn;\u0026thinsp;3.47\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e0.004\u003c/b\u003e\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eK\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4.07\u0026thinsp;\u0026plusmn;\u0026thinsp;0.53\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4.26\u0026thinsp;\u0026plusmn;\u0026thinsp;0.42\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e0.039\u003c/b\u003e\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCL\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e104.44\u0026thinsp;\u0026plusmn;\u0026thinsp;4.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e102.73\u0026thinsp;\u0026plusmn;\u0026thinsp;3.98\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e0.033\u003c/b\u003e\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2.16\u0026thinsp;\u0026plusmn;\u0026thinsp;0.28\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.03\u0026thinsp;\u0026plusmn;\u0026thinsp;0.22\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e0.018\u003c/b\u003e\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e8.27\u0026thinsp;\u0026plusmn;\u0026thinsp;0.64\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8.35\u0026thinsp;\u0026plusmn;\u0026thinsp;0.57\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.495\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBleedindWidth\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e22.38\u0026thinsp;\u0026plusmn;\u0026thinsp;11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0\u0026thinsp;\u0026plusmn;\u0026thinsp;0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBleedingLength\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e17.4\u0026thinsp;\u0026plusmn;\u0026thinsp;10.88\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0\u0026thinsp;\u0026plusmn;\u0026thinsp;0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBleedingHeight\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e18.72\u0026thinsp;\u0026plusmn;\u0026thinsp;8.7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0\u0026thinsp;\u0026plusmn;\u0026thinsp;0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBleedingVolume\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e12.41\u0026thinsp;\u0026plusmn;\u0026thinsp;25.15\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0\u0026thinsp;\u0026plusmn;\u0026thinsp;0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"4\" nameend=\"c4\" namest=\"c1\"\u003e\u003cp\u003e\u003csup\u003ea\u003c/sup\u003e: t-test, \u003csup\u003ed\u003c/sup\u003e: t-test with Welch correction\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThis study undertook a comprehensive evaluation of multiple machine learning algorithms to accurately classify brain hemorrhage risk and to delineate the most significant predictive factors. The performance characteristics of these models, assessed on both the training dataset (via 5-repeated 10-fold cross-validation) and an independent test dataset, are systematically presented.\u003c/p\u003e\u003cp\u003e\u003cb\u003eComparative Model Performance\u003c/b\u003e\u003c/p\u003e\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e provides a detailed comparison of all implemented machine learning models based on four key performance metrics: Accuracy, Brier Score (lower indicates better calibration), F-Score, and Area Under the Receiver Operating Characteristic Curve (AUC). On the crucial independent test dataset, several algorithms demonstrated notable predictive efficacy. \u003cb\u003eThe Bagging algorithm utilizing a Multivariate Adaptive Regression Splines (MARS) base learner (Bagging MARS)\u003c/b\u003e emerged as a particularly strong performer, achieving a test AUC of 0.8693, an accuracy of 0.8077 (21 out of 26 test cases correctly classified), a low Brier score of 0.1580, and a high F-Score of 0.8649. Other models exhibiting robust test performance included the \u003cb\u003eBoosting algorithm with a C5.0 base learner (Boosting C5.0)\u003c/b\u003e, which recorded a test AUC of 0.8235 and an accuracy similar to Bagging MARS at 0.8073. \u003cb\u003eThe Support Vector Machine (SVM) with a Radial Basis Function (RBF) kernel\u003c/b\u003e also showed excellent discriminatory power with a test AUC of 0.8562, albeit with a slightly lower accuracy of 0.7538. Similarly, \u003cb\u003eRandom Forests\u003c/b\u003e yielded a high test AUC of 0.8431, though its accuracy was 0.7308. Given its leading AUC on the test set and a strong, balanced performance across accuracy and F-Score, the Bagging MARS model was selected for subsequent in-depth interpretability analyses.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003ePerformance Metrics (Accuracy, Brier Score, F-Score, AUC) of Machine Learning Models on Training and Test Datasets for Brain Hemorrhage Classification.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSplit\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eBrier Score\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eF Score\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eAUC\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eBagging (CART learner)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7344\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2290\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7454\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.6828\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.6923\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.1944\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7712\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eBagging (MARS learner)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7878\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2275\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cb\u003e0.8790\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7916\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003e0.8077\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.1580\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cb\u003e0.8649\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cb\u003e0.8693\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eBagging (MLP learner)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7847\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2068\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7518\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7253\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7692\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.1869\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8333\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7582\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eBayesian Additive Trees\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7741\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2094\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8023\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7024\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.6538\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2020\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7907\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7353\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eBoosting (C5.0 learner)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003e0.8341\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2467\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7293\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7787\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.8073\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.1582\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8571\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.8235\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eBoosting (LightGBM)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7646\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2232\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7979\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.6453\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7538\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2264\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7907\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7686\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eBoosting (XGBoost)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7526\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2173\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7839\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.6033\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7923\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.1931\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7712\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eC5.0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.8025\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2974\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.6910\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.6991\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.8077\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e0.1576\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8485\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7876\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eC5.0 - Rules\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7407\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2263\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7579\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.6933\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7692\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.1823\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8235\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7451\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eCART\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.6832\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.3975\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.6848\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.6592\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.6923\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2800\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7143\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7908\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eElastic Net\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.8189\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2588\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7777\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cb\u003e0.8087\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7692\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2359\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8235\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7124\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eKNN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7667\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.3333\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7235\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.6547\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.6923\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.3077\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7500\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.6863\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eLasso\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.8185\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2582\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7789\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.8073\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7692\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2365\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8235\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7059\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eLogistic Regression\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.8156\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2838\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7699\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7291\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7692\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2308\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8235\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7157\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eMARS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7443\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2361\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7723\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.6432\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7692\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.1674\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8333\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7418\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eMultilayer Perceptron\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7854\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2153\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7457\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7413\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7692\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.1891\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8235\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7680\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eNaive Bayes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7780\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.3071\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7947\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.6367\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7692\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2083\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8421\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7889\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003ePartial Least Squares\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.8010\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.1947\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7645\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7262\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7308\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.1908\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.6536\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eRandom Forests\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7912\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2110\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8080\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.6464\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7308\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.1741\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8205\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.8431\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eRidge\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7741\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2101\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7550\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7300\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7692\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.1841\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8235\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.6863\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eRuleFit\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7793\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2855\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7431\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7869\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7310\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2650\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7880\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7120\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eSVM (Linear kernel)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.8076\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e0.1778\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7643\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7867\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7692\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.1604\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8235\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7124\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eSVM (Polynomial kernel)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7646\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2128\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7979\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7353\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7538\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.1792\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7907\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.8235\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eSVM (Radial kernel)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7646\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2232\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7979\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7351\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.7538\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.2264\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7907\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.8562\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cbr\u003e\u003cp\u003e\u003cstrong\u003e\u003cem\u003eIn-depth Analysis of the Bagging MARS Model\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA detailed examination of the selected Bagging MARS model was conducted to understand its predictive mechanisms and identify the primary determinants of brain hemorrhage risk.\u003c/p\u003e\n\u003cp\u003e\u0026middot; \u003cstrong\u003eVariable Importance:\u003c/strong\u003e \u003cstrong\u003e\u003cem\u003eFigure 1\u003c/em\u003e\u003c/strong\u003e illustrates the hierarchy of predictor importance as determined by the Bagging MARS model. Platelet count (PLT) was unequivocally identified as the most influential variable, with a relative importance score of 100.0. Following PLT, Sodium (Na) and Glomerular Filtration Rate (GFR) were found to be substantial contributors, with importance scores of 59.0 and 40.9, respectively. Other significant predictors included Postoperative C-Reactive Protein (CRP) (importance: 31.3) and Hemoglobin (importance: 30.8), underscoring their roles in the model\u0026apos;s risk assessment.\u003c/p\u003e\n\u003cp\u003e\u0026middot; \u003cstrong\u003eClassification Accuracy and Discriminatory Power on Test Data:\u003c/strong\u003e The confusion matrix (\u003cstrong\u003e\u003cem\u003eFigure 2\u003c/em\u003e\u003c/strong\u003e) for the Bagging MARS model on the test dataset (N=26) provides a granular view of its classification performance. The model correctly identified 16 out of 17 individuals without brain hemorrhage (True Negatives), resulting in a high specificity of 94.1%. For individuals who did experience brain hemorrhage, the model correctly classified 5 out of 9 cases (True Positives), yielding a sensitivity (recall) of 55.6%. The model made only 1 false positive prediction and 4 false negative predictions. The overall discriminatory capability of the model, as graphically represented by the Receiver Operating Characteristic (ROC) curve (\u003cstrong\u003e\u003cem\u003eFigure 3\u003c/em\u003e\u003c/strong\u003e), culminated in an Area Under the Curve (AUC) of 0.8693. This high AUC value signifies the model\u0026apos;s robust ability to distinguish between patients at high and low risk of brain hemorrhage.\u003c/p\u003e\n\u003cp\u003e\u0026middot; \u003cstrong\u003eGlobal Model Interpretability (SHAP Analysis):\u003c/strong\u003e The SHAP (SHapley Additive exPlanations) summary plot (\u003cstrong\u003e\u003cem\u003eFigure 4\u003c/em\u003e\u003c/strong\u003e) provides global insights into how feature values influence the model\u0026apos;s prediction (on a log-odds scale). For Platelet count (PLT), higher values (represented by redder dots) are generally associated with an increased positive impact on the prediction of brain hemorrhage (dots shift to the right of the zero SHAP value line). Conversely, for a feature like Alanine Aminotransferase (ALT), higher values tend to have a negative impact on the log-odds of predicting hemorrhage (redder dots shift to the left). This visualization effectively captures the magnitude and direction of each predictor\u0026apos;s influence across the dataset, for example, GFR shows that lower values (bluer dots) tend to increase the positive SHAP value, suggesting lower GFR is a risk factor.\u003c/p\u003e\n\u003cp\u003e\u0026middot; \u003cstrong\u003eLocal Model Interpretability (LIME Analysis):\u003c/strong\u003e LIME (Local Interpretable Model-agnostic Explanations) plots were employed to dissect individual predictions:\u003c/p\u003e\n\u003cp\u003eo \u003cstrong\u003e\u003cem\u003eFigure 5a\u003c/em\u003e\u003c/strong\u003e provides an explanation for a specific correct \u0026apos;Positive\u0026apos; prediction (Test Set Index: 3; Actual: \u0026apos;Positive\u0026apos;, Predicted: \u0026apos;Positive\u0026apos; with a probability of 0.64). For this instance, higher Age (contributing approximately +0.115 to the prediction weight), Na (+0.062), and BUN (+0.046) were the primary drivers pushing the prediction towards a positive hemorrhage outcome.\u003c/p\u003e\n\u003cp\u003eo \u003cstrong\u003e\u003cem\u003eFigure 5b\u003c/em\u003e\u003c/strong\u003e details a misclassification (Test Set Index: 8; Actual: \u0026apos;Negative\u0026apos;, Predicted: \u0026apos;Positive\u0026apos; with a probability of 0.34). Here, features such as Cl (+0.032), AST (+0.026), and Hemoglobin (+0.026) incorrectly influenced the model to predict a positive outcome. Notably, a lower Age value for this patient provided a strong counter-signal (contribution: -0.115), but was overridden by other factors. These local explanations are critical for understanding model behavior on a case-by-case basis and for identifying potential reasons for specific prediction outcomes.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003ePostoperative intracerebral hemorrhage remains a significant cause of morbidity and mortality following intracranial tumor surgery, despite advancements in surgical and perioperative care. Identifying patients at high risk for postoperative bleeding is crucial for optimizing intensive care management and reducing adverse outcomes. In this context, the present study demonstrates that machine learning algorithms\u0026mdash;particularly the Bagging MARS model\u0026mdash;can effectively predict hemorrhage risk using routinely collected clinical and laboratory data. The strong performance metrics of this model, including high accuracy and AUC values, underscore its potential as a clinical decision support tool for individualized risk stratification. In addition, Platelet count (PLT) was identified as the most influential variable with a relative importance score of 100.0. Following PLT, Sodium (Na) and Glomerular Filtration Rate (GFR) were found to be valuable contributors with importance scores of 59.0 and 40.9 respectively. Ziai et al stated in their study that platelet dysfunction is common in patients with spontaneous intracerebral hemorrhage and that low platelet count and platelet dysfunction are effective factors in the expansion of hematoma vol\u0026uuml;me [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eHyponatremia is a common electrolyte disorder in patients with neurological diseases; however, its predictive role for outcome in patients with supratentorial spontaneous intracerebral hemorrhage is controversial. In their study conducted in 2024, Qian et al. reported that hyponatremia occurring in the first 7 days after hemorrhage in patients with supratentorial spontaneous intracerebral hemorrhage was an independent predictor of 90-day mortality and adverse outcomes [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. They reiterated the need for meticulous electrolyte monitoring in patients treated surgically. In our study, low sodium levels were also associated with poor prognosis. It is the second important parameter affecting the occurrence of postoperative hemorrhage. The impact of chronic kidney disease (CKD) on the severity and prognosis of spontaneous intracerebral hemorrhage (ICH) has rarely been studied. Li et al. found that lower baseline eGFR was associated with an increased risk of in-hospital mortality, nonroutine discharge, hemorrhagic stroke severity, and in-hospital complications such as pneumonia, hydrocephalus, and hematoma evacuation in patients with acute intracerebral hemorrhage [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eChan et al retrospectively reviewed a group of 35 patients who underwent cranial surgery and demonstrated perioperative thrombocytopenia (platelet count less than 150,000/microliter). Fourteen of 35 patients (40%) developed postoperative intracranial hematomas requiring reoperation, and seven (20%) died within 2 weeks after surgery. Analysis revealed that a perioperative platelet count of less than 100,000/microliter was associated with a higher risk of postoperative hematoma formation [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e].\u003c/p\u003e\u003cp\u003ePostoperative bleeding rates after cranial surgery vary greatly in the literature, ranging from 0.77\u0026ndash;50% [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan additionalcitationids=\"CR16\" citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. However, the definition of postoperative bleeding varies. Series that include patients with radiographic evidence of bleeding have reported rates ranging from 10.8\u0026ndash;50% [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. In fact, the location of postoperative hematoma depends largely on preoperative pathology. \u003cem\u003eIntraparenchymal hematomas\u003c/em\u003e have been frequently observed as locations after cranial operations. In routine intensive care follow-up, blood coagulation cascades in particular are routinely examined. In our study, Plt values, Sodium and GFR values ​​stood out as the three most important predictive parameters for postoperative bleeding in \u003cem\u003eMachine Learning\u003c/em\u003e programs. We hope that MARS Bagging and the first three important parametric laboratory values ​​(Plt, Sodium and GFR) will contribute to making important decisions that will change patient prognosis.\u003c/p\u003e\u003cp\u003eGerlach et al retrospectively analyzed the data of 296 patients with and without hematoma who underwent surgery for meningioma to determine the risk factors associated with postoperative hematoma [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Meningioma surgery carries a higher risk for postoperative hematoma in the elderly. Thrombocytopenia and other hemostatic abnormalities are frequently associated with postoperative bleeding after meningioma surgery. Expansion of coagulation tests and specific replacement therapy may prevent hematoma formation and improve patient outcomes. Literature suggests that most hematomas are discovered within three days after the initial operation [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. In the planning of our study, we focused on the patients' brain CT scans and laboratory values ​​taken at the 48th postoperative hour. Kalfas reported that 35% of ICH cases occurred within 12 hours [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e], while Taylor reported that 44 out of 50 of 2,305 patients developed a hematoma within six hours after surgery [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. Of course, it should not be forgotten that longer periods of observation in the intensive care unit may be required after intracerebral tumor surgery.\u003c/p\u003e\u003cp\u003eVarious risk factors for hematoma development have been analyzed, and hypertension is usually the most important comorbidity in most studies [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. However, in our study, no statistically significant difference was found between the groups in terms of age, gender, hypertension (HT), diabetes mellitus (DM) and diet method. A statistically significant difference was observed between the groups in other parameters (length of stay in the intensive care unit, progression, nutrition, comorbidity, presence of cancer). Other risk factors include use of nonsteroidal anti-inflammatory drugs, use of anticoagulants, coagulopathies, thrombocytopenia, factor XIII deficiencies, and decreased fibrinogen levels.\u003c/p\u003e\u003cp\u003eOur study undertook a comprehensive evaluation of multiple machine learning algorithms to accurately classify the risk of postoperative cerebral hemorrhage and identify the most important predictive factors. Of all the machine learning models implemented based on four key performance metrics, the \u003cem\u003eBoosting Algorithm\u003c/em\u003e (Boosting C5.0) with C5.0-based learner recorded a test AUC of 0.8235 and an accuracy similar to Bagging MARS of 0.8073, demonstrating robust test performance. Similar to Bagging MARS, \u003cem\u003eRandom Forests\u003c/em\u003e produced a high test AUC of 0.8431, but its accuracy is 0.7308. A detailed review of the selected Bagging MARS model was performed to understand its predictive mechanisms and the primary determinants of cerebral hemorrhage risk were identified. In the Bagging MARS model; Platelet count (PLT) was determined as the undisputed most influential variable with a relative importance score of 100.0. Following PLT, Sodium (Na) and Glomerular Filtration Rate (GFR) were found to be significant contributors. Other significant predictors included Postoperative C-Reactive Protein (CRP) (importance: 31.3) and Hemoglobin (importance: 30.8), highlighting their role in risk assessment by the model. The model demonstrated its robust ability to distinguish between patients at high and low risk of brain hemorrhage.\u003c/p\u003e\u003cp\u003eIn a similar study, Yang et al. used radiomics models based on machine learning algorithms and effective computer-aided tools to predict ruptured intracranial aneurysms in 2023. The AdaBoost algorithm was found to be superior in the application of radiomics combined with machine learning algorithm to predict aneurysm ruptures [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. Mia et al. In their study published in 2024, they show the high efficiency of the ADASYN_RF algorithm on the cerebral stroke prediction dataset. Also, the AUC channels help to determine which type of categorization is best. Following this procedure, cerebral stroke can be predicted more accurately using ADASYN_RF methods [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e].\u003c/p\u003e\u003cp\u003e\u003cb\u003eLimitations of Study\u003c/b\u003e\u003c/p\u003e\u003cp\u003eThis study was conducted to predict the factors that may affect the development of postoperative hemorrhage in operated brain tumors using the machine learning method. Only one patient underwent reoperation with the diagnosis of developing intracerebral hematoma. However, this study only includes one-year tumor cases from a single center. These factors constitute the limitations of the study and longer-term studies can be conducted to make generalizations.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn conclusion, the present study demonstrates that machine learning algorithms\u0026mdash;particularly Bagging MARS\u0026mdash;can effectively predict postoperative hemorrhage risk following cerebral tumor surgery. Key predictors such as platelet count, sodium, and GFR offer actionable insights for clinical management. While promising, further validation with larger, multicenter datasets is necessary before clinical implementation. These predictive tools may support clinicians in early risk identification and tailored intensive care.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003e\u003cem\u003eContribution of Authors\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eDetailing the work; project preparation, data collection and writing; Exp.Dr.Yasin G\u0026ouml;kt\u0026uuml;rk, Exp.Dr. Şule G\u0026ouml;kt\u0026uuml;rk and Assoc.Prof.Dr.S.Kağan Başarslan have contributions. Ph.Dr.Hikmet Kocaman and Ph.Dr.\u003c/em\u003e\u003cem\u003eHasanYıldırım\u003c/em\u003e\u003cem\u003e\u0026nbsp;did the statistical analysis used the AI programmes fort he study. Assoc.Prof.Dr.S.Kağan Başarslan contributed by operating on the patients. Exp.Dr.Şule G\u0026ouml;k\u0026uuml;rk and Assoc.Prof.Dr.S.Kağan Başarslan are the responsible physicians of Neurosurgery Intensive Care and they personally contributed to the postoperative management. Full approval of the manuscript by all authors should be explicitly stated by including the following statement.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eInstitutional Review Board Statement\u003c/em\u003e\u003c/strong\u003e\u003cem\u003e:This study was conducted\u0026nbsp;\u003c/em\u003e\u003cem\u003ewith the approval of the Kayseri City Hospital Ethics Committee determined by the institutional or regional responsible committee (Ethics Committee approval number: 346/25.02.2025) and in accordance with the principles stated in the 1975 Helsinki Declaration, revised in 1983. \u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eConsent for publication\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e\u003cem\u003e:\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003cem\u003eConsent to participate was obtained from all patients participating in the study and the study was conducted after approval was obtained.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eData Availability Statement:\u003c/em\u003e\u003c/strong\u003e\u003cem\u003e\u0026nbsp;We encourage all authors of articles published in BMC journals to share their research data.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eFunding:\u003c/em\u003e\u003c/strong\u003e\u003cem\u003e\u0026nbsp;None.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eConflicts of interest:\u003c/em\u003e\u0026nbsp;\u003c/strong\u003e\u003cem\u003eThe authors declare no conflicts of interest.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003ePresentation:\u003c/em\u003e\u003c/strong\u003e\u003cem\u003e\u0026nbsp;none\u003c/em\u003e\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eRobba C, Bertuetti R, Rasulo F, Bertuccio A, Matta B. Coagulation management in patients undergoing neurosurgical procedures. Curr Opin Anaesthesiol. 2017;30(5):527\u0026ndash;33.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSeifman MA, Lewis PM, Rosenfeld JV, Hwang PYK. Postoperative intracranial haemorrhage: a review. Neurosurg Rev. 2011;34:393\u0026ndash;407.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChernov MF, Ivanov PI. Urgent Reoperation for Major Regional Complications After Removal of Intracranial Tumors: Outcome and Prognostic Factors in 100 Consecutive Case. Neurol Med Chir (Tokyo). 2007;47:243\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZetterling M, Ronne-Engstr\u0026ouml;m E. High intraoperative blood loss may be a risk factor for postoperative hematoma. J Neurosurg Anesthesiol. 2004;16(2):151\u0026ndash;5.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNittby HR, Maltese A, Stahl N. Early postoperative haematomas in neurosurgery. Acta Neurochir (Wien). 2016;158:837\u0026ndash;46.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNilsson CU, Strandberg K, Engstro\u0026uml;m M, Reinstrup P. Coagulation during elective neurosurgery with hydroxyethyl starch fluid therapy: an observational study with thromboelastometry, fibrinogen and factor XIII. Perioper Med (Lond). 2016;5:20.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePalmer JD, Sparrow OC, Iannotti F. Postoperative hematoma: a 5-year survey and identification of avoidable risk factors. Neurosurgery. 1994;35(6):1061\u0026ndash;4. discussion 1064-5.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFan W, Wu Z, Zhao W, Jia L, Li S, Wei W, Chen X. The performance of artificial intelligence in image-based prediction of hematoma enlargement: a systematic review and meta-analysis. Ann Med. 2025;57(1):2515473.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWagner MW, Namdar K, Biswas A, Monah S, Khalvati F, Ertl-Wagner BB. Radiomics, machine learning, and artificial intelligence\u0026mdash;what the neuroradiologist needs to know. Neuroradiology. 2021;63(12):1957\u0026ndash;67.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eXu C, Yang J, Xiong H, Cui X, Zhang Y, Gao M, He L, Fang Q, Han C, Liu W, Wang Y, Zhang J, Yuan Y, Zeng Z, Xu R. Machine learning and multi-omics analysis reveal key regulators of proneural-mesenchymal transition in glioblastoma. Sci Rep. 2025;15(1):19731.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZiai WC, Torbey MT, Kickler TS, Oh S, Bhardwaj A, Wityk RJ. Platelet count and function in spontaneous intracerebral hemorrhage. J Stroke Cerebrovasc Dis 2003 Jul-Aug;12(4):201\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eQian A, Zheng L, He Z, Zhou J, Tang S, Xing W. Predictive value of hyponatremia for short-term mortality in supratentorial spontaneous intracerebral hemorrhage: a single center study. Front Neurol. 2024;15:1301197.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi Z, Zixiao L, Zhou Q, Gu H, Wang Y, Zhao X. and Chinese Stroke Center Alliance investigators. Effects of estimated glomerular filtration rate on clinical outcomes in patients with intracerebral hemorrhage. BMC Neurol. 2022;22(1):19.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChan K, Mann K, Chan T. The significance of thrombocytopenia in the development of postoperative intracranial haematoma. J Neurosurg. 1989;71:38\u0026ndash;41. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3171/jns.1989.71.1.0038\u003c/span\u003e\u003cspan address=\"10.3171/jns.1989.71.1.0038\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBullock R, Hannemann CO, Murray L, Teasdale GM. Recurrent hematomas following craniotomy for traumatic intracranial mass. J Neurosurg. 1990;72:9\u0026ndash;14.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBasali A, Mascha EJ, Kalfas I, Schubert A. Relation between perioperative hypertension and intracranial hemorrhage after craniotomy. Anesthesiology. 2000;93:48\u0026ndash;54.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKalfas IH, Little JR. Postoperative hemorrhage: a survey of 4992 intracranial procedures. Neurosurgery. 1988;23:343\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGerlach R, Raabe A, Scharrer I, Meixensberger J, Seifert V. Post-operative hematoma after surgery for intracranial meningiomas: causes, avoidable risk factors and clinical outcome. Neurol Res. 2004;26:61\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTaylor WAS, Thomas NWM, Wellings JA, Bell BA. Timing of postoperative intracranial hematoma development and implications for the best use of neurosurgical intensive care. J Neurosurg. 1995;82:48\u0026ndash;50.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang W, Li W, Wu X, Zhong W, Wang J, Zhou Y, Huang T, Zhou L, Zhou Z. Comparison of Ruptured Intracranial Aneurysms Identification Using Different Machine Learning Algorithms and Radiomics. Diagnostics. 2023;13:2627.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMia R, Khanam S, Mahjabeen A, Hoque Ovy N, Ghimire D, Park M, Ara Begum MI. Sanwar Hosen ASM. Exploring Machine Learning for Predicting Cerebral Stroke: AStudyin Discover. Electronics 2024, 13, 686.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-informatics-and-decision-making","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"midm","sideBox":"Learn more about [BMC Medical Informatics and Decision Making](http://bmcmedinformdecismak.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/midm/default.aspx","title":"BMC Medical Informatics and Decision Making","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Brain tumor, Machine learning, Hemorrhage, Risk prediction, Artificial intelligence","lastPublishedDoi":"10.21203/rs.3.rs-7012786/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7012786/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e\u003cp\u003e\u003cem\u003ePostoperative intracranial hemorrhage is a critical complication following cerebral tumor surgery, often associated with increased morbidity and mortality. This study aimed to predict the risk of postoperative intracerebral hemorrhage in patients undergoing intracranial tumor surgery by employing machine learning (ML) algorithms for risk stratification and identifying key contributing factors.\u003c/em\u003e\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eA retrospective analysis was conducted on 118 patients who underwent intracranial tumor surgery and were monitored in the neurosurgical intensive care unit between January 2024 and January 2025. Patients with radiologically confirmed hematomas\u0026thinsp;\u0026ge;\u0026thinsp;5 cm\u0026sup3; on brain CT within 24\u0026ndash;48 hours postoperatively were classified as \"Positive\" for bleeding, while others were labeled \"Negative.\" Clinical and biochemical parameters were analyzed using SPSS and R. Multiple ML algorithms\u0026mdash;including Bagging MARS, Boosting C5.0, SVM, and Random Forests\u0026mdash;were developed and evaluated using performance metrics such as AUC, F-score, accuracy, and Brier score.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eThe Bagging MARS model demonstrated superior predictive performance, with a test AUC of 0.8693, accuracy of 80.8%, Brier score of 0.1580, and F-score of 0.8649. Platelet count, serum sodium level, and glomerular filtration rate (GFR) emerged as the most influential predictors of hemorrhage. Model explainability was enhanced using SHAP and LIME analyses, offering both global and local interpretability of the predictions.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e\u003cp\u003eML algorithms, particularly Bagging MARS, show high accuracy in predicting postoperative hemorrhage following brain tumor surgery. Biomarkers such as platelet count, sodium, and GFR offer clinically meaningful insights for early risk detection and intervention. Integration of these predictive models into clinical decision support systems may significantly improve postoperative monitoring and patient outcomes.\u003c/p\u003e","manuscriptTitle":"Prediction of postoperative haemorrhage after cerebral tumour surgery using machine learning algorithms","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-07-23 02:16:30","doi":"10.21203/rs.3.rs-7012786/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-08-13T09:40:30+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-08-11T23:56:02+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-08-06T13:08:07+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"212899861722046984885630298030852128533","date":"2025-07-23T04:45:54+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-07-21T18:29:16+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-07-20T11:01:57+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"259780843003150203285249174312937590724","date":"2025-07-18T13:58:33+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"168441128163825429531547501459640765316","date":"2025-07-18T05:22:46+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"121362788751923286048438709548487838078","date":"2025-07-18T02:07:42+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-07-17T23:19:41+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-07-16T23:14:41+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-07-15T07:34:01+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-07-14T19:04:39+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Informatics and Decision Making","date":"2025-07-14T19:01:56+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-informatics-and-decision-making","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"midm","sideBox":"Learn more about [BMC Medical Informatics and Decision Making](http://bmcmedinformdecismak.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/midm/default.aspx","title":"BMC Medical Informatics and Decision Making","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"77bb19a2-4847-4ba1-9114-7dbf9dfe5213","owner":[],"postedDate":"July 23rd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-10-27T16:28:37+00:00","versionOfRecord":{"articleIdentity":"rs-7012786","link":"https://doi.org/10.1186/s12911-025-03245-8","journal":{"identity":"bmc-medical-informatics-and-decision-making","isVorOnly":false,"title":"BMC Medical Informatics and Decision Making"},"publishedOn":"2025-10-23 16:17:28","publishedOnDateReadable":"October 23rd, 2025"},"versionCreatedAt":"2025-07-23 02:16:30","video":"","vorDoi":"10.1186/s12911-025-03245-8","vorDoiUrl":"https://doi.org/10.1186/s12911-025-03245-8","workflowStages":[]},"version":"v1","identity":"rs-7012786","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7012786","identity":"rs-7012786","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}