Machine Learning Integration of Baseline CT Imaging and Clinical Parameters Predicts 90-Day Functional Outcomes in Spontaneous Intracerebral Hemorrhage

Machine Learning Integration of Baseline CT Imaging and Clinical Parameters Predicts 90-Day Functional Outcomes in Spontaneous Intracerebral Hemorrhage | 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 Machine Learning Integration of Baseline CT Imaging and Clinical Parameters Predicts 90-Day Functional Outcomes in Spontaneous Intracerebral Hemorrhage Tianyu Yang, Shengkai Yang, Yunji Shan, Yangchun Gu, Jingshan Tao, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7363416/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 18 You are reading this latest preprint version Abstract Objective This study aims to investigate the predictive value of machine learning models integrating initial computed tomography (CT) imaging features and admission clinical parameters for 90-day functional outcomes in patients with spontaneous basal ganglia hemorrhage (BGH). The objective is to establish a clinically applicable tool for early identification of high-risk populations and guide personalized intervention strategies. Methods Patients with BGH admitted to Affiliated Binhai Hospital,Kangda College of Nanjing Medical University from January 2022 to December 2024 were retrospectively collected. Their clinical and imaging data at admission were gathered. The patients were divided into a training set and a test set in a 7:3 ratio. Univariate analysis and multivariate Logistic regression were applied to screen for risk factors. Based on the screened variables, three machine learning algorithms, namely LogisticRegression (LR), RandomForest (RF), and Support Vector Machine (SVM), were used to construct a clinical-only model, an imaging-only model, and a clinical-imaging integrated model through 5-fold cross-validation, aiming to predict the 90-day prognosis of patients with BGH. The predictive efficacy of the models was evaluated using receiver operating characteristic (ROC) curves and the area under the curve (AUC). Results Based on clinical and imaging features, we constructed 9 machine learning models. Among them, the clinical-imaging RF model showed an AUC of 0.97, a sensitivity of 0.86, and a specificity of 0.94; the LR model had an AUC of 0.89, a sensitivity of 0.85, and a specificity of 0.78; and the SVM model presented an AUC of 0.88, a sensitivity of 0.88, and a specificity of 0.74. These results indicated that the RF model had superior predictive performance. In addition, compared with the clinical RF model (AUC 0.86, sensitivity 0.85, specificity 0.73) and the imaging RF model (AUC 0.79, sensitivity 0.65, specificity 0.81), the predictive performance of the clinical-imaging RF model was significantly improved. A feature map was used to clarify the importance of variables in the optimal model. The AUC of the clinical-imaging RF model reached 0.84 in the test set, suggesting that this model has stability. Conclusion The RF model established in this study based on baseline hematoma volume at admission, minor/major axis ratio of the largest axial slice of the hematoma, perihematomal edema volume, neutrophil-to-lymphocyte ratio (NLR), and Glasgow Coma Scale (GCS) score exhibits reliable predictive performance for the 90-day prognosis of patients with BGH. It holds significant clinical application value and practical guiding significance. Basal ganglia hemorrhage CT scan Machine learning Prognosis Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 1. Introduction Basal ganglia hemorrhage(BGH)is the most common subtype of spontaneous intracerebral hemorrhage༈sICH༉, mostly caused by the rupture and bleeding of small arteries in the basal ganglia region due to long-term hypertension(1). It has an acute onset, with high mortality and disability rates, exerting significant negative impacts on families and society(2). Therefore, conducting prognostic assessment for patients in the early stage of onset is conducive to the priority management of high-risk patients and the selection of individualized treatment plans. Meanwhile, a reliable prediction model also has a positive impact on the verification and evaluation of new treatment regimens in clinical trials. Machine learning models constructed based on clinical features have shown certain value in the diagnosis and treatment of patients with cerebral hemorrhage(3), but their accuracy needs to be further improved. With the development of artificial intelligence technology, it has demonstrated strong advantages in image segmentation and analysis(4). The value of extracting imaging features through AI technology in predicting the prognosis of patients with cerebral hemorrhage requires further research. Therefore, in our study, artificial intelligence was used to analyze imaging images, combined with relevant clinical factors of patients at admission, to screen out factors affecting the short-term prognosis of patients with BGH. We comprehensively compared the predictive capabilities of different machine learning models and established an effective prognostic prediction tool to assist clinical decision-making. 2. Materials and Methods 2.1 Study Population In this study, patients with BGH who were admitted to Affiliated Binhai Hospital,Kangda College of Nanjing Medical University from January 2022 to December 2024 and evaluated by admission CT were retrospectively collected. All patients were followed up at 90 days using the modified Rankin Scale (mRS) for prognostic assessment, and were divided into a favorable prognosis group (mRS ≤ 2) and an poor prognosis group (mRS ≥ 3). The patient selection process, including exclusion criteria and inclusion criteria, is shown in Fig. 1 . This study was approved by the Ethics Committee of Affiliated Binhai Hospital,Kangda College of Nanjing Medical University, with the ethics approval number: 2025BYKYLL028. Since this was a retrospective analysis and the data had been de-identified, the Ethics Committee waived the requirement for obtaining signed informed consent from patients. The study was conducted in accordance with the Declaration of Helsinki and relevant international medical ethics guidelines. 2.2 Examination Methods A multi-detector spiral CT (DualSourceCT, SOMATOM, Germany) was used for cranial CT scanning. The patients were in the supine position with the head first, and the scanning range was from the skull base to the top of the skull. The scanning parameters were as follows: tube voltage 120 kV, tube current 200 mA, slice thickness 5 mm, matrix 512×512, and thin-slice reconstruction 1.25 mm. 2.3 Image Processing and Data Acquisition All baseline cranial CT images of the patients were exported from the PACS system in DICOM format and imported into the United Imaging uAI intelligent analysis platform (Version: 1.0.9127.181). Relevant parameters were automatically segmented and obtained via AI algorithms, including intracranial hematoma volume, the ratio of short diameter to long diameter at the maximum axial slice of the hematoma, volume of perilesional edema, intraventricular hematoma volume, and midline shift distance. The accuracy of AI segmentation was independently evaluated by two senior physicians with over 10 years of experience in neuroimaging diagnosis, who also reviewed and recorded imaging features: bleeding location, irregular morphology sign, intralesional hypodensity, swirl sign, black hole sign, blend sign, satellite sign, island sign, and fluid level. Any discrepancies were resolved through consensus reached by joint discussion. Clinical data of patients were collected from the hospital medical record system, including the following categories.Demographic characteristics: age and gender;Clinical history: hypertension, diabetes mellitus, and history of anticoagulant use;Vital signs within 24 hours after admission: maximum systolic blood pressure, diastolic blood pressure, pulse pressure, and initial Glasgow Coma Scale (GCS) score at admission;Laboratory indicators within 24 hours after admission:Routine blood tests: white blood cell count, absolute neutrophil count, absolute lymphocyte count, neutrophil-to-lymphocyte ratio (NLR), and platelet count;Metabolic indicators: blood glucose, total cholesterol, triglycerides, high-density lipoprotein (HDL), and low-density lipoprotein (LDL);Coagulation function: prothrombin time (PT), PT activity, international normalized ratio (INR), activated partial thromboplastin time (APTT), thrombin time (TT), fibrinogen (FIB), and D-dimer.The 90-day mRS score of patients obtained through outpatient follow-up or telephone follow-up was used as the study endpoint. 2.4 Statistical Analysis Data analysis was performed using the Shukun Research Platform. Univariate analyses were conducted for clinical and imaging features: the chi-square test was used for categorical variables; the independent samples t-test was applied to normally distributed continuous variables; and the Mann-Whitney U test was employed for non-normally distributed continuous variables. Variables with P < 0.05 were included in multivariate Logistic regression to identify independent risk factors for 90-day poor prognosis in patients with intracerebral hemorrhage. Based on the selected independent predictors, three machine learning models(LR, RF and SVM model)were constructed using 5-fold cross-validation. Model performance was assessed in the test set by generating receiver operating characteristic (ROC) curves and calculating the area under the curve (AUC). 3. Results A total of 209 patients with BGH were finally included in the study. Using random numbers, they were divided into a training set (n = 146, 70%) and a test set (n = 63, 30%). According to the modified Rankin Scale (mRS) score, the training set consisted of 71 patients with favorable prognosis and 75 patients with poor prognosis. A total of 37 features were included, including demographic characteristics, clinical history, vital signs within 24 hours of admission, first Glasgow Coma Scale (GCS) score at admission, laboratory indicators within 24 hours of admission, and imaging features. Univariate analysis showed that among clinical features, neutrophil count, fibrinogen (FG), GCS score, blood glucose, white blood cell count, and neutrophil-to-lymphocyte ratio (NLR) exhibited statistically significant differences between the two groups (P < 0.05). Among imaging features, initial hematoma volume, Minor/Major Axis ratio of the hematoma, perihematomal edema volume, midline shift, intraventricular extension, hypodense sign of hematoma, swirl sign, black hole sign, and mixed sign showed statistically significant differences between the two groups (P < 0.05) (Table 1 ).Univariate analysis revealed no significant differences in demographic or clinical laboratory variables between the progression and non-progression groups. However, significant differences were observed in imaging features, including initial hematoma volume, minor/major axis ratio, surrounding oedema volume, midline shift distance and the presence of hypodensities, swirl sign, black hole sign and island sign ( P < 0.05), as shown in Table 1 . Table 1 Univariate Analysis for 90-Day Outcome Prediction in BGH Prognosis Feature Poor(75) Favorable(71) P Value Demographic Age, mean (SD) 57.25 (± 15.06) 58.98 (± 12.61) 0.451 Gender(male), n (%) 51(71.8) 50(66.7) 0.499 Clinical characteristics Hypertension, n (%) 58(81.7) 62(82.7) 0.877 Diabetes Mellitus, n (%) 8(11.3) 10(13.3) 0/704 Diastolic pressure(mmHg), median (IQR) 95.0(19.5) 91.0(23.5) 0.372 Systolic pressure(mmHg), median (IQR) 156.0(29.0) 160.0(36.5) 0.882 GCS, median (IQR) 10.0(4.5) 14.0(3.0) < 0.001 D-dimer(mg/L), median (IQR) 188.0(507.5) 133.0(201.5) 0.065 Fibrinogen(g/L), median (IQR) 2.84(1.07) 2.34(0.95) 0.017 INR, mean (SD) 1.07(0.11) 1.05(0.08) 0.181 Prothrombin Time, median (IQR) 11.8(1.2) 11.6(0.85) 0.221 Prothrombin Activity(%), median (IQR) 97.0(15.0) 95.0(16.5) 0.378 Thrombin Time(s), median (IQR) 14.0(1.95) 14.3(1.75) 0.077 Platelet, mean (SD) 209.94 ± 62.15 205.16 ± 56.31 0.626 High-Density Lipoprotein, median (IQR) 1.15(0.45) 1.15(0.37) 0.647 Low-Density Lipoprotein, median (IQR) 2.23(0.82) 2.48(0.615) 0.392 Total Cholesterol, mean (SD) 4.56 ± 0.98 4.40 ± 0.91 0.310 Triglyceride, median (IQR) 1.17(0.69) 1.17(1.04) 0.619 Lymphocyte, median (IQR) 1.17(0.74) 1.30(0.72) 0.089 Neutrophil, median (IQR) 9.10(3.38) 7.06(3.29) < 0.001 NLR, median (IQR) 8.26(6.72) 5.94(4.45) < 0.001 White Blood Cell, median (IQR) 10.73(3.27) 8.66(3.23) < 0.001 Blood glucose, median (IQR) 7.34(3.04) 6.20(2.48) 0.005 Cholesterol, mean (SD) 4.56 ± 0.98 4.40 ± 0.91 0.310 Imaging features BHV(ml), median (IQR) 38.0(21.2) 14.4(10.65) < 0.001 PHEV (ml), median (IQR) 23.3(13.3) 11.1(7.7) < 0.001 Minor/Major axis, median (IQR) 0.64(0.20) 0.51(0.15) < 0.001 Midline shift (mm), median (IQR) 4.3(4.5) 0.0(0.0) < 0.001 Irrgular system sign,n(%) 48(67.6) 39(52) 0.055 Hypodensities, n (%) 42(59.2) 31(41.3) 0.031 Swirl sign, n (%) 39(54.9) 21(28) 0.001 Black hole sign, n (%) 19(26.8) 4(5.3) 0.001 Blend sign, n (%) 25(35.2) 12(16) 0.008 Satellite sign, n (%) 4(5.6) 0(0.0) 0.115 Island Sign, n (%) 10(14.1) 4(5.3) 0.073 Intraventricular Hemorrhage, n (%) 22(29.3%) 9(12.6%) 0.005 Hemorrhage from other sites, n (%) 15(19.5) 14(27.5) 0.362 GCS: Glasgow Coma Scale INR: International Normalized Ratio NLR: Neutrophil-to-Lymphocyte Ratio BHV:Baseline hematoma volume PHEV: Perihematoma edema volume Based on the screened clinical features, clinical feature models were constructed using LR, RF and SVM methods, respectively. The specificity, sensitivity, and AUC of the models in the training and testing sets are presented in Fig. 2 and Table 2 . Table 2 The specificity, sensitivity, and AUC of the clinical-only model in the training set and testing set ML Dataset AUC SEN SPE LR Train 0.86 0.85 0.73 Test 0.81 0.93 0.37 RF Train 0.85 0.80 0.80 Test 0.77 0.74 0.59 SVM Train 0.83 0.76 0.77 Test 0.79 0.83 0.62 ML: Machine Learning LR: LogisticRegression RF: RandomForest SVM: Support Vector Machine Based on the screened imaging features, imaging feature models were constructed using the aforementioned machine learning methods, respectively. The specificity, sensitivity, and AUC of the models in the training and testing sets are presented in Fig. 3 and Table 3 . Table 3 The specificity, sensitivity, and AUC of the imaging-only model in the training set and testing set ML Dataset AUC SEN SPE LR Train 0.79 0.65 0.81 Test 0.76 0.50 0.87 RF Train 0.83 0.80 0.69 Test 0.79 0.68 0.70 SVM Train 0.75 0.58 0.88 Test 0.70 0.34 0.87 ML: Machine Learning LR: LogisticRegression RF: RandomForest SVM: Support Vector Machine By integrating the features with the highest weight coefficients from the clinical-imaging integrated models, a clinical-imaging integrated model was constructed by selecting NLR and GCS from clinical features, as well as baseline hematoma volume, perihematomal edema volume, and the ratio of short diameter to long diameter of the hematoma from imaging features. The specificity, sensitivity, and AUC of the model in the training and testing sets are shown in Fig. 4 and Table 4 . Among them, the RF model performed excellently in the training set, with an AUC of 0.97 (95%CI [0.95, 0.99]), and an AUC of 0.84 (95% CI[0.74, 0.94]) in the testing set. For the RF model with the best predictive performance, the confusion matrices of the training and testing sets (5-A, 5-B), precision-recall curve (5-C), and clinical calibration curve (5-D) are presented in Fig. 5 . The weight coefficients of each feature in the RF model are shown in Table 5 and Fig. 6 . Table 4 The specificity, sensitivity, and AUC of the clinical-imaging integrated model in the training set and testing set ML Dataset AUC SEN SPE LR Train 0.89 0.85 0.78 Test 0.88 0.84 0.83 RF Train 0.97 0.86 0.94 Test 0.84 0.78 0.80 SVM Train 0.88 0.88 0.74 Test 0.81 0.84 0.61 ML: Machine Learning LR: LogisticRegression RF: RandomForest SVM: Support Vector Machine Table 5 The Importance of each feature in the clinical-imaging RF model Feature_name Importance Baseline hematoma volume 0.5250 Minor/major axis 0.1472 Perihematoma edema volume 0.1777 GCS 0.0808 NLR 0.0693 4. Discussion BGH as a common subtype of spontaneous intracerebral hemorrhage(sICH), is closely associated with factors such as age, history of hypertension, and lifestyle(5). Its incidence gradually increases with age. After the onset, the time available for physicians to select treatment regimens is limited, and accurate prediction of patient prognosis can assist clinicians in making treatment decisions(6). Meanwhile, for patients' families, timely understanding of the patients' survival and prognosis is also crucial. The main factors affecting the prognosis of sICH are the location and volume of hemorrhage(7), but this is not sufficiently accurate. Even experienced neurosurgeons face challenges in predicting the short-term prognosis of patients. Therefore, establishing an accurate early prognostic model for the subtype of basal ganglia hemorrhage holds important clinical value for the formulation of treatment decisions. In addition, accurate prediction models can also be used for the verification and evaluation of new treatment regimens in clinical trials. As a clinician-reported measure of global disability, the modified Rankin scale(mRS) is extensively utilized to assess outcomes in stroke patients and serves as a primary endpoint in randomized clinical trials(8). Its validity and reliability are well-established through multiple lines of evidence(9). This evaluation index focuses on measuring the recovery of patients' activities of daily living (such as self-care, walking, work, etc.) and serves as a reliable, simple, and easy-to-use tool for evaluating stroke prognosis(10). Therefore, in this study, using the 90-day mRS score of patients as the study endpoint can accurately assess the patients' current living status. In this study, relevant clinical indicators easily accessible in the emergency setting, such as medical history and laboratory tests, were included, and combined with hematoma-related imaging features obtained from non-contrast cranial CT scans for statistical analysis. Previously, the method of manual measurement combined with hematoma volume calculation formulas exhibited significantly reduced accuracy when assessing irregular hematomas and intraventricular hematomas(11,12). With the development of computer technology, the use of AI algorithms or predefined algorithms to automatically identify hematoma regions and calculate volumes has become the mainstream(13). Although their segmentation accuracy for extremely low-density or mixed-density hematomas may decrease, they still demonstrate high stability and accuracy in the diagnosis and treatment of patients with acute intracerebral hemorrhage. After univariate analysis and screening via multivariate logistic regression, among clinical features, a clinical-only model for predicting the 90-day prognosis of patients with BGH was constructed using neutrophil count, FG, fibrinogen, GCS, blood glucose, white blood cell count, and NLR. For this clinical-only model, the Random Forest (RF) model showed an AUC of 0.86, sensitivity of 0.85, and specificity of 0.73; the Logistic Regression (LR) model had an AUC of 0.85, sensitivity of 0.80, and specificity of 0.80; and the Support Vector Machine (SVM) model exhibited an AUC of 0.83, sensitivity of 0.76, and specificity of 0.77. Based on imaging features, 8 characteristic parameters were screened, including initial hematoma volume, ratio of hematoma short diameter to long diameter, perihematomal edema volume, midline shift distance, as well as hematoma hypodensity sign, swirl sign, black hole sign, and island sign, to construct an imaging-only model. For this imaging-only model, the RF model demonstrated an AUC of 0.79, sensitivity of 0.65, and specificity of 0.81; the LR model had an AUC of 0.83, sensitivity of 0.80, and specificity of 0.69; and the SVM model showed an AUC of 0.75, sensitivity of 0.58, and specificity of 0.88. The RF model exhibited better predictive performance than other prediction models. By integrating the 5 most predictive parameters from clinical and imaging parameters, a clinical-imaging integrated model was constructed. For this integrated model, the RF model showed an AUC of 0.97, sensitivity of 0.86, and specificity of 0.94; the LR model had an AUC of 0.89, sensitivity of 0.85, and specificity of 0.78; and the SVM model exhibited an AUC of 0.88, sensitivity of 0.88, and specificity of 0.74. The predictive performance of the clinical-imaging integrated model was significantly improved compared with the clinical-only model and the imaging-only model. Among them, the RF model performed better, with an AUC of 0.84 in the validation set, indicating that the RF model has stable generalization ability and applicability in predicting the 90-day prognosis of patients with BGH. Analysis of feature weights in the clinical-imaging RF model for 90-day prognosis in patients with BGH: Among these features, baseline hematoma volume is the most important. Previous studies have demonstrated that hematoma volume is the most critical factor affecting the prognosis of patients with intracerebral hemorrhage(14). The larger the hematoma volume, the greater the degree of damage to brain tissue, and the more irreversible the impairment of brain function(15).The minor/major axis ratio of the largest axial slice of the hematoma is an imaging feature for evaluating hematoma morphology(16). Hematomas with a larger ratio tend to be round in shape, indicating high intrinsic tension of the hematoma and weak resistance of surrounding brain tissue to hematoma expansion, ultimately leading to expansion in all directions. In contrast, a smaller ratio indicates a spindle-shaped hematoma, which expands along the interstitial spaces of brain tissue with lower tension, resulting in less damage to surrounding brain tissue compared to the high-ratio group(17).Perihematomal edema occurs in the hyperacute phase after intracerebral hemorrhage, appearing as a hypodense zone surrounding the hematoma shortly after its formation(18). It is recognized as an imaging marker of secondary injury following intracerebral hemorrhage, with pathophysiological manifestations including cytotoxic edema, ischemic necrosis, and neuroinflammation(19). This study confirms that the volume of early perihematomal edema is associated with patient prognosis. The neutrophil-to-lymphocyte ratio (NLR), as an inflammatory indicator, can reflect the degree of systemic inflammatory response induced by brain tissue damage(20). A high NLR is often accompanied by suppressed lymphocyte function, making patients prone to complications such as pulmonary or systemic multi-organ infections(21), which are important triggers for death in patients with intracerebral hemorrhage(22). Therefore, a significant increase in NLR may indirectly reflect the severity of the disease, thereby affecting patient prognosis.The Glasgow Coma Scale (GCS) is a classic scale for assessing the degree of consciousness disturbance(23), directly reflecting the functional status of the brain after intracerebral hemorrhage(24). Numerous previous studies have confirmed that the GCS score is an independent predictor of 30-day mortality in intracerebral hemorrhage; a lower GCS score indicates a higher risk of death(25).In summary, the multi-factor integrated model combining clinical and imaging indicators can significantly improve the accuracy of prediction. The strengths of this study lie in the use of AI-based extraction of imaging features, Artificial intelligence approaches exhibit exceptional proficiency in the automatic identification of intricate patterns within imaging data and the provision of quantitative outputs(26), which offers stability and reproducibility. By integrating clinical risk factors and comparing different machine learning models, the RF model was shown to have superior predictive performance. Results from the validation set also demonstrated that the RF model constructed based on clinical and imaging features has favorable predictive value for the short-term prognosis of patients with BGH. Compared with machine learning models developed in previous studies (3), our model significantly enhanced predictive performance by incorporating imaging features that assess the degree of brain tissue damage caused by intracranial hematoma. This study has certain limitations. Firstly, as a retrospective study, its data were derived from previous clinical records. Restricted by the study design, selection bias may have existed during case enrollment and data extraction, which could affect the generalizability of the research results. Additionally, the model in this study was constructed only based on an internal dataset, lacking validation with external independent cohorts. Therefore, future research urgently needs to conduct multicenter, large-sample prospective studies, combined with external validation across multiple regions and institutions, to further evaluate the accuracy of the model. 5. Conclusion The RF model established based on clinical and imaging features at admission exhibits reliable predictive performance for the 90-day prognosis of patients with BGH, and holds significant clinical application value as well as practical guiding significance. Declarations Author Contributions TYY and ZZ substantially contributed to the conception and design of the study. YJS and JST performed research and prepared figures. YTY and SKY analyzed and interpreted the data and drafted the article. ZZ made critical revisions related to the important intellectual content of the manuscript and final approval. JST and YCG provided supervision. All authors read and approved the final manuscript. Funding This work was supported by Science Foundation of Kangda College of Nanjing Medical University (KD2024KYJJ176),the Beijing Medical Award Foundation (YXJL-2024-0299-0062), and the Scientific Research Project of Yancheng Municipal Health Commission (YK2024067). Availability of data and materials The data presented in this study are available upon request from the corresponding author. Data Sharing Statement The data presented in this study are available on request from the corresponding author. Ethics Statement The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Institutional Review Board of the Affiliated Binhai Hospital,Kangda College of Nanjing Medical University (ethics numbers: 2025BYKYLL028). Consent for Publication Informed consent was obtained from all subjects involved in the study. Disclosure The authors report no conflicts of in this work. References Sheth KN. Spontaneous Intracerebral Hemorrhage. N Engl J Med. 2022 Oct 27;387(17):1589–96. Vandertop WP, Can A, Post R. Spontaneous Intracerebral Hemorrhage. N Engl J Med. 2023 Jan 12;388(2):191–2. Geng Z, Yang C, Zhao Z, Yan Y, Guo T, Liu C, et al. Development and validation of a machine learning-based predictive model for assessing the 90-day prognostic outcome of patients with spontaneous intracerebral hemorrhage. J Transl Med. 2024 Mar 4;22:236. Baeßler B, Engelhardt S, Hekalo A, Hennemuth A, Hüllebrand M, Laube A, et al. Perfect Match: Radiomics and Artificial Intelligence in Cardiac Imaging. Circ Cardiovasc Imaging. 2024 Jun;17(6):e015490. Liu D, Zhang G, Wang Y, Li J, Cao P, Yin X, et al. Geometric features of middle cerebral artery are associated with spontaneous basal ganglia intracerebral haemorrhage. Stroke Vasc Neurol. 2022 Mar 9;7(5):399–405. Al-Kawaz MN, Hanley DF, Ziai W. Advances in Therapeutic Approaches for Spontaneous Intracerebral Hemorrhage. Neurotherapeutics. 2020 Oct;17(4):1757–67. He M, Lu Z, Lv Y, Cheng Z, Zhang Q, Jin X, et al. Machine learning-based prediction of 6-month functional recovery in hypertensive cerebral hemorrhage: insights from XGBoost and SHAP analysis. Front Neurol. 2025;16:1608341. Jp B, O A, J E. Evolution of the Modified Rankin Scale and Its Use in Future Stroke Trials[J]. Stroke, Stroke, 2017, 48(7). Banks JL, Marotta CA. Outcomes validity and reliability of the modified Rankin scale: implications for stroke clinical trials: a literature review and synthesis. Stroke. 2007 Mar;38(3):1091–6. Haggag H, Hodgson C. Clinimetrics: Modified Rankin Scale (mRS). Journal of Physiotherapy. 2022 Oct 1;68(4):281. B Z, Wb J, Ly Z, et al. 1/2SH: A Simple, Accurate, and Reliable Method of Calculating the Hematoma Volume of Spontaneous Intracerebral Hemorrhage[J]. Stroke, Stroke, 2020, 51(1). Xu X, Chen X, Zhang J, Zheng Y, Sun G, Yu X, et al. Comparison of the Tada formula with software slicer: precise and low-cost method for volume assessment of intracerebral hematoma. Stroke. 2014 Nov;45(11):3433–5. Petrov A, Kashevnik A, Haleev M, Ali A, Ivanov A, Samochernykh K, et al. AI-Based Approach to One-Click Chronic Subdural Hematoma Segmentation Using Computed Tomography Images. Sensors (Basel). 2024 Jan 23;24(3):721. M L, Z W, X M, Y Z, X H, L L, et al. Predictive Nomogram for Unfavorable Outcome of Spontaneous Intracerebral Hemorrhage[J]. World neurosurgery, World Neurosurg, 2022, 164. Kc T, Sm F, Wcy L, Iyh L, Yk W, Omy C, et al. Location-Specific Hematoma Volume Cutoff and Clinical Outcomes in Intracerebral Hemorrhage[J]. Stroke, Stroke,2023, 54(6). Yang T, Zhao Z, Gu Y, Yang S, Zhang Y, Li L, et al. Evaluating the value of machine learning models for predicting hematoma expansion in acute spontaneous intracerebral hemorrhage based on CT imaging features of hematomas and surrounding oedema. Front Neurol. 2025;16:1567525. R R, M S, S T, M T, M K, K K, et al. Mechanism of Spontaneous Intracerebral Hemorrhage Formation: An Anatomical Specimens-Based Study[J]. Stroke, Stroke,2022, 53(11). Y C, S C, J C, et al. Perihematomal Edema After Intracerebral Hemorrhage: An Update on Pathogenesis, Risk Factors, and Therapeutic Advances[J]. Frontiers in immunology, Front Immunol, 2021, 12. C J, H G, Z Z, et al. Molecular, Pathological, Clinical, and Therapeutic Aspects of Perihematomal Edema in Different Stages of Intracerebral Hemorrhage[J]. Oxidative medicine and cellular longevity, Oxid Med Cell Longev, 2022, 2022. P G, Y L, Y G, et al. The association of neutrophil to lymphocyte ratio, platelet to lymphocyte ratio, and lymphocyte to monocyte ratio with post-thrombolysis early neurological outcomes in patients with acute ischemic stroke[J]. Journal of neuroinflammation, J Neuroinflammation, 2021, 18(1). Y Z, W P, X Z. The prognostic value of the combined neutrophil-to-lymphocyte ratio (NLR) and neutrophil-to-platelet ratio (NPR) in sepsis[J]. Scientific reports, Sci Rep, 2024, 14(1). Js B, Am B. Complications of intracerebral haemorrhage[J]. The Lancet. Neurology, Lancet Neurol, 2012, 11(1). k Z, Tu D, G T, et al. The Message of the Glasgow Coma Scale: A Comprehensive Bibliometric Analysis and Systematic Review of Clinical Practice Guidelines Spanning the Past 50 years[J]. World neurosurgery, World Neurosurg, 2024, 185. A L, Fd A, Am V, et al. Admission Glasgow Coma Scale Score as a Predictor of Outcome in Patients Without Traumatic Brain Injury[J]. American journal of critical care : an official publication, American Association of Critical-Care Nurses, Am J Crit Care, 2021, 30(5). Cw W, Yj L, Yh L, et al. Hematoma shape, hematoma size, Glasgow coma scale score and ICH score: which predicts the 30-day mortality better for intracerebral hematoma?[J]. PloS one, PLoS One, 2014, 9(7). Bs K, C J, Sm B, et al. Radiology artificial intelligence: a systematic review and evaluation of methods (RAISE)[J]. European radiology, Eur Radiol, 2022, 32(11). Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 05 Nov, 2025 Reviews received at journal 24 Oct, 2025 Reviews received at journal 24 Oct, 2025 Reviews received at journal 20 Oct, 2025 Reviews received at journal 13 Oct, 2025 Reviews received at journal 09 Oct, 2025 Reviewers agreed at journal 07 Oct, 2025 Reviewers agreed at journal 05 Oct, 2025 Reviewers agreed at journal 04 Oct, 2025 Reviewers agreed at journal 04 Oct, 2025 Reviews received at journal 03 Oct, 2025 Reviewers agreed at journal 02 Oct, 2025 Reviewers agreed at journal 02 Oct, 2025 Reviewers agreed at journal 02 Oct, 2025 Reviewers invited by journal 02 Oct, 2025 Editor assigned by journal 14 Aug, 2025 Submission checks completed at journal 14 Aug, 2025 First submitted to journal 13 Aug, 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-7363416","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":528842917,"identity":"870e848a-03ef-4dab-aed9-96af4eeb67a2","order_by":0,"name":"Tianyu Yang","email":"","orcid":"","institution":"Kangda College of Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Tianyu","middleName":"","lastName":"Yang","suffix":""},{"id":528842918,"identity":"f1304c22-8ae5-4ef4-8dce-4bf73dc8d4fa","order_by":1,"name":"Shengkai Yang","email":"","orcid":"","institution":"Kangda College of Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Shengkai","middleName":"","lastName":"Yang","suffix":""},{"id":528842919,"identity":"47edcf93-11b0-487f-a34c-1d40d9062ab8","order_by":2,"name":"Yunji Shan","email":"","orcid":"","institution":"Kangda College of Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yunji","middleName":"","lastName":"Shan","suffix":""},{"id":528842920,"identity":"1acfe512-ada1-4b58-a5c4-58e12c9b842e","order_by":3,"name":"Yangchun Gu","email":"","orcid":"","institution":"Kangda College of Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yangchun","middleName":"","lastName":"Gu","suffix":""},{"id":528842921,"identity":"3000132b-72ea-42a7-9edc-3ceb2519b25d","order_by":4,"name":"Jingshan Tao","email":"","orcid":"","institution":"Kangda College of Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Jingshan","middleName":"","lastName":"Tao","suffix":""},{"id":528842922,"identity":"8e2186cb-2709-4973-86f4-a069b4ce02f7","order_by":5,"name":"Zhen Zhao","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA5ElEQVRIie3PMQrCMBiG4RShXQKufynUK/wSKA7FszQI7VJE6AVaCnEpuFb0ELo5VoJOHqCjuYGTKAhqd2l1c8gL2fKQfITodP8YvI+REtK3srS6oO9+T+xCZqqchex7gnWUM3qRvFMMVrlS950/tYu9cHzsBcSSh00bMdZHhvYpTN5bhBOjOSU0DOs20oPAg6GQfNm8EiNNCFCvlZgQXYGLJ9/UXDgjBJ52EQqxB3tRNSRnBLGbAMTJMBOT5mOZKjBgZteWQRlt1V2M+cKan6vb4+n2LXlsJR/W/XZdp9PpdJ96AZ9rTvAQN21iAAAAAElFTkSuQmCC","orcid":"","institution":"Zhongda Hospital, Southeast University","correspondingAuthor":true,"prefix":"","firstName":"Zhen","middleName":"","lastName":"Zhao","suffix":""}],"badges":[],"createdAt":"2025-08-13 09:38:40","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7363416/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7363416/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":93616602,"identity":"2aef6672-329a-4251-94c8-b6889b0fba3a","added_by":"auto","created_at":"2025-10-15 17:00:59","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":958633,"visible":true,"origin":"","legend":"","description":"","filename":"Manuscript.docx","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/15f37c7d12dcc091b5101b9c.docx"},{"id":93616597,"identity":"eac1611e-a7ae-4953-b16f-aad88b5a7b64","added_by":"auto","created_at":"2025-10-15 17:00:59","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8517,"visible":true,"origin":"","legend":"","description":"","filename":"4d8fee1607d64e698a029f32dfd1c13e.json","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/66356f863de23ccfb82cae1f.json"},{"id":93616606,"identity":"b161efea-3f38-4d18-b417-23bc804b15ae","added_by":"auto","created_at":"2025-10-15 17:00:59","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":78389,"visible":true,"origin":"","legend":"","description":"","filename":"4d8fee1607d64e698a029f32dfd1c13e1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/fb73d8952249f1942a511f86.xml"},{"id":93616599,"identity":"0b67960a-b715-4265-bf31-95da9eb6da07","added_by":"auto","created_at":"2025-10-15 17:00:59","extension":"jpeg","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":375145,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/a8255c255b90780b23bb2a03.jpeg"},{"id":93618685,"identity":"95fd4dc3-931f-4491-90f6-c5e2ce8be760","added_by":"auto","created_at":"2025-10-15 17:24:59","extension":"jpeg","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1074,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/d667a0bb7b2649b911e08627.jpeg"},{"id":93616607,"identity":"1e0cccea-5dc6-4a2c-b00d-3b356cc51f2c","added_by":"auto","created_at":"2025-10-15 17:00:59","extension":"jpeg","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":294121,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/7a6257862b020170df57c203.jpeg"},{"id":93618087,"identity":"fd502460-94f3-46a5-8229-96271b30f412","added_by":"auto","created_at":"2025-10-15 17:16:59","extension":"jpeg","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":307914,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/233ade83fa95a55848a1d3d6.jpeg"},{"id":93618086,"identity":"1a579394-f5d7-4103-944e-d17470773be4","added_by":"auto","created_at":"2025-10-15 17:16:59","extension":"jpeg","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1074,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/3bc7f767aab4dc6f9c6ea599.jpeg"},{"id":93617605,"identity":"c04b37bf-2ab2-4c18-83de-3f8054a56e4d","added_by":"auto","created_at":"2025-10-15 17:08:59","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":171919,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/fa5b4e4ebda5ea334422c440.png"},{"id":93616619,"identity":"0468d3bd-387c-48e5-87db-577fa3fd904c","added_by":"auto","created_at":"2025-10-15 17:00:59","extension":"jpeg","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":316247,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/e301fb5e6c0157fb74a38f1f.jpeg"},{"id":93616610,"identity":"499cdc83-61b1-4e7c-bff5-75d4a57514f9","added_by":"auto","created_at":"2025-10-15 17:00:59","extension":"jpeg","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":50238,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage8.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/6a8ce28c5b30457d208d281f.jpeg"},{"id":93617610,"identity":"157be03c-b855-4ab6-aabb-6497d2e9671b","added_by":"auto","created_at":"2025-10-15 17:08:59","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":38128,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/80e2221ab8cafeb44ee6b56f.png"},{"id":93618093,"identity":"4608fa46-d684-449f-bca6-b5d7f1c5a0b9","added_by":"auto","created_at":"2025-10-15 17:16:59","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":196940,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/6e4bb0d9de31446861fef752.png"},{"id":93618687,"identity":"4cf554a3-0ad5-42f3-b3bb-103a74a58a33","added_by":"auto","created_at":"2025-10-15 17:24:59","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":935,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/248d2d8b5c22d21b70f33771.png"},{"id":93617602,"identity":"3c575796-dd33-4ad2-bf8d-f9511e76afd9","added_by":"auto","created_at":"2025-10-15 17:08:59","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":56238,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/6a84c817bb96bb2b9d933d12.png"},{"id":93616617,"identity":"5f8d85d0-f98e-4373-a847-52562b74d087","added_by":"auto","created_at":"2025-10-15 17:00:59","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":57188,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/352070041f8c8cca68ac05b5.png"},{"id":93618088,"identity":"1becf625-330d-41c9-b7fe-8a748d927655","added_by":"auto","created_at":"2025-10-15 17:16:59","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":935,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/9d9e6b090f8dab7dd4273465.png"},{"id":93617608,"identity":"0a170c2f-944b-42c3-bff1-438a10688bd3","added_by":"auto","created_at":"2025-10-15 17:08:59","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":31982,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/40136d0f72824828cbfdbe97.png"},{"id":93616613,"identity":"26852c80-e806-4a41-8762-9ba984e28796","added_by":"auto","created_at":"2025-10-15 17:00:59","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":64893,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/8d98c33656ee43aad9ed7744.png"},{"id":93618688,"identity":"d29c7481-1792-4582-a775-f983f2a1ceed","added_by":"auto","created_at":"2025-10-15 17:24:59","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9825,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/3b5752ef15df16b4249cad8d.png"},{"id":93616620,"identity":"127e10cb-c73c-40e1-abe3-3b0665204095","added_by":"auto","created_at":"2025-10-15 17:00:59","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11593,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/635e6945da4ac9aa0b9e3d76.png"},{"id":93616624,"identity":"bbd634af-9ed5-439e-ab90-fb692a44b45e","added_by":"auto","created_at":"2025-10-15 17:00:59","extension":"xml","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":76687,"visible":true,"origin":"","legend":"","description":"","filename":"4d8fee1607d64e698a029f32dfd1c13e1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/c0618b2cb5b502ae83427cc4.xml"},{"id":93616623,"identity":"fe4ecd1f-3a32-4184-9fea-c1338236b822","added_by":"auto","created_at":"2025-10-15 17:00:59","extension":"html","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":84310,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/644197bc4610f6c0ad54c1bd.html"},{"id":93616596,"identity":"205248a4-af54-4ebd-91d3-e4126191d21f","added_by":"auto","created_at":"2025-10-15 17:00:59","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":82868,"visible":true,"origin":"","legend":"\u003cp\u003ePatient Selection Flowchart\u003c/p\u003e","description":"","filename":"1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/d14d0d89c3305c5eadc53cfa.jpg"},{"id":93616595,"identity":"bb215122-8e0f-41b3-b770-94592fa674a3","added_by":"auto","created_at":"2025-10-15 17:00:59","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":63043,"visible":true,"origin":"","legend":"\u003cp\u003eThe specificity, sensitivity, and AUC of the clinical-only model in the training set and testing set\u003c/p\u003e","description":"","filename":"2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/d283dbcc24c06149bd4710f1.jpg"},{"id":93617598,"identity":"47e64a23-5bf2-4502-9726-9a1e44b37cd4","added_by":"auto","created_at":"2025-10-15 17:08:59","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":66762,"visible":true,"origin":"","legend":"\u003cp\u003eThe specificity, sensitivity, and AUC of the imaging-only model in the training set and testing set\u003c/p\u003e","description":"","filename":"3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/c02201f417df837049923135.jpg"},{"id":93618091,"identity":"48f605e9-6ec3-4fd7-8ef9-c0d71cb6eb62","added_by":"auto","created_at":"2025-10-15 17:16:59","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":61543,"visible":true,"origin":"","legend":"\u003cp\u003eThe specificity, sensitivity, and AUC of the clinical-imaging integrated model in the training set and testing set.\u003c/p\u003e","description":"","filename":"4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/50409daed488b4eb893589f6.jpg"},{"id":93619131,"identity":"4ef3cc72-6b08-466d-b17a-8832ed4201ae","added_by":"auto","created_at":"2025-10-15 17:32:59","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":65305,"visible":true,"origin":"","legend":"\u003cp\u003ePerformance evaluation of the random forest model in the training and testing sets: confusion matrices, precision-recall curves, and clinical calibration curves\u003c/p\u003e","description":"","filename":"5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/c30fe99cdff1eaadfc877d4e.jpg"},{"id":93618083,"identity":"8a29d36f-9e19-499c-aed8-46cdf15cb3d7","added_by":"auto","created_at":"2025-10-15 17:16:59","extension":"jpg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":33206,"visible":true,"origin":"","legend":"\u003cp\u003eWeighting factors for each clinical characteristic.\u003c/p\u003e","description":"","filename":"6.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/f9a346e2b7198c8fb7f199f8.jpg"},{"id":93681990,"identity":"f9697c71-f40f-4937-938a-1fc753bd6e0b","added_by":"auto","created_at":"2025-10-16 12:30:22","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1123513,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7363416/v1/b73634e7-94aa-4b8f-9826-b5c1a092ebcb.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Machine Learning Integration of Baseline CT Imaging and Clinical Parameters Predicts 90-Day Functional Outcomes in Spontaneous Intracerebral Hemorrhage","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eBasal ganglia hemorrhage(BGH)is the most common subtype of spontaneous intracerebral hemorrhage༈sICH༉, mostly caused by the rupture and bleeding of small arteries in the basal ganglia region due to long-term hypertension(1). It has an acute onset, with high mortality and disability rates, exerting significant negative impacts on families and society(2). Therefore, conducting prognostic assessment for patients in the early stage of onset is conducive to the priority management of high-risk patients and the selection of individualized treatment plans. Meanwhile, a reliable prediction model also has a positive impact on the verification and evaluation of new treatment regimens in clinical trials.\u003c/p\u003e\u003cp\u003eMachine learning models constructed based on clinical features have shown certain value in the diagnosis and treatment of patients with cerebral hemorrhage(3), but their accuracy needs to be further improved. With the development of artificial intelligence technology, it has demonstrated strong advantages in image segmentation and analysis(4). The value of extracting imaging features through AI technology in predicting the prognosis of patients with cerebral hemorrhage requires further research.\u003c/p\u003e\u003cp\u003eTherefore, in our study, artificial intelligence was used to analyze imaging images, combined with relevant clinical factors of patients at admission, to screen out factors affecting the short-term prognosis of patients with BGH. We comprehensively compared the predictive capabilities of different machine learning models and established an effective prognostic prediction tool to assist clinical decision-making.\u003c/p\u003e"},{"header":"2. Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Study Population\u003c/h2\u003e\u003cp\u003eIn this study, patients with BGH who were admitted to Affiliated Binhai Hospital,Kangda College of Nanjing Medical University from January 2022 to December 2024 and evaluated by admission CT were retrospectively collected. All patients were followed up at 90 days using the modified Rankin Scale (mRS) for prognostic assessment, and were divided into a favorable prognosis group (mRS\u0026thinsp;\u0026le;\u0026thinsp;2) and an poor prognosis group (mRS\u0026thinsp;\u0026ge;\u0026thinsp;3). The patient selection process, including exclusion criteria and inclusion criteria, is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e This study was approved by the Ethics Committee of Affiliated Binhai Hospital,Kangda College of Nanjing Medical University, with the ethics approval number: 2025BYKYLL028. Since this was a retrospective analysis and the data had been de-identified, the Ethics Committee waived the requirement for obtaining signed informed consent from patients. The study was conducted in accordance with the Declaration of Helsinki and relevant international medical ethics guidelines.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 Examination Methods\u003c/h2\u003e\u003cp\u003eA multi-detector spiral CT (DualSourceCT, SOMATOM, Germany) was used for cranial CT scanning. The patients were in the supine position with the head first, and the scanning range was from the skull base to the top of the skull. The scanning parameters were as follows: tube voltage 120 kV, tube current 200 mA, slice thickness 5 mm, matrix 512\u0026times;512, and thin-slice reconstruction 1.25 mm.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3 Image Processing and Data Acquisition\u003c/h2\u003e\u003cp\u003eAll baseline cranial CT images of the patients were exported from the PACS system in DICOM format and imported into the United Imaging uAI intelligent analysis platform (Version: 1.0.9127.181). Relevant parameters were automatically segmented and obtained via AI algorithms, including intracranial hematoma volume, the ratio of short diameter to long diameter at the maximum axial slice of the hematoma, volume of perilesional edema, intraventricular hematoma volume, and midline shift distance. The accuracy of AI segmentation was independently evaluated by two senior physicians with over 10 years of experience in neuroimaging diagnosis, who also reviewed and recorded imaging features: bleeding location, irregular morphology sign, intralesional hypodensity, swirl sign, black hole sign, blend sign, satellite sign, island sign, and fluid level. Any discrepancies were resolved through consensus reached by joint discussion.\u003c/p\u003e\u003cp\u003eClinical data of patients were collected from the hospital medical record system, including the following categories.Demographic characteristics: age and gender;Clinical history: hypertension, diabetes mellitus, and history of anticoagulant use;Vital signs within 24 hours after admission: maximum systolic blood pressure, diastolic blood pressure, pulse pressure, and initial Glasgow Coma Scale (GCS) score at admission;Laboratory indicators within 24 hours after admission:Routine blood tests: white blood cell count, absolute neutrophil count, absolute lymphocyte count, neutrophil-to-lymphocyte ratio (NLR), and platelet count;Metabolic indicators: blood glucose, total cholesterol, triglycerides, high-density lipoprotein (HDL), and low-density lipoprotein (LDL);Coagulation function: prothrombin time (PT), PT activity, international normalized ratio (INR), activated partial thromboplastin time (APTT), thrombin time (TT), fibrinogen (FIB), and D-dimer.The 90-day mRS score of patients obtained through outpatient follow-up or telephone follow-up was used as the study endpoint.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e2.4 Statistical Analysis\u003c/h2\u003e\u003cp\u003eData analysis was performed using the Shukun Research Platform. Univariate analyses were conducted for clinical and imaging features: the chi-square test was used for categorical variables; the independent samples t-test was applied to normally distributed continuous variables; and the Mann-Whitney U test was employed for non-normally distributed continuous variables. Variables with P\u0026thinsp;\u0026lt;\u0026thinsp;0.05 were included in multivariate Logistic regression to identify independent risk factors for 90-day poor prognosis in patients with intracerebral hemorrhage. Based on the selected independent predictors, three machine learning models(LR, RF and SVM model)were constructed using 5-fold cross-validation. Model performance was assessed in the test set by generating receiver operating characteristic (ROC) curves and calculating the area under the curve (AUC).\u003c/p\u003e\u003c/div\u003e"},{"header":"3. Results","content":"\u003cp\u003eA total of 209 patients with BGH were finally included in the study. Using random numbers, they were divided into a training set (n\u0026thinsp;=\u0026thinsp;146, 70%) and a test set (n\u0026thinsp;=\u0026thinsp;63, 30%). According to the modified Rankin Scale (mRS) score, the training set consisted of 71 patients with favorable prognosis and 75 patients with poor prognosis. A total of 37 features were included, including demographic characteristics, clinical history, vital signs within 24 hours of admission, first Glasgow Coma Scale (GCS) score at admission, laboratory indicators within 24 hours of admission, and imaging features.\u003c/p\u003e\u003cp\u003eUnivariate analysis showed that among clinical features, neutrophil count, fibrinogen (FG), GCS score, blood glucose, white blood cell count, and neutrophil-to-lymphocyte ratio (NLR) exhibited statistically significant differences between the two groups (P\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Among imaging features, initial hematoma volume, Minor/Major Axis ratio of the hematoma, perihematomal edema volume, midline shift, intraventricular extension, hypodense sign of hematoma, swirl sign, black hole sign, and mixed sign showed statistically significant differences between the two groups (P\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).Univariate analysis revealed no significant differences in demographic or clinical laboratory variables between the progression and non-progression groups. However, significant differences were observed in imaging features, including initial hematoma volume, minor/major axis ratio, surrounding oedema volume, midline shift distance and the presence of hypodensities, swirl sign, black hole sign and island sign (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05), as shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\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\u003eUnivariate Analysis for 90-Day Outcome Prediction in BGH\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\u003ePrognosis\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eFeature\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePoor(75)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eFavorable(71)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003eP\u003c/em\u003e Value\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"4\" nameend=\"c4\" namest=\"c1\"\u003e\u003cp\u003e\u003cb\u003eDemographic\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge, mean (SD)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e57.25 (\u0026plusmn;\u0026thinsp;15.06)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e58.98 (\u0026plusmn;\u0026thinsp;12.61)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.451\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGender(male), n (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e51(71.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e50(66.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.499\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e\u003cp\u003e\u003cb\u003eClinical characteristics\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHypertension, n (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e58(81.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e62(82.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.877\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDiabetes Mellitus, n (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e8(11.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e10(13.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0/704\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDiastolic pressure(mmHg), median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e95.0(19.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e91.0(23.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.372\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSystolic pressure(mmHg), median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e156.0(29.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e160.0(36.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.882\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGCS, median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e10.0(4.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e14.0(3.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eD-dimer(mg/L), median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e188.0(507.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e133.0(201.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.065\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFibrinogen(g/L), median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2.84(1.07)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.34(0.95)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.017\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eINR, mean (SD)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.07(0.11)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.05(0.08)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.181\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eProthrombin Time, median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e11.8(1.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e11.6(0.85)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.221\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eProthrombin Activity(%), median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e97.0(15.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e95.0(16.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.378\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eThrombin Time(s), median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e14.0(1.95)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e14.3(1.75)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.077\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePlatelet, mean (SD)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e209.94\u0026thinsp;\u0026plusmn;\u0026thinsp;62.15\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e205.16\u0026thinsp;\u0026plusmn;\u0026thinsp;56.31\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.626\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHigh-Density Lipoprotein, median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.15(0.45)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.15(0.37)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.647\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLow-Density Lipoprotein, median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2.23(0.82)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.48(0.615)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.392\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTotal Cholesterol, mean (SD)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4.56\u0026thinsp;\u0026plusmn;\u0026thinsp;0.98\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4.40\u0026thinsp;\u0026plusmn;\u0026thinsp;0.91\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.310\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTriglyceride, median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.17(0.69)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.17(1.04)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.619\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLymphocyte, median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.17(0.74)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.30(0.72)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.089\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNeutrophil, median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e9.10(3.38)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7.06(3.29)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNLR, median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e8.26(6.72)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5.94(4.45)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWhite Blood Cell, median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e10.73(3.27)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8.66(3.23)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBlood glucose, median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e7.34(3.04)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6.20(2.48)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.005\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCholesterol, mean (SD)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4.56\u0026thinsp;\u0026plusmn;\u0026thinsp;0.98\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4.40\u0026thinsp;\u0026plusmn;\u0026thinsp;0.91\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.310\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"4\" nameend=\"c4\" namest=\"c1\"\u003e\u003cp\u003e\u003cb\u003eImaging features\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBHV(ml), median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e38.0(21.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e14.4(10.65)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePHEV (ml), median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e23.3(13.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e11.1(7.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMinor/Major axis, median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.64(0.20)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.51(0.15)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMidline shift (mm), median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4.3(4.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.0(0.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIrrgular system sign,n(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e48(67.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e39(52)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.055\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHypodensities, n (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e42(59.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e31(41.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.031\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSwirl sign, n (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e39(54.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e21(28)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBlack hole sign, n (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e19(26.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4(5.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBlend sign, n (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e25(35.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e12(16)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.008\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSatellite sign, n (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4(5.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0(0.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.115\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIsland Sign, n (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e10(14.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4(5.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.073\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIntraventricular Hemorrhage, n (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e22(29.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e9(12.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.005\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHemorrhage from other sites, n (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e15(19.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e14(27.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.362\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"4\"\u003eGCS: Glasgow Coma Scale\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"4\"\u003eINR: International Normalized Ratio\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"4\"\u003eNLR: Neutrophil-to-Lymphocyte Ratio\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"4\"\u003eBHV:Baseline hematoma volume\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"4\"\u003ePHEV: Perihematoma edema volume\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eBased on the screened clinical features, clinical feature models were constructed using LR, RF and SVM methods, respectively. The specificity, sensitivity, and AUC of the models in the training and testing sets are presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e and Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\u003cp\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 specificity, sensitivity, and AUC of the clinical-only model in the training set and testing set\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=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eML\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDataset\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAUC\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSEN\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eSPE\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eLR\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.86\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.85\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.73\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.81\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.93\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.37\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eRF\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.85\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.80\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.80\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.77\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.74\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.59\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eSVM\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.83\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.76\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.77\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.79\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.83\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.62\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"5\"\u003eML: Machine Learning\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"5\"\u003eLR: LogisticRegression\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"5\"\u003eRF: RandomForest\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"5\"\u003eSVM: Support Vector Machine\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eBased on the screened imaging features, imaging feature models were constructed using the aforementioned machine learning methods, respectively. The specificity, sensitivity, and AUC of the models in the training and testing sets are presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e and Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\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\u003eThe specificity, sensitivity, and AUC of the imaging-only model in the training set and testing set\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=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eML\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDataset\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAUC\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSEN\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eSPE\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eLR\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.79\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.65\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.81\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.76\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.87\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eRF\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.83\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.80\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.69\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.79\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.68\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.70\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eSVM\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.75\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.58\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.88\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.70\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.34\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.87\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"5\"\u003eML: Machine Learning\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"5\"\u003eLR: LogisticRegression\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"5\"\u003eRF: RandomForest\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"5\"\u003eSVM: Support Vector Machine\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eBy integrating the features with the highest weight coefficients from the clinical-imaging integrated models, a clinical-imaging integrated model was constructed by selecting NLR and GCS from clinical features, as well as baseline hematoma volume, perihematomal edema volume, and the ratio of short diameter to long diameter of the hematoma from imaging features. The specificity, sensitivity, and AUC of the model in the training and testing sets are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e and Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e. Among them, the RF model performed excellently in the training set, with an AUC of 0.97 (95%CI [0.95, 0.99]), and an AUC of 0.84 (95% CI[0.74, 0.94]) in the testing set. For the RF model with the best predictive performance, the confusion matrices of the training and testing sets (5-A, 5-B), precision-recall curve (5-C), and clinical calibration curve (5-D) are presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e. The weight coefficients of each feature in the RF model are shown in Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e and \u003cb\u003eFig.\u0026nbsp;6\u003c/b\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eThe specificity, sensitivity, and AUC of the clinical-imaging integrated model in the training set and testing set\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=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eML\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDataset\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAUC\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSEN\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eSPE\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eLR\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.89\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.85\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.78\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.88\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.84\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.83\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eRF\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.97\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.86\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.94\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.84\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.78\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.80\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eSVM\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.88\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.88\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.74\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.81\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.84\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.61\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"5\"\u003eML: Machine Learning\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"5\"\u003eLR: LogisticRegression\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"5\"\u003eRF: RandomForest\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"5\"\u003eSVM: Support Vector Machine\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eThe Importance of each feature in the clinical-imaging RF model\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"2\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFeature_name\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eImportance\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBaseline hematoma volume\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.5250\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMinor/major axis\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.1472\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePerihematoma edema volume\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.1777\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGCS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.0808\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNLR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.0693\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eBGH as a common subtype of spontaneous intracerebral hemorrhage(sICH), is closely associated with factors such as age, history of hypertension, and lifestyle(5). Its incidence gradually increases with age. After the onset, the time available for physicians to select treatment regimens is limited, and accurate prediction of patient prognosis can assist clinicians in making treatment decisions(6). Meanwhile, for patients' families, timely understanding of the patients' survival and prognosis is also crucial. The main factors affecting the prognosis of sICH are the location and volume of hemorrhage(7), but this is not sufficiently accurate. Even experienced neurosurgeons face challenges in predicting the short-term prognosis of patients. Therefore, establishing an accurate early prognostic model for the subtype of basal ganglia hemorrhage holds important clinical value for the formulation of treatment decisions. In addition, accurate prediction models can also be used for the verification and evaluation of new treatment regimens in clinical trials.\u003c/p\u003e\u003cp\u003eAs a clinician-reported measure of global disability, the modified Rankin scale(mRS) is extensively utilized to assess outcomes in stroke patients and serves as a primary endpoint in randomized clinical trials(8). Its validity and reliability are well-established through multiple lines of evidence(9). This evaluation index focuses on measuring the recovery of patients' activities of daily living (such as self-care, walking, work, etc.) and serves as a reliable, simple, and easy-to-use tool for evaluating stroke prognosis(10). Therefore, in this study, using the 90-day mRS score of patients as the study endpoint can accurately assess the patients' current living status.\u003c/p\u003e\u003cp\u003eIn this study, relevant clinical indicators easily accessible in the emergency setting, such as medical history and laboratory tests, were included, and combined with hematoma-related imaging features obtained from non-contrast cranial CT scans for statistical analysis. Previously, the method of manual measurement combined with hematoma volume calculation formulas exhibited significantly reduced accuracy when assessing irregular hematomas and intraventricular hematomas(11,12). With the development of computer technology, the use of AI algorithms or predefined algorithms to automatically identify hematoma regions and calculate volumes has become the mainstream(13). Although their segmentation accuracy for extremely low-density or mixed-density hematomas may decrease, they still demonstrate high stability and accuracy in the diagnosis and treatment of patients with acute intracerebral hemorrhage.\u003c/p\u003e\u003cp\u003eAfter univariate analysis and screening via multivariate logistic regression, among clinical features, a clinical-only model for predicting the 90-day prognosis of patients with BGH was constructed using neutrophil count, FG, fibrinogen, GCS, blood glucose, white blood cell count, and NLR. For this clinical-only model, the Random Forest (RF) model showed an AUC of 0.86, sensitivity of 0.85, and specificity of 0.73; the Logistic Regression (LR) model had an AUC of 0.85, sensitivity of 0.80, and specificity of 0.80; and the Support Vector Machine (SVM) model exhibited an AUC of 0.83, sensitivity of 0.76, and specificity of 0.77.\u003c/p\u003e\u003cp\u003eBased on imaging features, 8 characteristic parameters were screened, including initial hematoma volume, ratio of hematoma short diameter to long diameter, perihematomal edema volume, midline shift distance, as well as hematoma hypodensity sign, swirl sign, black hole sign, and island sign, to construct an imaging-only model. For this imaging-only model, the RF model demonstrated an AUC of 0.79, sensitivity of 0.65, and specificity of 0.81; the LR model had an AUC of 0.83, sensitivity of 0.80, and specificity of 0.69; and the SVM model showed an AUC of 0.75, sensitivity of 0.58, and specificity of 0.88. The RF model exhibited better predictive performance than other prediction models.\u003c/p\u003e\u003cp\u003eBy integrating the 5 most predictive parameters from clinical and imaging parameters, a clinical-imaging integrated model was constructed. For this integrated model, the RF model showed an AUC of 0.97, sensitivity of 0.86, and specificity of 0.94; the LR model had an AUC of 0.89, sensitivity of 0.85, and specificity of 0.78; and the SVM model exhibited an AUC of 0.88, sensitivity of 0.88, and specificity of 0.74. The predictive performance of the clinical-imaging integrated model was significantly improved compared with the clinical-only model and the imaging-only model. Among them, the RF model performed better, with an AUC of 0.84 in the validation set, indicating that the RF model has stable generalization ability and applicability in predicting the 90-day prognosis of patients with BGH.\u003c/p\u003e\u003cp\u003eAnalysis of feature weights in the clinical-imaging RF model for 90-day prognosis in patients with BGH: Among these features, baseline hematoma volume is the most important. Previous studies have demonstrated that hematoma volume is the most critical factor affecting the prognosis of patients with intracerebral hemorrhage(14). The larger the hematoma volume, the greater the degree of damage to brain tissue, and the more irreversible the impairment of brain function(15).The minor/major axis ratio of the largest axial slice of the hematoma is an imaging feature for evaluating hematoma morphology(16). Hematomas with a larger ratio tend to be round in shape, indicating high intrinsic tension of the hematoma and weak resistance of surrounding brain tissue to hematoma expansion, ultimately leading to expansion in all directions. In contrast, a smaller ratio indicates a spindle-shaped hematoma, which expands along the interstitial spaces of brain tissue with lower tension, resulting in less damage to surrounding brain tissue compared to the high-ratio group(17).Perihematomal edema occurs in the hyperacute phase after intracerebral hemorrhage, appearing as a hypodense zone surrounding the hematoma shortly after its formation(18). It is recognized as an imaging marker of secondary injury following intracerebral hemorrhage, with pathophysiological manifestations including cytotoxic edema, ischemic necrosis, and neuroinflammation(19). This study confirms that the volume of early perihematomal edema is associated with patient prognosis.\u003c/p\u003e\u003cp\u003eThe neutrophil-to-lymphocyte ratio (NLR), as an inflammatory indicator, can reflect the degree of systemic inflammatory response induced by brain tissue damage(20). A high NLR is often accompanied by suppressed lymphocyte function, making patients prone to complications such as pulmonary or systemic multi-organ infections(21), which are important triggers for death in patients with intracerebral hemorrhage(22). Therefore, a significant increase in NLR may indirectly reflect the severity of the disease, thereby affecting patient prognosis.The Glasgow Coma Scale (GCS) is a classic scale for assessing the degree of consciousness disturbance(23), directly reflecting the functional status of the brain after intracerebral hemorrhage(24). Numerous previous studies have confirmed that the GCS score is an independent predictor of 30-day mortality in intracerebral hemorrhage; a lower GCS score indicates a higher risk of death(25).In summary, the multi-factor integrated model combining clinical and imaging indicators can significantly improve the accuracy of prediction.\u003c/p\u003e\u003cp\u003eThe strengths of this study lie in the use of AI-based extraction of imaging features, Artificial intelligence approaches exhibit exceptional proficiency in the automatic identification of intricate patterns within imaging data and the provision of quantitative outputs(26), which offers stability and reproducibility. By integrating clinical risk factors and comparing different machine learning models, the RF model was shown to have superior predictive performance. Results from the validation set also demonstrated that the RF model constructed based on clinical and imaging features has favorable predictive value for the short-term prognosis of patients with BGH. Compared with machine learning models developed in previous studies (3), our model significantly enhanced predictive performance by incorporating imaging features that assess the degree of brain tissue damage caused by intracranial hematoma.\u003c/p\u003e\u003cp\u003eThis study has certain limitations. Firstly, as a retrospective study, its data were derived from previous clinical records. Restricted by the study design, selection bias may have existed during case enrollment and data extraction, which could affect the generalizability of the research results. Additionally, the model in this study was constructed only based on an internal dataset, lacking validation with external independent cohorts. Therefore, future research urgently needs to conduct multicenter, large-sample prospective studies, combined with external validation across multiple regions and institutions, to further evaluate the accuracy of the model.\u003c/p\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eThe RF model established based on clinical and imaging features at admission exhibits reliable predictive performance for the 90-day prognosis of patients with BGH, and holds significant clinical application value as well as practical guiding significance.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTYY and ZZ substantially contributed to the conception and design of the study. YJS and JST performed research and prepared figures. YTY and SKY analyzed and interpreted the data and drafted the article. ZZ made critical revisions related to the important intellectual content of the manuscript and final approval. JST and YCG provided supervision. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by Science Foundation of Kangda College of Nanjing Medical University (KD2024KYJJ176),the Beijing Medical Award Foundation (YXJL-2024-0299-0062), and the Scientific Research Project of Yancheng Municipal Health Commission (YK2024067).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data presented in this study are available upon request from the corresponding author.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Sharing Statement\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe data presented in this study are available on request from the corresponding author.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Institutional Review Board of the Affiliated Binhai Hospital,Kangda College of Nanjing Medical University (ethics numbers: 2025BYKYLL028).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for Publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eInformed consent was obtained from all subjects involved in the study.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDisclosure\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors report no conflicts of in this work.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSheth KN. Spontaneous Intracerebral Hemorrhage. N Engl J Med. 2022 Oct 27;387(17):1589\u0026ndash;96. \u003c/li\u003e\n\u003cli\u003eVandertop WP, Can A, Post R. Spontaneous Intracerebral Hemorrhage. N Engl J Med. 2023 Jan 12;388(2):191\u0026ndash;2. \u003c/li\u003e\n\u003cli\u003eGeng Z, Yang C, Zhao Z, Yan Y, Guo T, Liu C, et al. Development and validation of a machine learning-based predictive model for assessing the 90-day prognostic outcome of patients with spontaneous intracerebral hemorrhage. J Transl Med. 2024 Mar 4;22:236. \u003c/li\u003e\n\u003cli\u003eBae\u0026szlig;ler B, Engelhardt S, Hekalo A, Hennemuth A, H\u0026uuml;llebrand M, Laube A, et al. Perfect Match: Radiomics and Artificial Intelligence in Cardiac Imaging. Circ Cardiovasc Imaging. 2024 Jun;17(6):e015490. \u003c/li\u003e\n\u003cli\u003eLiu D, Zhang G, Wang Y, Li J, Cao P, Yin X, et al. Geometric features of middle cerebral artery are associated with spontaneous basal ganglia intracerebral haemorrhage. Stroke Vasc Neurol. 2022 Mar 9;7(5):399\u0026ndash;405. \u003c/li\u003e\n\u003cli\u003eAl-Kawaz MN, Hanley DF, Ziai W. Advances in Therapeutic Approaches for Spontaneous Intracerebral Hemorrhage. Neurotherapeutics. 2020 Oct;17(4):1757\u0026ndash;67. \u003c/li\u003e\n\u003cli\u003eHe M, Lu Z, Lv Y, Cheng Z, Zhang Q, Jin X, et al. Machine learning-based prediction of 6-month functional recovery in hypertensive cerebral hemorrhage: insights from XGBoost and SHAP analysis. Front Neurol. 2025;16:1608341. \u003c/li\u003e\n\u003cli\u003eJp B, O A, J E. Evolution of the Modified Rankin Scale and Its Use in Future Stroke Trials[J]. Stroke, Stroke, 2017, 48(7).\u003c/li\u003e\n\u003cli\u003eBanks JL, Marotta CA. Outcomes validity and reliability of the modified Rankin scale: implications for stroke clinical trials: a literature review and synthesis. Stroke. 2007 Mar;38(3):1091\u0026ndash;6. \u003c/li\u003e\n\u003cli\u003eHaggag H, Hodgson C. Clinimetrics: Modified Rankin Scale (mRS). Journal of Physiotherapy. 2022 Oct 1;68(4):281. \u003c/li\u003e\n\u003cli\u003eB Z, Wb J, Ly Z, et al. 1/2SH: A Simple, Accurate, and Reliable Method of Calculating the Hematoma Volume of Spontaneous Intracerebral Hemorrhage[J]. Stroke, Stroke, 2020, 51(1).\u003c/li\u003e\n\u003cli\u003eXu X, Chen X, Zhang J, Zheng Y, Sun G, Yu X, et al. Comparison of the Tada formula with software slicer: precise and low-cost method for volume assessment of intracerebral hematoma. Stroke. 2014 Nov;45(11):3433\u0026ndash;5. \u003c/li\u003e\n\u003cli\u003ePetrov A, Kashevnik A, Haleev M, Ali A, Ivanov A, Samochernykh K, et al. AI-Based Approach to One-Click Chronic Subdural Hematoma Segmentation Using Computed Tomography Images. Sensors (Basel). 2024 Jan 23;24(3):721. \u003c/li\u003e\n\u003cli\u003eM L, Z W, X M, Y Z, X H, L L, et al. Predictive Nomogram for Unfavorable Outcome of Spontaneous Intracerebral Hemorrhage[J]. World neurosurgery, World Neurosurg, 2022, 164.\u003c/li\u003e\n\u003cli\u003eKc T, Sm F, Wcy L, Iyh L, Yk W, Omy C, et al. Location-Specific Hematoma Volume Cutoff and Clinical Outcomes in Intracerebral Hemorrhage[J]. Stroke, Stroke,2023, 54(6).\u003c/li\u003e\n\u003cli\u003eYang T, Zhao Z, Gu Y, Yang S, Zhang Y, Li L, et al. Evaluating the value of machine learning models for predicting hematoma expansion in acute spontaneous intracerebral hemorrhage based on CT imaging features of hematomas and surrounding oedema. Front Neurol. 2025;16:1567525. \u003c/li\u003e\n\u003cli\u003eR R, M S, S T, M T, M K, K K, et al. Mechanism of Spontaneous Intracerebral Hemorrhage Formation: An Anatomical Specimens-Based Study[J]. Stroke, Stroke,2022, 53(11).\u003c/li\u003e\n\u003cli\u003eY C, S C, J C, et al. Perihematomal Edema After Intracerebral Hemorrhage: An Update on Pathogenesis, Risk Factors, and Therapeutic Advances[J]. Frontiers in immunology, Front Immunol, 2021, 12.\u003c/li\u003e\n\u003cli\u003eC J, H G, Z Z, et al. Molecular, Pathological, Clinical, and Therapeutic Aspects of Perihematomal Edema in Different Stages of Intracerebral Hemorrhage[J]. Oxidative medicine and cellular longevity, Oxid Med Cell Longev, 2022, 2022.\u003c/li\u003e\n\u003cli\u003eP G, Y L, Y G, et al. The association of neutrophil to lymphocyte ratio, platelet to lymphocyte ratio, and lymphocyte to monocyte ratio with post-thrombolysis early neurological outcomes in patients with acute ischemic stroke[J]. Journal of neuroinflammation, J Neuroinflammation, 2021, 18(1).\u003c/li\u003e\n\u003cli\u003eY Z, W P, X Z. The prognostic value of the combined neutrophil-to-lymphocyte ratio (NLR) and neutrophil-to-platelet ratio (NPR) in sepsis[J]. Scientific reports, Sci Rep, 2024, 14(1).\u003c/li\u003e\n\u003cli\u003eJs B, Am B. Complications of intracerebral haemorrhage[J]. The Lancet. Neurology, Lancet Neurol, 2012, 11(1).\u003c/li\u003e\n\u003cli\u003ek Z, Tu D, G T, et al. The Message of the Glasgow Coma Scale: A Comprehensive Bibliometric Analysis and Systematic Review of Clinical Practice Guidelines Spanning the Past 50 years[J]. World neurosurgery, World Neurosurg, 2024, 185.\u003c/li\u003e\n\u003cli\u003eA L, Fd A, Am V, et al. Admission Glasgow Coma Scale Score as a Predictor of Outcome in Patients Without Traumatic Brain Injury[J]. American journal of critical care : an official publication, American Association of Critical-Care Nurses, Am J Crit Care, 2021, 30(5).\u003c/li\u003e\n\u003cli\u003eCw W, Yj L, Yh L, et al. Hematoma shape, hematoma size, Glasgow coma scale score and ICH score: which predicts the 30-day mortality better for intracerebral hematoma?[J]. PloS one, PLoS One, 2014, 9(7).\u003c/li\u003e\n\u003cli\u003eBs K, C J, Sm B, et al. Radiology artificial intelligence: a systematic review and evaluation of methods (RAISE)[J]. European radiology, Eur Radiol, 2022, 32(11).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Basal ganglia hemorrhage, CT scan, Machine learning, Prognosis","lastPublishedDoi":"10.21203/rs.3.rs-7363416/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7363416/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eObjective\u003c/h2\u003e\u003cp\u003eThis study aims to investigate the predictive value of machine learning models integrating initial computed tomography (CT) imaging features and admission clinical parameters for 90-day functional outcomes in patients with spontaneous basal ganglia hemorrhage (BGH). The objective is to establish a clinically applicable tool for early identification of high-risk populations and guide personalized intervention strategies.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003ePatients with BGH admitted to Affiliated Binhai Hospital,Kangda College of Nanjing Medical University from January 2022 to December 2024 were retrospectively collected. Their clinical and imaging data at admission were gathered. The patients were divided into a training set and a test set in a 7:3 ratio. Univariate analysis and multivariate Logistic regression were applied to screen for risk factors. Based on the screened variables, three machine learning algorithms, namely LogisticRegression (LR), RandomForest (RF), and Support Vector Machine (SVM), were used to construct a clinical-only model, an imaging-only model, and a clinical-imaging integrated model through 5-fold cross-validation, aiming to predict the 90-day prognosis of patients with BGH. The predictive efficacy of the models was evaluated using receiver operating characteristic (ROC) curves and the area under the curve (AUC).\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eBased on clinical and imaging features, we constructed 9 machine learning models. Among them, the clinical-imaging RF model showed an AUC of 0.97, a sensitivity of 0.86, and a specificity of 0.94; the LR model had an AUC of 0.89, a sensitivity of 0.85, and a specificity of 0.78; and the SVM model presented an AUC of 0.88, a sensitivity of 0.88, and a specificity of 0.74. These results indicated that the RF model had superior predictive performance. In addition, compared with the clinical RF model (AUC 0.86, sensitivity 0.85, specificity 0.73) and the imaging RF model (AUC 0.79, sensitivity 0.65, specificity 0.81), the predictive performance of the clinical-imaging RF model was significantly improved. A feature map was used to clarify the importance of variables in the optimal model. The AUC of the clinical-imaging RF model reached 0.84 in the test set, suggesting that this model has stability.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e\u003cp\u003eThe RF model established in this study based on baseline hematoma volume at admission, minor/major axis ratio of the largest axial slice of the hematoma, perihematomal edema volume, neutrophil-to-lymphocyte ratio (NLR), and Glasgow Coma Scale (GCS) score exhibits reliable predictive performance for the 90-day prognosis of patients with BGH. It holds significant clinical application value and practical guiding significance.\u003c/p\u003e","manuscriptTitle":"Machine Learning Integration of Baseline CT Imaging and Clinical Parameters Predicts 90-Day Functional Outcomes in Spontaneous Intracerebral Hemorrhage","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-15 17:00:54","doi":"10.21203/rs.3.rs-7363416/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-11-06T04:35:40+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-24T15:47:29+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-24T15:41:49+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-20T08:24:08+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-13T12:46:31+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-09T06:49:27+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"261251289253322743482733259737190992652","date":"2025-10-07T04:10:56+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"122588888320526454934903380668837951707","date":"2025-10-06T01:42:23+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"162916839339458107683425189224807746374","date":"2025-10-05T02:18:10+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"246989813351161090163825948218123037025","date":"2025-10-04T23:55:32+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-03T15:46:58+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"136959520480052030457882381609224671685","date":"2025-10-02T21:32:57+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"200689043626429597304919216106339997977","date":"2025-10-02T21:11:28+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"31452103197475840237131644533074536098","date":"2025-10-02T19:51:56+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-10-02T19:45:45+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-08-14T12:19:17+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-08-14T12:18:59+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Imaging","date":"2025-08-13T09:28:33+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"3cdc40a0-05df-4c41-bfaf-44dde15e511d","owner":[],"postedDate":"October 15th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-01-21T11:40:21+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-15 17:00:54","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7363416","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7363416","identity":"rs-7363416","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}