Functional Outcome Prediction Across Multiple Timescales Intracerebral Hemorrhage Using a Radiomics Model

Functional Outcome Prediction Across Multiple Timescales Intracerebral Hemorrhage Using a Radiomics Model | 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 Article Functional Outcome Prediction Across Multiple Timescales Intracerebral Hemorrhage Using a Radiomics Model Jianmin Liu, Dan Wang, Jing Zhang, Shenglin Li, Guojin Zhang, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7191002/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Purpose To establish an outcome prediction model for multiple timescales itracerebral hemorrhage (ICH). Materials and methods ICH patients with an onset-to-imaging time (OIT) of less than 72 h were retrospectively collected. Patients were divided into three groups according to their OIT. Group 1–3: OIT < 6h, 6 ≤ OIT<24h, 24 ≤ OIT<72h. The first preoperative review in Group 1 within 72 h were recorded for Group 4. A binary logistic regression classifier was used to construct outcome prediction models for each group. The predictive performance of each group’s model was compared with the application of the Group 1 model across different groups to explore its accuracy and applicability in predicting the outcomes for patients in various groups. Results A total of 2,136 patients with ICH were included in the study. In the training set, the AUC value obtained by directly applying the group 1 radiomics model to group 2 patients was 0.773. The AUC value for the direct application of the Group 1 combined model to Group 3 patients was 0.811. The AUC for applying the Group 1 radiomics model directly to Group 4 patients was 0.779. The AUC of the radiomics model, constructed by combining the key radiomics features of Groups 1 and 4, was 0.788. The AUC of the independent radiomics model for Group 4 was 0.815. Conclusion The radiomics and combined models for predicting outcomes in ICH patients with OIT < 6 h can be applied to all patients with OIT < 72 h, allowing for early and accurate outcome prediction across multiple timescales. Health sciences/Medical research Health sciences/Neurology Intracerebral Hemorrhage Radiomics Outcome Prediction Timescales Non-contrast Computed Tomography Figures Figure 1 Figure 2 Figure 3 Figure 4 1. Introduction Intracerebral hemorrhage (ICH) is the most fatal type of stroke[ 1 – 3 ]. The effectiveness of treatment for ICH is highly time-dependent. The phrase “Time is brain” applies to ICH as well, as a large number of neurons can be damaged in the short period between the initial onset and hematoma expansion [ 4 ]. Unlike patients with ischemic stroke, patients with ICH show little or no improvement in the near term due to a lack of effective treatment options [ 5 ]. Thus, accurate prediction of the outcome of ICH remains an unmet need [ 6 ] and is a pressing concern for physicians [ 7 ]. Hematoma expansion is an independent risk factor for poor ICH[ 8 ]. More than 30% of hematoma expansions occur within 24 h after the onset of symptoms[ 9 , 10 ]. Therefore, radiological studies related to ICH outcomes have focused on most patients within 24 h of onset-to-imaging time (OIT). In the real world, OIT is usually more than 6 h in some ICH patients due to factors such as low public awareness, socioeconomic status, or limited medical resources [ 11 ]. The outcome status of this group cannot be ignored. Recent advances in radiomics have shown promise in the field of ICH outcome prediction[ 12 , 13 ]. Our previous non-contrast computed tomography (NCCT)-based radiomics studies have also achieved positive results [ 14 ]. However, it remained uncertain whether reexamination data from the same patient could provide more information and whether the key outcome features identified in patients with OIT 6 h. In this study, we will compare the radiomics model of the first examination of patients with OIT < 6 hours with the radiomics model of the first preoperative follow-up examination within 72 h. Observe whether the reexamination data can provide more outcome information. Second, we applied the outcome prediction model for patients with OIT < 6 h to ICH patients admitted with different time windows and compared it with the outcome prediction model specific to patients with ICH with different time windows to analyze the applicability and accuracy of the OIT 6 h. To develop a model that accurately stratifies ICH patient outcome risks across multiple timescales. 2. Materials and methods This retrospective study was approved by the ethics committees of three medical institutions. Ethical approval was obtained from the Center 1-Lanzhou University Second Hospital, Center 2-Gansu Provincial Hospital, and Center 3-Xi’an Central Hospital institutional review boards (Approval number: 2022A-096, 2022–275 and LW-2022–011). As the study was a retrospective study, informed consent was waived by the Ethics Committee. This retrospective study collected NCCT data, clinical data, and laboratory data from ICH patients with OIT < 72 h after admission to Center 1(January 1, 2016, to October 1, 2020), Center 2, and Center 3(June 1, 2021, to November 30, 2021), and for patients with OIT < 6 h for their first preoperative review at Center 1 and their NCCT data. Clinical data included OIT, age, sex, admission systolic blood pressure (SBP), and Glasgow coma scale (GCS) score; GCS < 9 was defined as a low score, GCS ≥ 9 was defined as a high score scale. The Modified Rankin Scale (MRS) 90 days post-symptom onset was used to evaluate the outcome. A score of 0–3 was defined as good outcome, 4–6 was poor outcome. Laboratory data included glucose (GLU), triglyceride (TG), white blood cell count (WBC), etc. The patients were divided into three groups based on OIT: Group 1(OIT < 6 h), Group 2(6 OIT < 24 h), and Group 3(24 OIT < 72 h). Group 4 included NCCT data of patients with OIT 18 years, baseline NCCT scan within 72 h of OIT, and complete clinical and laboratory data after admission. Exclusion criteria refer to previous studies[ 14 ]. Patients from Group 1–3 were further divided based on their treatment center: training set (C1: January 1, 2016, to December 31, 2018), internal validation set (C1: January 1, 2019, to October 1, 2020), and external validation set (C2 + C3). 2.1 Feature Selection NCCT parameters, segmentation of regions of interest, and radiomic feature extraction procedure are detailed in the supplementary materials. Clinical features (include clinical, radiological, and laboratory data) were initially screened using one-way analysis, Kruskal-Wallis test, Chi-Square Test, and Mann-Whitney U test(P < 0.05). Subsequently, multiple logistic regression was employed to identify factors independently associated with outcome, retaining features with P < 0.05. Radiomics features were screened using a two-step process. First, Mann–Whitney U was used to screen the radiomics features related to the outcome, and the features with P < 0.01 were retained. Second, multiple logistic regression and stepwise regression analyses were used to screen the key radiomic features independently related to the outcome, and the final key features with P < 0.05 were retained. 2.2 Model construction and validation Clinical and radiomics features were analyzed by univariate and multivariate analyses. Logistic regression was then used to construct outcome models for Groups 1, 2, 3, and 4, resulting in Models 1, 2, 3, and 4, respectively. Second, we applied the prediction models (Models 1–4) and compared their performance with Model 4 to assess whether the inclusion of NCCT radiomics features can provide additional information for outcome prediction and further improve prediction accuracy. Last, involved applying the radiomics feature in Model 1 directly to Groups 2 and 3, followed by comparing their predictive performance to determine the feasibility and accuracy of Model 1 for different OIT groups. The detailed model exploration and construction are shown in Fig. 1 . The prediction model based on key radiomics features from Group 1 was named Rad.G1, while the Clinical model (based on clinical features) was named Clinical.G1. The comprehensive prediction model incorporating radiomics and clinical features was named COMB.G1. Similarly, the radiomics, clinical, and comprehensive model were named Rad.G2, Clinical.G2, and COMB.G2, respectively. For Group 3, the radiomic model was named Rad.G3, the clinical model was named Clinical.G3, and the comprehensive model was named COMB.G3. The radiomics model for Group 4 was named Rad.G4. To determine whether key radiomics features from preoperative review NCCT provide additional information for outcome prediction, we conducted the following analysis: ①The Rad.G1 model was directly applied to Group 4 patients, denoted as Rad.G1.G4. Similarly, COMB.G1 and Clinical.G1 were applied to Group 4, denoted as COMB.G1.G4 and Clinical.G1.G4; ② Radiomics features from the Rad.G1 model were combined with those extracted from Group 4, and the combined features were refined using stepwise regression. The final key features, identified through screening were labeled as Rad.G1wG4 for constructing the outcome prediction model; ③Independent clinical features were combined with Rad.G1wG4 features to construct the comprehensive Group 4 outcome model, COMB.G1wG4;④The predictive performance of Rad.G1.G4, Rad.G1wG4, Rad.G4 was compared across multiple dimensions to assess the added value of key radiomics features from NCCT data in predicting outcomes. To assess the applicability of the Group 1 outcome prediction model in predicting outcomes for ICH patients across different OIT groups, we conducted the following experiments: ①The models Rad.G1, Clinical.G1, and COMB.G1 were applied to Group 2 patients, denoted as Rad.G1.G2, Clinical.G1.G2, and COMB.G1.G2. Simultaneously, the accuracy of the outcome prediction models Rad.G2, Clinical.G2, and COMB.G2, which were constructed using data from Group 2, was compared and analyzed across multiple dimensions to assess the applicability of the Group 1 model in Group 2. The optimal model from Group 1, COMB.G1, was then applied to Group 3 and is denoted as COMB.G1.G3. Additionally, multi-model and multi-dimensional comparisons were conducted using Rad.G3, Clinical.G3, and COMB.G3 models constructed using the data of patients in Group 3 to evaluate the applicability of the Group 1 model in Group 3. 2.3 Statistical Analyses Continuous variables are reported as medians and interquartile ranges (IQR), while categorical variables are presented as numbers and percentages. Univariate analysis included one-way ANOVA, Kruskal-Wallis test, chi-square test, and Mann-Whitney U test. A significance threshold of p < 0.05 was used, with a stricter threshold of p < 0.01 for radiomic feature screening. Multivariate logistic regression was employed to identify factors independently associated with functional outcomes, and forward-backward stepwise regression was used to refine radiomics features (p < 0.05). Model performance was evaluated using receiver operating characteristic (ROC) analysis, with accuracy assessed by the area under the curve (AUC), accuracy (ACC), sensitivity (SEN), specificity (SPE), positive predictive value (PPV), and negative predictive value (NPV). The DeLong test revealed significant differences in AUC between models (p < 0.05). The Hosmer-Lemeshow test was used to assess the goodness of fit between the model’s predictions and actual outcomes, and decision curve analysis (DCA) evaluated the clinical utility of the model. Statistical analyses were performed using SPSS version 25 (IBM, Armonk, NY, USA) and R (version 4.1.0; R Foundation for Statistical Computing, Vienna, Austria). 3. Results 3.1 Characteristics of the Study Sample A total of 2,136 patients with ICH were included from three medical centers. The number of patients in groups 1–4 was 1,098, 660, 378 and 584. There were 1,235 male patients (58%) and 901 female patients (42%). The median patient age was 59 years, and the median OIT was 5 h. In groups 1–3, Group 3 patients had the lowest rate of poor outcomes, with rates of 703(64%), 384(58%), and 162(43%) patients in the three groups, respectively. The high score rate of GCS score of group 3 was also higher compared to groups 1 and 2 by 83%, 64%, and 72%, respectively (Table 1 ). Table 1 Baseline Patient Characteristics Parameter Group Group 1 Group 2 Group 3 Group 4 All Data n = 1098 n = 660 n = 378 n = 584 n = 2136 Poor outcome* 703 (64) 384 (58) 162 (43) 346 (59) 1247 (58) OIT (h) 3.0 (1.5, 4.0) 10.0 (7.0, 13.0) 31.6 ± 11.2 3.0 (1.5, 4.0) 5.0(3.0, 13.0) Gender* Male 652 (59) 372 (56) 211 (56) 345 (59) 1235(58) Female 446 (41) 288 (44) 167 (44) 239 (41) 901(42) Age (y) 60 (52, 70) 60 (51, 69) 57 (50, 67) 59 (51, 68) 59 (51, 69) Location* Deep 977 (89) 545 (83) 281 (74) 534 (91) 1803 (85) Lobar 121 (11) 115 (17) 97 (26) 50 (9) 333 (15) Midline shift* 282 (26) 132 (20) 47 (12) 83 (14) 461 (22) IVH* 546 (50) 317 (48) 137 (36) 249 (43) 1001 (47) SAH* 208 (19) 131 (20) 55 (15) 58 (10) 394 (18) Hypodensities* 729 (66) 382 (58) 223 (59) 379 (65) 1334 (63) ICH volume (ml) 31.1 (13.9, 70.5) 33.2 (14.9, 60.8) 23.2 (10.4, 40.7) 24.7 (13.4, 44.9) 29.8 (13.2, 60.9) PHE volume (ml) 11.2 (4.9, 23.4) 13.3 (5.9, 24.9) 12.6 (5.7, 24.3) 8.9 (4.5, 16.6) 12.2 (5.3, 23.9) Temperature (℃) 36.6 (36.4, 36.8) 36.6 (36.5, 36.8) 36.6 (36.5, 36.8) 36.6 (36.5, 36.8) 36.6 (36.5, 36.8) Smoking* 189 (17) 87 (13) 31 (8) 60 (10) 308 (14) SBP (mmHg) 172(153, 191) 162 (145, 179) 159 (143, 176) 170 (153, 189) 166 (148, 186) GCS* Low (score<9) 399 (36) 185 (28) 63 (17) 147 (25) 647 (30) High (score ≥ 9) 699 (64) 475 (72) 315 (83) 437 (75) 1489 (70) GLU (mmol/L) 7.90 (6.40, 9.89) 7.55 (6.50, 9.39) 6.73 (5.63, 7.90) 7.48 (6.30, 9.30) 7.50 (6.25, 9.34) TG (mmol/L) 1.26 (0.99, 1.99) 0.95 (0.64, 1.45) 1.08 (0.80, 1.88) 1.26 (0.76, 1.98) 1.10 (0.74, 1.84) WBC (10^9/L) 8.61 (6.38, 11.79) 9.22 (7.04, 11.97) 8.74 (6.76, 11.46) 8.20 (6.53, 10.60) 8.84 (6.66, 11.72) NE (10^9/L) 6.94 (4.46, 10.10) 7.93 (5.62, 10.70) 7.10 (5.08, 9.86) 6.53 (4.25, 9.33) 7.25 (4.85, 10.24) LY (10^9/L) 1.04 (0.70, 1.60) 0.84 (0.59, 1.18) 1.04 (0.71, 1.47) 1.00(0.69, 1.51) 0.96 (0.67, 1.42) NLR 6.25 (3.35, 12.25) 9.89 (5.39, 15.97) 6.75 (3.98, 11.87) 6.26 (3.38, 11.77) 7.55 (3.96, 13.37) HGB (g/L) 148 (136, 161) 150 (137, 164) 151 (137, 164) 148 (136, 159) 149 (137, 162) INR 0.99 (0.94, 1.05) 1.00 (0.94, 1.05) 1.01 (0.94, 1.05) 0.98 (0.93, 1.04) 0.99 (0.94, 1.05) Note.— Unless otherwise indicated, data are presented as median (interquartile range) and data in parentheses are percentages. OIT = onset-to-imaging time, IVH = Intraventricular hemorrhage, SAH = Subarachnoid hemorrhage, SBP = Systolic blood pressure, GCS = Glasgow coma scale, GLU = Glucose, TG = Triglyceride, WBC = White blood cell, NE = Neutrophil, LY = Lymphocyte, NLR = Neutrophil-lymphocyte ratio, HGB = Hemoglobin, INR = International normalized ratio * Data are percentages n (%). The incidence of positive hypodensity was higher in Group 1 (approximately 66%), while intraventricular hemorrhage and midline shift also had the highest incidences in the group with 50% and 26% respectively. The median hematoma volumes for Groups 1, 2, and 3 were 31.1 ml, 33.2 ml, and 23.2 ml, respectively. After performing univariate and multivariate logistic regression analyses on all clinical features in Group 2 and Group 3. The age, hematoma volume, GCS score, WBC count, and hematoma location were independent independently associated with outcome in Group 2. In Group 3, the independent risk factors were age, hematoma volume, GCS score, and hematoma location. 3.2 Radiomics feature selection and model construction 12 features were included in Group 1[ 14 ], 8 in Group 2, 3 in Group 3, and 8 in Group 4. Seven key features were retained after the screening via stepwise forward-backward regression for Group 1, combined with similarly named radiomics features from Group 4(Supplementary Tables 1). 3.3 Models performance The AUC value of Rad.G1.G4 in the training set was 0.779. The AUC value of COMB.G1.G4 were 0.864. The AUC and ACC value of Rad.G1wG4 were 0.788 and 0.726; Rad.G4 were 0.815and 0.768, Table 2 and Supplementary Fig. 2. The Delong test comparing the AUCs of different models showed that the training cohort was the p-value for Rad.G1.G4 vs. Rad.G4 significant at 0.04, while all other comparisons had p > 0.05(Table 3 ). In addition, the direct application of COMB.G1.G4 maintained high predictive performance and good stability in the three cohorts, with AUC values of 0.864, 0.804, and 0.800. Table 2 Performance of each model in three cohort Model AUC ACC SEN SPE Cohort Rad.G1.G4 0.779(0.727–0.830) 0.729(0.677–0.778) 0.741(0.619–0.810) 0.712(0.552–0.792) Training cohort Rad.G1wG4 0.788(0.737–0.838) 0.726(0.673–0.775) 0.688(0.518–0.788) 0.784(0.664–0.856) Rad.G4 0.815(0.768–0.862) 0.768(0.717–0.813) 0.862(0.714–0.910) 0.624(0.472–0.704) Clinical.G1.G4 0.828(0.784–0.872) 0.729(0.677–0.778) 0.698(0.603–0.778) 0.776(0.640–0.856) COMB.G1.G4 0.864(0.825–0.903) 0.768(0.717–0.813) 0.683(0.571–0.767) 0.896(0.792–0.952) COMB.G1wG4 0.872(0.834–0.909) 0.796(0.747–0.839) 0.757(0.672–0.826) 0.856(0.752–0.928) Rad.G1.G4 0.721(0.659–0.784) 0.639(0.577–0.698) 0.653(0.551–0.823) 0.620(0.500-0.769) Internal validation cohort Rad.G1wG4 0.751(0.691–0.812) 0.710(0.650–0.765) 0.673(0.422–0.755) 0.759(0.583–0.843) Rad.G4 0.740(0.679-0.800) 0.682(0.621–0.739) 0.810(0.680–0.912) 0.509(0.407–0.621) Clinical.G1.G4 0.790(0.734–0.845) 0.686(0.625–0.743) 0.633(0.537–0.816) 0.759(0.667–0.870) COMB.G1.G4 0.804(0.751–0.857) 0.706(0.646–0.761) 0.612(0.469–0.748) 0.833(0.759–0.898) COMB.G1wG4 0.800(0.745–0.854) 0.698(0.638–0.754) 0.612(0.496–0.748) 0.815(0.750–0.908) Rad.G1.G4 0.560(0.179–0.941) 0.533(0.266–0.787) 0.600(0.300-1.000) 0.400(0.000–1.000) External validation cohort Rad.G1wG4 0.600(0.233–0.967) 0.667(0.384–0.882) 0.600(0.000-0.900) 0.800(0.200-1.000) Rad.G4 0.700(0.324-1.000) 0.733(0.449–0.922) 1.000(0.798-1.000) 0.200(0.000–1.000) Clinical.G1.G4 0.820(0.587-1.000) 0.800(0.519–0.957) 1.000(0.600-1.000) 0.400(0.000–1.000) COMB.G1.G4 0.800(0.560-1.000) 0.600(0.323–0.837) 0.700(0.400-1.000) 0.400(0.200-1.000) COMB.G1wG4 0.720(0.342-1.000) 0.800(0.519–0.957) 1.000(0.598-1.000) 0.400(0.195-1.000) Table 3 Delong test of model AUC area Name P value Cohort Rad.G1.G4-Rad.G1wG4 0.50 Training cohort Rad.G1.G4-Rad.G4 0.04 Rad.G1wG4-Rad.G4 0.09 COMB.G1.G4-COMB.G1wG4 0.36 Rad.G1.G4-Rad.G1wG4 0.11 Internal validation cohort Rad.G1.G4-Rad.G4 0.40 Rad.G1wG4-Rad.G4 0.57 COMB.G1.G4-COMB.G1wG4 0.76 Rad.G1.G4-Rad.G1wG4 0.60 External validation cohort Rad.G1.G4-Rad.G4 0.20 Rad.G1wG4-Rad.G4 0.19 COMB.G1.G4-COMB.G1wG4 0.64 The Hosmer-Lemeshow test showed that in both the training and internal validation cohorts, all models closely followed the diagonal, representing the predicted probability of an ideal model Fig. 2 A-C. Decision curves analysis (DCA) for the training and internal validation cohort showed a good net benefit in predicting functional outcomes of ICH patients (Fig. 2 D-F). The AUC values for Rad.G1.G2 and Rad. G2 were 0.773 and 0.795 in training cohort (Table 4 and Supplementary Fig. 3). The Delong test revealed no significant difference in the AUCs for Rad.G1.G2 vs. Rad.G2 across the three cohorts. After adding clinical and laboratory-independent risk factors, the AUC values of COMB.G1.G2 vs. COMB.G2 were still not statistically significant Supplementary Table 2. Table 4 Performance of each model in three cohort Model AUC ACC SEN SPE PPV NPV Cohort Rad.G1.G2 0.773(0.723–0.822) 0.747(0.699–0.792) 0.752(0.614–0.819) 0.740(0.514–0.808) 0.806(0.772–0.819) 0.675(0.591–0.694) Training cohort Clinical.G1.G2 0.859(0.821–0.898) 0.803(0.758–0.843) 0.776(0.619–0.829) 0.842(0.698–0.897) 0.876(0.850–0.883) 0.724(0.685–0.736) COMB.G1.G2 0.860(0.822–0.899) 0.795(0.749–0.836) 0.786(0.695–0.838) 0.808(0.698–0.884) 0.855(0.839–0.863) 0.724(0.694–0.741) Rad.G2 0.786(0.739–0.833) 0.728(0.678–0.773) 0.671(0.566–0.729) 0.808(0.636–0.877) 0.834(0.809–0.845) 0.631(0.574–0.650) Clinical.G2 0.863(0.825-0.900) 0.787(0.740–0.828) 0.762(0.657–0.829) 0.822(0.692–0.884) 0.860(0.841–0.870) 0.706(0.669–0.721) COMB.G2 0.869(0.832–0.906) 0.792(0.746–0.833) 0.752(0.638–0.824) 0.849(0.740–0.911) 0.878(0.859–0.887) 0.705(0.675–0.719) Rad.G1.G2 0.795(0.742–0.849) 0.714(0.656–0.767) 0.743(0.651–0.868) 0.675(0.590–0.812) 0.748(0.723–0.776) 0.669(0.639–0.709) Internal validation cohort Clinical.G1.G2 0.866(0.825–0.908) 0.762(0.707–0.812) 0.724(0.625–0.836) 0.812(0.692–0.923) 0.833(0.812–0.852) 0.693(0.659–0.720) COMB.G1.G2 0.872(0.831–0.912) 0.796(0.742–0.842) 0.783(0.657–0.855) 0.812(0.675–0.881) 0.844(0.820–0.855) 0.742(0.705–0.757) Rad.G2 0.812(0.761–0.864) 0.743(0.687–0.795) 0.704(0.586–0.796) 0.795(0.658–0.889) 0.817(0.788–0.835) 0.674(0.631–0.698) Clinical.G2 0.875(0.835–0.915) 0.773(0.718–0.822) 0.730(0.651–0.829) 0.829(0.743–0.949) 0.847(0.832–0.863) 0.703(0.680–0.730) COMB.G2 0.883(0.844–0.921) 0.773(0.718–0.822) 0.717(0.632–0.816) 0.846(0.752–0.957) 0.858(0.842–0.873) 0.697(0.672–0.723) Rad.G1.G2 0.706(0.528–0.884) 0.600(0.421–0.761) 0.471(0.118–0.941) 0.722(0.500-0.944) 0.615(0.286–0.762) 0.591(0.500-0.654) External validation cohort Clinical.G1.G2 0.837(0.683–0.990) 0.800(0.631–0.916) 0.824(0.471-1.000) 0.778(0.054-1.000) 0.778(0.667–0.810) 0.824(0.245–0.857) COMB.G1.G2 0.866(0.744–0.988) 0.771(0.599–0.896) 0.824(0.587-1.000) 0.722(0.499–0.946) 0.737(0.666–0.773) 0.812(0.749–0.850) Rad.G2 0.807(0.661–0.954) 0.714(0.537–0.854) 0.412(0.353–0.824) 1.000(0.611-1.000) 1.000(1.000–1.000) 0.643(0.524–0.643) Clinical.G2 0.879(0.749-1.000) 0.829(0.664–0.934) 0.882(0.706-1.000) 0.778(0.054-1.000) 0.789(0.750–0.810) 0.875(0.328-0.900) COMB.G2 0.915(0.805-1.000) 0.857(0.697–0.952) 0.824(0.587-1.000) 0.889(0.667-1.000) 0.875(0.833–0.895) 0.842(0.800-0.857) The Hosmer–Lemeshow test calibration curves indicated that the predicted probabilities of each model in the training and internal validation cohorts closely aligned with actual probabilities (Fig. 3 A-C). DCA showed that in the training cohorts, both COMB.G1.G2 and COMB.G2 achieved high net benefits (Fig. 4 A-C). Rad. G3 performed at a moderate level in the internal validation cohorts, with AUC values of 0.717. COMB.G3 had a higher classification performance in the training cohort, with AUC values of 0.817. COMB.G1.G3 exhibited classification performance comparable to COMB.G3 for predicting ICH outcomes. The other parameters for each model are shown in Table 5 and Supplementary Figs. 3. The Delong test showed no significant difference in the AUCs for COMB.G1.G3 vs. COMB.G3 across the three cohorts. The Hosmer-Lemeshow test calibration curves demonstrated that COMB.G1.G3 and COMB.G3 were the models closest to the actual prediction probability in the internal validation cohort (Fig. 3 D-f). The DCA indicated that COMB.G1.G3 achieved high clinical net benefits within the threshold range of 0.3–0.9 in the training cohort (Fig. 4 D-f). Table 5 Performance of each model in three cohort Model AUC ACC SEN SPE PPV NPV Cohort COMB.G1.G3 0.811(0.755–0.867) 0.781(0.722–0.832) 0.629(0.467–0.724) 0.906(0.711–0.961) 0.846(0.803–0.864) 0.748(0.700-0.759) Training cohort Rad.G3 0.678(0.609–0.746) 0.657(0.592–0.717) 0.562(0.410–0.638) 0.734(0.555–0.828) 0.634(0.558–0.663) 0.671(0.607–0.697) Clinical.G3 0.814(0.759–0.869) 0.768(0.709–0.821) 0.686(0.533–0.762) 0.836(0.633–0.922) 0.774(0.727–0.792) 0.764(0.710–0.781) COMB.G3 0.817(0.763–0.872) 0.768(0.709–0.821) 0.648(0.514–0.733) 0.867(0.703–0.953) 0.800(0.760–0.819) 0.750(0.709–0.767) COMB.G1.G3 0.847(0.780–0.914) 0.792(0.712–0.858) 0.566(0.340–0.699) 0.948(0.805-1.000) 0.882(0.818–0.902) 0.760(0.729–0.770) Internal validation cohort Rad.G3 0.717(0.627–0.806) 0.692(0.605–0.770) 0.547(0.377–0.736) 0.792(0.610–0.909) 0.644(0.556–0.709) 0.718(0.662–0.745) Clinical.G3 0.838(0.766–0.909) 0.769(0.687–0.839) 0.679(0.528–0.793) 0.831(0.649–0.948) 0.735(0.683–0.764) 0.790(0.746–0.811) COMB.G3 0.854(0.790–0.918) 0.777(0.696–0.845) 0.623(0.472–0.736) 0.883(0.714–0.987) 0.786(0.735–0.812) 0.773(0.733–0.792) COMB.G1.G3 0.944(0.813-1.000) 0.867(0.595–0.983) 0.667(0.325-1.000) 0.917(0.750-1.000) 0.667(0.494–0.750) 0.917(0.900-0.923) External validation cohort Rad.G3 0.972(0.895-1.000) 0.933(0.681–0.998) 0.667(0.333-1.000) 1.000(0.833-1.000) 1.000(1.000–1.000) 0.923(0.909–0.923) Clinical.G3 0.944(0.823-1.000) 0.933(0.681–0.998) 1.000(0.000–1.000) 0.917(0.750-1.000) 0.750(0.000-0.750) 1.000(1.000–1.000) COMB.G3 0.944(0.823-1.000) 0.867(0.595–0.983) 0.667(0.000–1.000) 0.917(0.750-1.000) 0.667(0.000-0.750) 0.917(0.900-0.923) 4. Discussion Our study developed and validated a radiomics model for predicting 90-day functional outcomes in patients with ICH. Not only can the model accurately predict functional outcomes for ICH patients with OIT < 6 h, but by exploring its feasibility across different timescales. This fills a gap in the literature regarding ICH prognosis. The results showed that Group 3 patients were the latest to be admitted to hospital in groups 1–3, but had the lowest rate of poor prognosis in the three groups. These findings contradict the principle of “Time is brain.” Further analysis revealed that the age and hematoma volume of patients in Group 3 were the lowest among the three groups. The incidence of high GCS scores upon admission was the highest. Due to the relatively small amount of bleeding, these patients exhibited mild clinical symptoms with minimal neurological dysfunction. These patients may present only with persistent headaches. Moreover, due to an insufficient understanding of stroke symptoms by patients or their families, stroke attacks are often not identified in a timely manner, leading to delayed admission. Nevertheless, it is important to note that up to 40% of ICH patients in this group still have a poor prognosis. Therefore, accurate stratification of prognosis in this subset of patients is crucial for reducing the disease burden of ICH. The radiomics model constructed from the first NCCT scan of patients with OIT < 6 h was directly applied to follow-up patients and achieved good performance in prognosis prediction. This result indicates that a review of NCCT data does not provide additional value for predicting the prognosis of patients with ICH. This suggests that the COMB.G1 model for hyperacute phase ICH, developed in previous studies, can accurately predict the 90-day functional outcomes in ICH patients. The key radiomic features of Rad.G1 reflect critical pathophysiological changes that impact prognosis, key molecular expressions related to outcomes, and inflammatory mechanisms. By comparing clinical and laboratory outcome-independent risk factors, it was found that elevated WBC count in Group 2 was an independent risk factor affecting prognosis. Increasing evidence [ 15 , 16 ] indicates that inflammatory injury plays a key role in ICH-induced secondary brain injury and that the inflammatory response is an important defense mechanism following cerebral hemorrhage. When intracerebral hemorrhage occurs, blood components, including red and WBC, macrophages, and plasma proteins such as thrombin, enter the brain parenchyma, triggering an inflammatory response[ 17 , 18 ]. Leukocyte activation is most active within the first 24 h after ICH onset, marked by the accumulation and activation of inflammatory cells. The exudation of white blood cells is central to the inflammatory process, followed by the lysis of red blood cells, which release cytotoxic substances like Hb, heme, and iron, further exacerbating brain damage. The inflammatory response following a hemorrhage aggravates cerebral hemorrhage-induced brain injury, leading to neuronal injury, destruction of the blood-brain barrier, and widespread brain cell death [ 19 ]. This may be the main reason why an elevated WBC count was an independent risk factor for poor prognosis in Group 2. In group 2, after combining clinical and laboratory factors, the AUC values for COMB.G1.G2 and COMB.G2 exceeded 0.85 in all three cohorts, with no statistical difference between the two models. This indicates that the comprehensive model with OIT < 6 h is also applicable to patients with 6 ≤ OIT<24 h. Calibration and DCA showed that the COMB.G1.G2 model had good accuracy and clinical utility in Group 2 patients. This result demonstrates that applying the Group 1 model to predict the functional prognosis of patients with ICH in Group 2 can achieve high predictive accuracy. Specifically, the 90-day outcome prediction model constructed for ICH patients with OIT < 6 h can also be applied to those with OIT < 24 h with good accuracy and stability. When comparing the applicability of the comprehensive model, COMB.G1 to Group 3, it was found that COMB.G1.G3 and COMB.G3 achieved similar predictive accuracies. In the training cohort, the AUC values were 0.811 for COMB.G1.G3 and 0.817 for COMB.G3. Notably, COMB.G1.G3 demonstrated higher accuracy and specificity than COMB.G3. In both Groups 1 and 3, the same clinical and laboratory risk factors were independently associated with outcome. Calibration and DCA showed that COMB.G1.G3 had good accuracy and clinical utility in Group 3 patients. These results suggest that the Group 1 model can also be used to predict the functional prognosis of ICH patients in Group 3. In other words, the 90-day outcome prediction model constructed for ICH patients with OIT < 6 h can also be applied to 24 ≤ OIT<72h ICH patients with good accuracy and stability. This study had several limitations. First, it was based on retrospective data, and outcomes were obtained via telephone follow-up, introducing inevitable recall bias. To minimize this bias, patient prognosis was divided into two categories according to the MRS, using independent walking as the classification criterion, which could reduce recall bias. Second, while multi-dimensional comparisons confirmed the applicability of COMB.G1 to multiple timescales, the pathophysiological changes reflected by key radiomic features in each model could not be fully elucidated. Finally, the large sample size and manual segmentation were time-consuming, resulting in a substantial workload. In future research, we plan to develop precise deep-learning segmentation models to achieve faster and more accurate region of interest segmentation. In conclusion, the results of this study suggest that the 90-day outcome prediction model for hyperacute ICH patients can be applied to outcome classification across different time windows. We constructed a universal model, COMB.G1, capable of predicting outcomes in all ICH patients with OIT < 72 h. This model provides a foundation for future research to quantify key radiomic features and explore underlying pathological mechanisms. Abbreviations ICH Intracerebral hemorrhage NCCT Non-contrast computed tomography OIT Onset-to-imaging time ROC Receiver operating characteristic AUC Area under curve ACC Accuracy SBP Systolic blood pressure GCS Glasgow coma scale GLU Glucose TG Triglyceride WBC White blood cell NE Neutrophil LY Lymphocyte NLR Neutrophil-lymphocyte ratio HGB Hemoglobin INR International normalized ratio SEN Sensitivity SPE Specificity PPV Positive predictive value NPV Negative predictive value DCA Decision curve analysis IVH Intraventricular hemorrhage SAH Subarachnoid hemorrhage CI Confidence interval OR Odds ratio Rad-score Radiomics score Declarations Acknowledgments: We extend our deepest appreciation to the patient volunteers and their families. We thank, Gansu Provincial Hospital, Lanzhou University Second Hospital and Xi’an Central Hospital (Department of Radiology) for freely providing data. We would like to thank Editage (www.editage.cn) for English language editing. Data availability statement The data that support the findings of this study are available from the corresponding author upon reasonable request. Funding This work was supported by the Guizhou Provincial Basic Research (Natural Science) Program (grant number QKHJC [2024] youth 319; recipient: Xiaoyu Huang). Author contributions: All authors attest that they meet the current International Committee of Medical Journal Editors (ICMJE) criteria for Authorship. References Al-Kawaz MN, Hanley DF, Ziai W (2020) Advances in Therapeutic Approaches for Spontaneous Intracerebral Hemorrhage. Neurotherapeutics 17:1757–1767. https://doi.org/10.1007/s13311-020-00902-w Puy L, Parry-Jones AR, Sandset EC, et al (2023) Intracerebral haemorrhage. Nat Rev Dis Primers 9:14. https://doi.org/10.1038/s41572-023-00424-7 Tsao CW, Aday AW, Almarzooq ZI, et al (2023) Heart Disease and Stroke Statistics-2023 Update: A Report From the American Heart Association. Circulation 147:e93–e621. https://doi.org/10.1161/CIR.0000000000001123 O’Carroll CB, Brown BL, Freeman WD (2021) Intracerebral Hemorrhage: A Common yet Disproportionately Deadly Stroke Subtype. Mayo Clin Proc. https://doi.org/10.1016/j.mayocp.2020.10.034 Greenberg SM, Ziai WC, Cordonnier C, et al (2022) 2022 Guideline for the Management of Patients With Spontaneous Intracerebral Hemorrhage: A Guideline From the American Heart Association/American Stroke Association. Stroke 53:e282–e361. https://doi.org/10.1161/STR.0000000000000407 Morotti A, Arba F, Boulouis G, Charidimou A (2020) Noncontrast CT markers of intracerebral hemorrhage expansion and poor outcome: A meta-analysis. Neurology 95:632–643. https://doi.org/10.1212/WNL.0000000000010660 Song Z, Tang Z, Liu H, et al (2021) A clinical-radiomics nomogram may provide a personalized 90-day functional outcome assessment for spontaneous intracerebral hemorrhage. Eur Radiol 31:4949–4959. https://doi.org/10.1007/s00330-021-07828-7 Davis SM, Broderick J, Hennerici M, et al (2006) Hematoma growth is a determinant of mortality and poor outcome after intracerebral hemorrhage. Neurology 66:1175–1181. https://doi.org/10.1212/01.wnl.0000208408.98482.99 Brott T, Broderick J, Kothari R, et al (1997) Early hemorrhage growth in patients with intracerebral hemorrhage. Stroke 28:1–5. https://doi.org/10.1161/01.str.28.1.1 Davis SM, Broderick J, Hennerici M, et al (2006) Hematoma growth is a determinant of mortality and poor outcome after intracerebral hemorrhage. Neurology 66:1175–1181. https://doi.org/10.1212/01.wnl.0000208408.98482.99 Li S, Cui L-Y, Anderson C, et al (2019) Public Awareness of Stroke and the Appropriate Responses in China: A Cross-Sectional Community-Based Study (FAST-RIGHT). Stroke 50:455–462. https://doi.org/10.1161/STROKEAHA.118.023317 Zhang J (2024) Harnessing radiomics to predict hematoma expansion in intracerebral hemorrhage: a step towards personalized care. Eur Radiol. https://doi.org/10.1007/s00330-024-11014-w Feng C, Ding Z, Lao Q, et al (2024) Prediction of early hematoma expansion of spontaneous intracerebral hemorrhage based on deep learning radiomics features of noncontrast computed tomography. Eur Radiol 34:2908–2920. https://doi.org/10.1007/s00330-023-10410-y Huang X, Wang D, Zhang Q, et al (2022) Radiomics for prediction of intracerebral hemorrhage outcomes: A retrospective multicenter study. Neuroimage Clin 36:103242. https://doi.org/10.1016/j.nicl.2022.103242 Zhou Y, Wang Y, Wang J, et al (2014) Inflammation in intracerebral hemorrhage: from mechanisms to clinical translation. Prog Neurobiol 115:25–44. https://doi.org/10.1016/j.pneurobio.2013.11.003 Iglesias-Rey R, Rodríguez-Yáñez M, Arias S, et al (2018) Inflammation, edema and poor outcome are associated with hyperthermia in hypertensive intracerebral hemorrhages. Eur J Neurol 25:1161–1168. https://doi.org/10.1111/ene.13677 d-Prostaglandin J2 activates peroxisome proliferator-activated receptor-gamma, promotes expression of catalase, and reduces inflammation, behavioral dysfunction, and neuronal loss after intracerebral hemorrhage in rats - PubMed. https://pubmed.ncbi.nlm.nih.gov/16208315/. Accessed 15 Feb 2023 Wang J, Tsirka SE (2005) Contribution of extracellular proteolysis and microglia to intracerebral hemorrhage. Neurocrit Care 3:77–85. https://doi.org/10.1385/NCC:3:1:077 Wang J, Doré S (2007) Inflammation after intracerebral hemorrhage. J Cereb Blood Flow Metab 27:894–908. https://doi.org/10.1038/sj.jcbfm.9600403 Additional Declarations No competing interests reported. Supplementary Files SupplementaryMaterial.docx Cite Share Download PDF Status: Posted Version 1 posted 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-7191002","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":508580834,"identity":"3beace68-ee3a-4466-b87f-57ee67b33e4c","order_by":0,"name":"Jianmin Liu","email":"","orcid":"","institution":"The Fifth Affiliated Hospital of Zunyi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Jianmin","middleName":"","lastName":"Liu","suffix":""},{"id":508580835,"identity":"5d341641-f729-41c5-8b91-e997387ae07d","order_by":1,"name":"Dan Wang","email":"","orcid":"","institution":"The Fifth Affiliated Hospital of Zunyi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Dan","middleName":"","lastName":"Wang","suffix":""},{"id":508580836,"identity":"ab626275-5300-4257-abb3-01aa73ad1da8","order_by":2,"name":"Jing Zhang","email":"","orcid":"","institution":"The Fifth Affiliated Hospital of Zunyi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Jing","middleName":"","lastName":"Zhang","suffix":""},{"id":508580837,"identity":"762c596a-96e2-4ab2-9260-2bb7076dd8b8","order_by":3,"name":"Shenglin Li","email":"","orcid":"","institution":"University of Electronic Science and Technology of China Sichuan Provincial People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Shenglin","middleName":"","lastName":"Li","suffix":""},{"id":508580838,"identity":"8fffcf22-a2bb-4159-8063-90b0d8189308","order_by":4,"name":"Guojin Zhang","email":"","orcid":"","institution":"University of Electronic Science and Technology of China Sichuan Provincial People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Guojin","middleName":"","lastName":"Zhang","suffix":""},{"id":508580839,"identity":"c23ba197-5b33-42e3-9a3a-aa21978526bc","order_by":5,"name":"Jialiang Ren","email":"","orcid":"","institution":"GE Healthcare","correspondingAuthor":false,"prefix":"","firstName":"Jialiang","middleName":"","lastName":"Ren","suffix":""},{"id":508580840,"identity":"8b07184f-95cf-40bd-a1b8-dc7b5b03fa61","order_by":6,"name":"Shikui Zhang","email":"","orcid":"","institution":"Guangdong Hospital of Integrated Traditional Chinese and Western Medicine","correspondingAuthor":false,"prefix":"","firstName":"Shikui","middleName":"","lastName":"Zhang","suffix":""},{"id":508580841,"identity":"ddb5e42e-a709-4bd5-a39a-6175546ca258","order_by":7,"name":"Jinhong Xiao","email":"","orcid":"","institution":"The Fifth Affiliated Hospital of Zunyi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Jinhong","middleName":"","lastName":"Xiao","suffix":""},{"id":508580842,"identity":"73688b49-0a5b-4939-a634-7f6849ae7850","order_by":8,"name":"Shaoxia Li","email":"","orcid":"","institution":"Guangdong Hospital of Integrated Traditional Chinese and Western Medicine","correspondingAuthor":false,"prefix":"","firstName":"Shaoxia","middleName":"","lastName":"Li","suffix":""},{"id":508580843,"identity":"b6d9df78-da85-43d5-9064-34a3f8e8b2f2","order_by":9,"name":"Yingyu Lin","email":"","orcid":"","institution":"Guangdong Hospital of Integrated Traditional Chinese and Western Medicine","correspondingAuthor":false,"prefix":"","firstName":"Yingyu","middleName":"","lastName":"Lin","suffix":""},{"id":508580844,"identity":"86bbd71b-ab93-4e3a-a078-ceaca10d0d29","order_by":10,"name":"Xiaoyu Huang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA60lEQVRIiWNgGAWjYBACAwglAeVWSMjJk6jljIWxYQNxWqCAsa0ikeEAAS3mEsnPHn5ts5Dnl8gxk+adJ5HA2MD88NENPFosZ6SZG8uckTCcOSPHTHLmNok8dgY2Y+McfA67kWAmLVEhwbjhRo6ZxMdtEsWMDTxs0vi1pH+TljCQsAdrSZwjkdhwgKAWoHs+VEgkQmxpIEbLmTdl0gxnJJJn9jwrtpxxTMLYsJmQX46nb5P82VZn28+evPE2T02dnDx788PH+LSAADMPiBRIYIFEKDMB5SDA+ANE8h9g/kCE4lEwCkbBKBiBAABLXEiwyoMyngAAAABJRU5ErkJggg==","orcid":"","institution":"The Fifth Affiliated Hospital of Zunyi Medical University","correspondingAuthor":true,"prefix":"","firstName":"Xiaoyu","middleName":"","lastName":"Huang","suffix":""}],"badges":[],"createdAt":"2025-07-23 01:23:19","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7191002/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7191002/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":90485957,"identity":"9aa2ccc7-a2b8-436d-9ee5-0d8fb0cbe4ab","added_by":"auto","created_at":"2025-09-03 08:52:43","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":7744158,"visible":true,"origin":"","legend":"\u003cp\u003eModel exploration and construction steps\u003c/p\u003e","description":"","filename":"Figure1.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7191002/v1/9548148e69d098617d0d52c3.jpg"},{"id":90484984,"identity":"fb4e24e3-87af-469e-a351-245c625756af","added_by":"auto","created_at":"2025-09-03 08:44:43","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":9040159,"visible":true,"origin":"","legend":"\u003cp\u003eCalibration curves A-C and Decision curves D-F of the prognostic prediction model in three cohort\u003c/p\u003e","description":"","filename":"Figure2.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7191002/v1/490c9e03c39acac32d3e014e.jpg"},{"id":90484988,"identity":"68f747c0-eb81-4826-8e9c-8617de218bb1","added_by":"auto","created_at":"2025-09-03 08:44:43","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":8374616,"visible":true,"origin":"","legend":"\u003cp\u003eCalibration curves of Group 2(A-C) and Group 3(D-F) in three cohort\u003c/p\u003e","description":"","filename":"Figure3.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7191002/v1/9813a42bf179110b82fa7b98.jpg"},{"id":90485958,"identity":"cab73e04-99ec-4d8a-bac4-9210425a865d","added_by":"auto","created_at":"2025-09-03 08:52:43","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":7000559,"visible":true,"origin":"","legend":"\u003cp\u003eDecision curves of Group 2(A-C) and Group 3(D-F) in three cohort\u003c/p\u003e","description":"","filename":"Figure4.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7191002/v1/d16660a43776b30e69177134.jpg"},{"id":108805190,"identity":"752233a2-77e9-4adb-a0f4-a5079c6a6fa5","added_by":"auto","created_at":"2026-05-08 15:25:09","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":32698079,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7191002/v1/40146e6b-5dba-4b7c-af79-e171a8043336.pdf"},{"id":90484987,"identity":"803ad891-dff6-4abc-954e-dfda84be6f16","added_by":"auto","created_at":"2025-09-03 08:44:43","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":631303,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-7191002/v1/6671b03dc476af2b82a665f7.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Functional Outcome Prediction Across Multiple Timescales Intracerebral Hemorrhage Using a Radiomics Model","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eIntracerebral hemorrhage (ICH) is the most fatal type of stroke[\u003cspan additionalcitationids=\"CR2\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. The effectiveness of treatment for ICH is highly time-dependent. The phrase \u0026ldquo;Time is brain\u0026rdquo; applies to ICH as well, as a large number of neurons can be damaged in the short period between the initial onset and hematoma expansion [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Unlike patients with ischemic stroke, patients with ICH show little or no improvement in the near term due to a lack of effective treatment options [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Thus, accurate prediction of the outcome of ICH remains an unmet need [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e] and is a pressing concern for physicians [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Hematoma expansion is an independent risk factor for poor ICH[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. More than 30% of hematoma expansions occur within 24 h after the onset of symptoms[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Therefore, radiological studies related to ICH outcomes have focused on most patients within 24 h of onset-to-imaging time (OIT).\u003c/p\u003e\u003cp\u003eIn the real world, OIT is usually more than 6 h in some ICH patients due to factors such as low public awareness, socioeconomic status, or limited medical resources [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. The outcome status of this group cannot be ignored. Recent advances in radiomics have shown promise in the field of ICH outcome prediction[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Our previous non-contrast computed tomography (NCCT)-based radiomics studies have also achieved positive results [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. However, it remained uncertain whether reexamination data from the same patient could provide more information and whether the key outcome features identified in patients with OIT\u0026thinsp;\u0026lt;\u0026thinsp;6 h could be applied to those with OIT\u0026thinsp;\u0026gt;\u0026thinsp;6 h.\u003c/p\u003e\u003cp\u003eIn this study, we will compare the radiomics model of the first examination of patients with OIT\u0026thinsp;\u0026lt;\u0026thinsp;6 hours with the radiomics model of the first preoperative follow-up examination within 72 h. Observe whether the reexamination data can provide more outcome information. Second, we applied the outcome prediction model for patients with OIT\u0026thinsp;\u0026lt;\u0026thinsp;6 h to ICH patients admitted with different time windows and compared it with the outcome prediction model specific to patients with ICH with different time windows to analyze the applicability and accuracy of the OIT\u0026thinsp;\u0026lt;\u0026thinsp;6 h model in predicting the outcome of ICH patients with OIT\u0026thinsp;\u0026gt;\u0026thinsp;6 h. To develop a model that accurately stratifies ICH patient outcome risks across multiple timescales.\u003c/p\u003e"},{"header":"2. Materials and methods","content":"\u003cp\u003eThis retrospective study was approved by the ethics committees of three medical institutions. Ethical approval was obtained from the Center 1-Lanzhou University Second Hospital, Center 2-Gansu Provincial Hospital, and Center 3-Xi\u0026rsquo;an Central Hospital institutional review boards (Approval number: 2022A-096, 2022\u0026ndash;275 and LW-2022\u0026ndash;011). As the study was a retrospective study, informed consent was waived by the Ethics Committee.\u003c/p\u003e\u003cp\u003eThis retrospective study collected NCCT data, clinical data, and laboratory data from ICH patients with OIT\u0026thinsp;\u0026lt;\u0026thinsp;72 h after admission to Center 1(January 1, 2016, to October 1, 2020), Center 2, and Center 3(June 1, 2021, to November 30, 2021), and for patients with OIT\u0026thinsp;\u0026lt;\u0026thinsp;6 h for their first preoperative review at Center 1 and their NCCT data. Clinical data included OIT, age, sex, admission systolic blood pressure (SBP), and Glasgow coma scale (GCS) score; GCS\u0026thinsp;\u0026lt;\u0026thinsp;9 was defined as a low score, GCS\u0026thinsp;\u0026ge;\u0026thinsp;9 was defined as a high score scale. The Modified Rankin Scale (MRS) 90 days post-symptom onset was used to evaluate the outcome. A score of 0\u0026ndash;3 was defined as good outcome, 4\u0026ndash;6 was poor outcome. Laboratory data included glucose (GLU), triglyceride (TG), white blood cell count (WBC), etc.\u003c/p\u003e\u003cp\u003eThe patients were divided into three groups based on OIT: Group 1(OIT\u0026thinsp;\u0026lt;\u0026thinsp;6 h), Group 2(6 OIT\u0026thinsp;\u0026lt;\u0026thinsp;24 h), and Group 3(24 OIT\u0026thinsp;\u0026lt;\u0026thinsp;72 h). Group 4 included NCCT data of patients with OIT\u0026thinsp;\u0026lt;\u0026thinsp;6 h ICH who underwent their first preoperative review within 72 h after admission (Center 1). The inclusion criteria were clinical diagnosis of ICH, age\u0026thinsp;\u0026gt;\u0026thinsp;18 years, baseline NCCT scan within 72 h of OIT, and complete clinical and laboratory data after admission. Exclusion criteria refer to previous studies[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Patients from Group 1\u0026ndash;3 were further divided based on their treatment center: training set (C1: January 1, 2016, to December 31, 2018), internal validation set (C1: January 1, 2019, to October 1, 2020), and external validation set (C2\u0026thinsp;+\u0026thinsp;C3).\u003c/p\u003e\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Feature Selection\u003c/h2\u003e\u003cp\u003eNCCT parameters, segmentation of regions of interest, and radiomic feature extraction procedure are detailed in the supplementary materials. Clinical features (include clinical, radiological, and laboratory data) were initially screened using one-way analysis, Kruskal-Wallis test, Chi-Square Test, and Mann-Whitney U test(P\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Subsequently, multiple logistic regression was employed to identify factors independently associated with outcome, retaining features with P\u0026thinsp;\u0026lt;\u0026thinsp;0.05. Radiomics features were screened using a two-step process. First, Mann\u0026ndash;Whitney U was used to screen the radiomics features related to the outcome, and the features with P\u0026thinsp;\u0026lt;\u0026thinsp;0.01 were retained. Second, multiple logistic regression and stepwise regression analyses were used to screen the key radiomic features independently related to the outcome, and the final key features with P\u0026thinsp;\u0026lt;\u0026thinsp;0.05 were retained.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 Model construction and validation\u003c/h2\u003e\u003cp\u003eClinical and radiomics features were analyzed by univariate and multivariate analyses. Logistic regression was then used to construct outcome models for Groups 1, 2, 3, and 4, resulting in Models 1, 2, 3, and 4, respectively. Second, we applied the prediction models (Models 1\u0026ndash;4) and compared their performance with Model 4 to assess whether the inclusion of NCCT radiomics features can provide additional information for outcome prediction and further improve prediction accuracy. Last, involved applying the radiomics feature in Model 1 directly to Groups 2 and 3, followed by comparing their predictive performance to determine the feasibility and accuracy of Model 1 for different OIT groups. The detailed model exploration and construction are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe prediction model based on key radiomics features from Group 1 was named Rad.G1, while the Clinical model (based on clinical features) was named Clinical.G1. The comprehensive prediction model incorporating radiomics and clinical features was named COMB.G1. Similarly, the radiomics, clinical, and comprehensive model were named Rad.G2, Clinical.G2, and COMB.G2, respectively. For Group 3, the radiomic model was named Rad.G3, the clinical model was named Clinical.G3, and the comprehensive model was named COMB.G3. The radiomics model for Group 4 was named Rad.G4.\u003c/p\u003e\u003cp\u003eTo determine whether key radiomics features from preoperative review NCCT provide additional information for outcome prediction, we conducted the following analysis: ①The Rad.G1 model was directly applied to Group 4 patients, denoted as Rad.G1.G4. Similarly, COMB.G1 and Clinical.G1 were applied to Group 4, denoted as COMB.G1.G4 and Clinical.G1.G4; ② Radiomics features from the Rad.G1 model were combined with those extracted from Group 4, and the combined features were refined using stepwise regression. The final key features, identified through screening were labeled as Rad.G1wG4 for constructing the outcome prediction model; ③Independent clinical features were combined with Rad.G1wG4 features to construct the comprehensive Group 4 outcome model, COMB.G1wG4;④The predictive performance of Rad.G1.G4, Rad.G1wG4, Rad.G4 was compared across multiple dimensions to assess the added value of key radiomics features from NCCT data in predicting outcomes.\u003c/p\u003e\u003cp\u003eTo assess the applicability of the Group 1 outcome prediction model in predicting outcomes for ICH patients across different OIT groups, we conducted the following experiments: ①The models Rad.G1, Clinical.G1, and COMB.G1 were applied to Group 2 patients, denoted as Rad.G1.G2, Clinical.G1.G2, and COMB.G1.G2. Simultaneously, the accuracy of the outcome prediction models Rad.G2, Clinical.G2, and COMB.G2, which were constructed using data from Group 2, was compared and analyzed across multiple dimensions to assess the applicability of the Group 1 model in Group 2. The optimal model from Group 1, COMB.G1, was then applied to Group 3 and is denoted as COMB.G1.G3. Additionally, multi-model and multi-dimensional comparisons were conducted using Rad.G3, Clinical.G3, and COMB.G3 models constructed using the data of patients in Group 3 to evaluate the applicability of the Group 1 model in Group 3.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3 Statistical Analyses\u003c/h2\u003e\u003cp\u003eContinuous variables are reported as medians and interquartile ranges (IQR), while categorical variables are presented as numbers and percentages. Univariate analysis included one-way ANOVA, Kruskal-Wallis test, chi-square test, and Mann-Whitney U test. A significance threshold of p\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was used, with a stricter threshold of p\u0026thinsp;\u0026lt;\u0026thinsp;0.01 for radiomic feature screening. Multivariate logistic regression was employed to identify factors independently associated with functional outcomes, and forward-backward stepwise regression was used to refine radiomics features (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Model performance was evaluated using receiver operating characteristic (ROC) analysis, with accuracy assessed by the area under the curve (AUC), accuracy (ACC), sensitivity (SEN), specificity (SPE), positive predictive value (PPV), and negative predictive value (NPV). The DeLong test revealed significant differences in AUC between models (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). The Hosmer-Lemeshow test was used to assess the goodness of fit between the model\u0026rsquo;s predictions and actual outcomes, and decision curve analysis (DCA) evaluated the clinical utility of the model. Statistical analyses were performed using SPSS version 25 (IBM, Armonk, NY, USA) and R (version 4.1.0; R Foundation for Statistical Computing, Vienna, Austria).\u003c/p\u003e\u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Characteristics of the Study Sample\u003c/h2\u003e\u003cp\u003eA total of 2,136 patients with ICH were included from three medical centers. The number of patients in groups 1\u0026ndash;4 was 1,098, 660, 378 and 584. There were 1,235 male patients (58%) and 901 female patients (42%). The median patient age was 59 years, and the median OIT was 5 h. In groups 1\u0026ndash;3, Group 3 patients had the lowest rate of poor outcomes, with rates of 703(64%), 384(58%), and 162(43%) patients in the three groups, respectively. The high score rate of GCS score of group 3 was also higher compared to groups 1 and 2 by 83%, 64%, and 72%, respectively (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\u003eBaseline Patient Characteristics\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eParameter\u003c/p\u003e\u003cp\u003e\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;Group\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGroup 1\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eGroup 2\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eGroup 3\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eGroup 4\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eAll Data\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003en\u0026thinsp;=\u0026thinsp;1098\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003en\u0026thinsp;=\u0026thinsp;660\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003en\u0026thinsp;=\u0026thinsp;378\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003en\u0026thinsp;=\u0026thinsp;584\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003en\u0026thinsp;=\u0026thinsp;2136\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePoor outcome*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e703 (64)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e384 (58)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e162 (43)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e346 (59)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1247 (58)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOIT (h)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3.0 (1.5, 4.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e10.0 (7.0, 13.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e31.6\u0026thinsp;\u0026plusmn;\u0026thinsp;11.2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e3.0 (1.5, 4.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e5.0(3.0, 13.0)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGender*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e652 (59)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e372 (56)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e211 (56)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e345 (59)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1235(58)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFemale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e446 (41)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e288 (44)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e167 (44)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e239 (41)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e901(42)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge (y)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e60 (52, 70)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e60 (51, 69)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e57 (50, 67)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e59 (51, 68)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e59 (51, 69)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLocation*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDeep\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e977 (89)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e545 (83)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e281 (74)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e534 (91)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1803 (85)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLobar\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e121 (11)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e115 (17)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e97 (26)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e50 (9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e333 (15)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMidline shift*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e282 (26)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e132 (20)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e47 (12)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e83 (14)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e461 (22)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIVH*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e546 (50)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e317 (48)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e137 (36)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e249 (43)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1001 (47)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSAH*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e208 (19)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e131 (20)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e55 (15)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e58 (10)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e394 (18)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHypodensities*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e729 (66)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e382 (58)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e223 (59)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e379 (65)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1334 (63)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eICH volume (ml)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e31.1 (13.9, 70.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e33.2 (14.9, 60.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e23.2 (10.4, 40.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e24.7 (13.4, 44.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e29.8 (13.2, 60.9)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePHE volume (ml)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e11.2 (4.9, 23.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e13.3 (5.9, 24.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e12.6 (5.7, 24.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e8.9 (4.5, 16.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e12.2 (5.3, 23.9)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTemperature (℃)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e36.6 (36.4, 36.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e36.6 (36.5, 36.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e36.6 (36.5, 36.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e36.6 (36.5, 36.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e36.6 (36.5, 36.8)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSmoking*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e189 (17)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e87 (13)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e31 (8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e60 (10)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e308 (14)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSBP (mmHg)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e172(153, 191)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e162 (145, 179)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e159 (143, 176)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e170 (153, 189)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e166 (148, 186)\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=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLow (score\u0026lt;9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e399 (36)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e185 (28)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e63 (17)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e147 (25)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e647 (30)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHigh (score\u0026thinsp;\u0026ge;\u0026thinsp;9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e699 (64)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e475 (72)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e315 (83)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e437 (75)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1489 (70)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGLU (mmol/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e7.90 (6.40, 9.89)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7.55 (6.50, 9.39)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e6.73 (5.63, 7.90)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e7.48 (6.30, 9.30)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e7.50 (6.25, 9.34)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTG (mmol/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.26 (0.99, 1.99)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.95 (0.64, 1.45)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1.08 (0.80, 1.88)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e1.26 (0.76, 1.98)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1.10 (0.74, 1.84)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWBC (10^9/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e8.61 (6.38, 11.79)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e9.22 (7.04, 11.97)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e8.74 (6.76, 11.46)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e8.20 (6.53, 10.60)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e8.84 (6.66, 11.72)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNE (10^9/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e6.94 (4.46, 10.10)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7.93 (5.62, 10.70)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e7.10 (5.08, 9.86)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e6.53 (4.25, 9.33)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e7.25 (4.85, 10.24)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLY (10^9/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.04 (0.70, 1.60)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.84 (0.59, 1.18)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1.04 (0.71, 1.47)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e1.00(0.69, 1.51)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.96 (0.67, 1.42)\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=\"left\" colname=\"c2\"\u003e\u003cp\u003e6.25 (3.35, 12.25)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e9.89 (5.39, 15.97)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e6.75 (3.98, 11.87)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e6.26 (3.38, 11.77)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e7.55 (3.96, 13.37)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHGB (g/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e148 (136, 161)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e150 (137, 164)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e151 (137, 164)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e148 (136, 159)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e149 (137, 162)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eINR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.99 (0.94, 1.05)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.00 (0.94, 1.05)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1.01 (0.94, 1.05)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.98 (0.93, 1.04)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.99 (0.94, 1.05)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"6\"\u003eNote.\u0026mdash; Unless otherwise indicated, data are presented as median (interquartile range) and data in parentheses are percentages.\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"6\"\u003eOIT\u0026thinsp;=\u0026thinsp;onset-to-imaging time, IVH\u0026thinsp;=\u0026thinsp;Intraventricular hemorrhage, SAH\u0026thinsp;=\u0026thinsp;Subarachnoid hemorrhage, SBP\u0026thinsp;=\u0026thinsp;Systolic blood pressure, GCS\u0026thinsp;=\u0026thinsp;Glasgow coma scale, GLU\u0026thinsp;=\u0026thinsp;Glucose, TG\u0026thinsp;=\u0026thinsp;Triglyceride, WBC\u0026thinsp;=\u0026thinsp;White blood cell, NE\u0026thinsp;=\u0026thinsp;Neutrophil, LY\u0026thinsp;=\u0026thinsp;Lymphocyte, NLR\u0026thinsp;=\u0026thinsp;Neutrophil-lymphocyte ratio, HGB\u0026thinsp;=\u0026thinsp;Hemoglobin, INR\u0026thinsp;=\u0026thinsp;International normalized ratio\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"6\"\u003e* Data are percentages n (%).\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThe incidence of positive hypodensity was higher in Group 1 (approximately 66%), while intraventricular hemorrhage and midline shift also had the highest incidences in the group with 50% and 26% respectively. The median hematoma volumes for Groups 1, 2, and 3 were 31.1 ml, 33.2 ml, and 23.2 ml, respectively.\u003c/p\u003e\u003cp\u003eAfter performing univariate and multivariate logistic regression analyses on all clinical features in Group 2 and Group 3. The age, hematoma volume, GCS score, WBC count, and hematoma location were independent independently associated with outcome in Group 2. In Group 3, the independent risk factors were age, hematoma volume, GCS score, and hematoma location.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Radiomics feature selection and model construction\u003c/h2\u003e\u003cp\u003e12 features were included in Group 1[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e], 8 in Group 2, 3 in Group 3, and 8 in Group 4. Seven key features were retained after the screening via stepwise forward-backward regression for Group 1, combined with similarly named radiomics features from Group 4(Supplementary Tables\u0026nbsp;1).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Models performance\u003c/h2\u003e\u003cp\u003eThe AUC value of Rad.G1.G4 in the training set was 0.779. The AUC value of COMB.G1.G4 were 0.864. The AUC and ACC value of Rad.G1wG4 were 0.788 and 0.726; Rad.G4 were 0.815and 0.768, Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e and Supplementary Fig.\u0026nbsp;2. The Delong test comparing the AUCs of different models showed that the training cohort was the p-value for Rad.G1.G4 vs. Rad.G4 significant at 0.04, while all other comparisons had p\u0026thinsp;\u0026gt;\u0026thinsp;0.05(Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). In addition, the direct application of COMB.G1.G4 maintained high predictive performance and good stability in the three cohorts, with AUC values of 0.864, 0.804, and 0.800.\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\u003ePerformance of each model in three cohort\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" 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\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAUC\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eACC\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\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eCohort\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G1.G4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.779(0.727\u0026ndash;0.830)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.729(0.677\u0026ndash;0.778)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.741(0.619\u0026ndash;0.810)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.712(0.552\u0026ndash;0.792)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\" morerows=\"5\" rowspan=\"6\"\u003e\u003cp\u003eTraining cohort\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G1wG4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.788(0.737\u0026ndash;0.838)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.726(0.673\u0026ndash;0.775)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.688(0.518\u0026ndash;0.788)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.784(0.664\u0026ndash;0.856)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.815(0.768\u0026ndash;0.862)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.768(0.717\u0026ndash;0.813)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.862(0.714\u0026ndash;0.910)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.624(0.472\u0026ndash;0.704)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eClinical.G1.G4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.828(0.784\u0026ndash;0.872)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.729(0.677\u0026ndash;0.778)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.698(0.603\u0026ndash;0.778)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.776(0.640\u0026ndash;0.856)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G1.G4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.864(0.825\u0026ndash;0.903)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.768(0.717\u0026ndash;0.813)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.683(0.571\u0026ndash;0.767)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.896(0.792\u0026ndash;0.952)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G1wG4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.872(0.834\u0026ndash;0.909)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.796(0.747\u0026ndash;0.839)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.757(0.672\u0026ndash;0.826)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.856(0.752\u0026ndash;0.928)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G1.G4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.721(0.659\u0026ndash;0.784)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.639(0.577\u0026ndash;0.698)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.653(0.551\u0026ndash;0.823)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.620(0.500-0.769)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\" morerows=\"5\" rowspan=\"6\"\u003e\u003cp\u003eInternal validation cohort\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G1wG4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.751(0.691\u0026ndash;0.812)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.710(0.650\u0026ndash;0.765)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.673(0.422\u0026ndash;0.755)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.759(0.583\u0026ndash;0.843)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.740(0.679-0.800)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.682(0.621\u0026ndash;0.739)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.810(0.680\u0026ndash;0.912)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.509(0.407\u0026ndash;0.621)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eClinical.G1.G4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.790(0.734\u0026ndash;0.845)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.686(0.625\u0026ndash;0.743)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.633(0.537\u0026ndash;0.816)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.759(0.667\u0026ndash;0.870)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G1.G4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.804(0.751\u0026ndash;0.857)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.706(0.646\u0026ndash;0.761)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.612(0.469\u0026ndash;0.748)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.833(0.759\u0026ndash;0.898)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G1wG4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.800(0.745\u0026ndash;0.854)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.698(0.638\u0026ndash;0.754)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.612(0.496\u0026ndash;0.748)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.815(0.750\u0026ndash;0.908)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G1.G4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.560(0.179\u0026ndash;0.941)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.533(0.266\u0026ndash;0.787)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.600(0.300-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.400(0.000\u0026ndash;1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\" morerows=\"5\" rowspan=\"6\"\u003e\u003cp\u003eExternal validation cohort\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G1wG4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.600(0.233\u0026ndash;0.967)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.667(0.384\u0026ndash;0.882)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.600(0.000-0.900)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.800(0.200-1.000)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.700(0.324-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.733(0.449\u0026ndash;0.922)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.000(0.798-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.200(0.000\u0026ndash;1.000)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eClinical.G1.G4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.820(0.587-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.800(0.519\u0026ndash;0.957)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.000(0.600-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.400(0.000\u0026ndash;1.000)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G1.G4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.800(0.560-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.600(0.323\u0026ndash;0.837)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.700(0.400-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.400(0.200-1.000)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G1wG4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.720(0.342-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.800(0.519\u0026ndash;0.957)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.000(0.598-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.400(0.195-1.000)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eDelong test of model AUC area\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\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\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eName\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eP value\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eCohort\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G1.G4-Rad.G1wG4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\" morerows=\"3\" rowspan=\"4\"\u003e\u003cp\u003eTraining cohort\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G1.G4-Rad.G4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.04\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G1wG4-Rad.G4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.09\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G1.G4-COMB.G1wG4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.36\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G1.G4-Rad.G1wG4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\" morerows=\"3\" rowspan=\"4\"\u003e\u003cp\u003eInternal validation cohort\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G1.G4-Rad.G4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.40\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G1wG4-Rad.G4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.57\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G1.G4-COMB.G1wG4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.76\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G1.G4-Rad.G1wG4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.60\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\" morerows=\"3\" rowspan=\"4\"\u003e\u003cp\u003eExternal validation cohort\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G1.G4-Rad.G4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.20\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G1wG4-Rad.G4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.19\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G1.G4-COMB.G1wG4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.64\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThe Hosmer-Lemeshow test showed that in both the training and internal validation cohorts, all models closely followed the diagonal, representing the predicted probability of an ideal model Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA-C. Decision curves analysis (DCA) for the training and internal validation cohort showed a good net benefit in predicting functional outcomes of ICH patients (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eD-F).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe AUC values for Rad.G1.G2 and Rad. G2 were 0.773 and 0.795 in training cohort (Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e and Supplementary Fig.\u0026nbsp;3). The Delong test revealed no significant difference in the AUCs for Rad.G1.G2 vs. Rad.G2 across the three cohorts. After adding clinical and laboratory-independent risk factors, the AUC values of COMB.G1.G2 vs. COMB.G2 were still not statistically significant Supplementary Table\u0026nbsp;2.\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\u003ePerformance of each model in three cohort\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"8\"\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\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\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAUC\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eACC\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\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003ePPV\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003eNPV\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u003cp\u003eCohort\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G1.G2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.773(0.723\u0026ndash;0.822)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.747(0.699\u0026ndash;0.792)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.752(0.614\u0026ndash;0.819)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.740(0.514\u0026ndash;0.808)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.806(0.772\u0026ndash;0.819)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.675(0.591\u0026ndash;0.694)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\" morerows=\"5\" rowspan=\"6\"\u003e\u003cp\u003eTraining cohort\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eClinical.G1.G2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.859(0.821\u0026ndash;0.898)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.803(0.758\u0026ndash;0.843)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.776(0.619\u0026ndash;0.829)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.842(0.698\u0026ndash;0.897)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.876(0.850\u0026ndash;0.883)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.724(0.685\u0026ndash;0.736)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G1.G2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.860(0.822\u0026ndash;0.899)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.795(0.749\u0026ndash;0.836)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.786(0.695\u0026ndash;0.838)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.808(0.698\u0026ndash;0.884)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.855(0.839\u0026ndash;0.863)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.724(0.694\u0026ndash;0.741)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.786(0.739\u0026ndash;0.833)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.728(0.678\u0026ndash;0.773)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.671(0.566\u0026ndash;0.729)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.808(0.636\u0026ndash;0.877)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.834(0.809\u0026ndash;0.845)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.631(0.574\u0026ndash;0.650)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eClinical.G2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.863(0.825-0.900)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.787(0.740\u0026ndash;0.828)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.762(0.657\u0026ndash;0.829)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.822(0.692\u0026ndash;0.884)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.860(0.841\u0026ndash;0.870)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.706(0.669\u0026ndash;0.721)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.869(0.832\u0026ndash;0.906)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.792(0.746\u0026ndash;0.833)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.752(0.638\u0026ndash;0.824)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.849(0.740\u0026ndash;0.911)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.878(0.859\u0026ndash;0.887)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.705(0.675\u0026ndash;0.719)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G1.G2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.795(0.742\u0026ndash;0.849)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.714(0.656\u0026ndash;0.767)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.743(0.651\u0026ndash;0.868)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.675(0.590\u0026ndash;0.812)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.748(0.723\u0026ndash;0.776)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.669(0.639\u0026ndash;0.709)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\" morerows=\"5\" rowspan=\"6\"\u003e\u003cp\u003eInternal validation cohort\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eClinical.G1.G2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.866(0.825\u0026ndash;0.908)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.762(0.707\u0026ndash;0.812)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.724(0.625\u0026ndash;0.836)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.812(0.692\u0026ndash;0.923)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.833(0.812\u0026ndash;0.852)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.693(0.659\u0026ndash;0.720)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G1.G2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.872(0.831\u0026ndash;0.912)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.796(0.742\u0026ndash;0.842)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.783(0.657\u0026ndash;0.855)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.812(0.675\u0026ndash;0.881)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.844(0.820\u0026ndash;0.855)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.742(0.705\u0026ndash;0.757)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.812(0.761\u0026ndash;0.864)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.743(0.687\u0026ndash;0.795)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.704(0.586\u0026ndash;0.796)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.795(0.658\u0026ndash;0.889)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.817(0.788\u0026ndash;0.835)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.674(0.631\u0026ndash;0.698)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eClinical.G2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.875(0.835\u0026ndash;0.915)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.773(0.718\u0026ndash;0.822)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.730(0.651\u0026ndash;0.829)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.829(0.743\u0026ndash;0.949)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.847(0.832\u0026ndash;0.863)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.703(0.680\u0026ndash;0.730)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.883(0.844\u0026ndash;0.921)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.773(0.718\u0026ndash;0.822)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.717(0.632\u0026ndash;0.816)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.846(0.752\u0026ndash;0.957)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.858(0.842\u0026ndash;0.873)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.697(0.672\u0026ndash;0.723)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G1.G2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.706(0.528\u0026ndash;0.884)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.600(0.421\u0026ndash;0.761)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.471(0.118\u0026ndash;0.941)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.722(0.500-0.944)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.615(0.286\u0026ndash;0.762)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.591(0.500-0.654)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\" morerows=\"5\" rowspan=\"6\"\u003e\u003cp\u003eExternal validation cohort\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eClinical.G1.G2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.837(0.683\u0026ndash;0.990)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.800(0.631\u0026ndash;0.916)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.824(0.471-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.778(0.054-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.778(0.667\u0026ndash;0.810)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.824(0.245\u0026ndash;0.857)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G1.G2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.866(0.744\u0026ndash;0.988)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.771(0.599\u0026ndash;0.896)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.824(0.587-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.722(0.499\u0026ndash;0.946)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.737(0.666\u0026ndash;0.773)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.812(0.749\u0026ndash;0.850)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.807(0.661\u0026ndash;0.954)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.714(0.537\u0026ndash;0.854)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.412(0.353\u0026ndash;0.824)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.000(0.611-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.000(1.000\u0026ndash;1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.643(0.524\u0026ndash;0.643)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eClinical.G2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.879(0.749-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.829(0.664\u0026ndash;0.934)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.882(0.706-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.778(0.054-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.789(0.750\u0026ndash;0.810)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.875(0.328-0.900)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.915(0.805-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.857(0.697\u0026ndash;0.952)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.824(0.587-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.889(0.667-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.875(0.833\u0026ndash;0.895)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.842(0.800-0.857)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThe Hosmer\u0026ndash;Lemeshow test calibration curves indicated that the predicted probabilities of each model in the training and internal validation cohorts closely aligned with actual probabilities (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA-C). DCA showed that in the training cohorts, both COMB.G1.G2 and COMB.G2 achieved high net benefits (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA-C).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eRad. G3 performed at a moderate level in the internal validation cohorts, with AUC values of 0.717. COMB.G3 had a higher classification performance in the training cohort, with AUC values of 0.817. COMB.G1.G3 exhibited classification performance comparable to COMB.G3 for predicting ICH outcomes. The other parameters for each model are shown in Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e and Supplementary Figs.\u0026nbsp;3. The Delong test showed no significant difference in the AUCs for COMB.G1.G3 vs. COMB.G3 across the three cohorts. The Hosmer-Lemeshow test calibration curves demonstrated that COMB.G1.G3 and COMB.G3 were the models closest to the actual prediction probability in the internal validation cohort (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eD-f). The DCA indicated that COMB.G1.G3 achieved high clinical net benefits within the threshold range of 0.3\u0026ndash;0.9 in the training cohort (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eD-f).\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\u003ePerformance of each model in three cohort\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"8\"\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\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\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAUC\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eACC\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\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003ePPV\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003eNPV\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u003cp\u003eCohort\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G1.G3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.811(0.755\u0026ndash;0.867)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.781(0.722\u0026ndash;0.832)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.629(0.467\u0026ndash;0.724)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.906(0.711\u0026ndash;0.961)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.846(0.803\u0026ndash;0.864)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.748(0.700-0.759)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\" morerows=\"3\" rowspan=\"4\"\u003e\u003cp\u003eTraining cohort\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.678(0.609\u0026ndash;0.746)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.657(0.592\u0026ndash;0.717)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.562(0.410\u0026ndash;0.638)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.734(0.555\u0026ndash;0.828)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.634(0.558\u0026ndash;0.663)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.671(0.607\u0026ndash;0.697)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eClinical.G3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.814(0.759\u0026ndash;0.869)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.768(0.709\u0026ndash;0.821)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.686(0.533\u0026ndash;0.762)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.836(0.633\u0026ndash;0.922)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.774(0.727\u0026ndash;0.792)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.764(0.710\u0026ndash;0.781)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.817(0.763\u0026ndash;0.872)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.768(0.709\u0026ndash;0.821)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.648(0.514\u0026ndash;0.733)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.867(0.703\u0026ndash;0.953)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.800(0.760\u0026ndash;0.819)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.750(0.709\u0026ndash;0.767)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G1.G3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.847(0.780\u0026ndash;0.914)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.792(0.712\u0026ndash;0.858)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.566(0.340\u0026ndash;0.699)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.948(0.805-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.882(0.818\u0026ndash;0.902)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.760(0.729\u0026ndash;0.770)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\" morerows=\"3\" rowspan=\"4\"\u003e\u003cp\u003eInternal validation cohort\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.717(0.627\u0026ndash;0.806)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.692(0.605\u0026ndash;0.770)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.547(0.377\u0026ndash;0.736)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.792(0.610\u0026ndash;0.909)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.644(0.556\u0026ndash;0.709)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.718(0.662\u0026ndash;0.745)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eClinical.G3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.838(0.766\u0026ndash;0.909)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.769(0.687\u0026ndash;0.839)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.679(0.528\u0026ndash;0.793)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.831(0.649\u0026ndash;0.948)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.735(0.683\u0026ndash;0.764)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.790(0.746\u0026ndash;0.811)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.854(0.790\u0026ndash;0.918)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.777(0.696\u0026ndash;0.845)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.623(0.472\u0026ndash;0.736)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.883(0.714\u0026ndash;0.987)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.786(0.735\u0026ndash;0.812)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.773(0.733\u0026ndash;0.792)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G1.G3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.944(0.813-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.867(0.595\u0026ndash;0.983)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.667(0.325-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.917(0.750-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.667(0.494\u0026ndash;0.750)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.917(0.900-0.923)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\" morerows=\"3\" rowspan=\"4\"\u003e\u003cp\u003eExternal validation cohort\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRad.G3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.972(0.895-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.933(0.681\u0026ndash;0.998)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.667(0.333-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.000(0.833-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.000(1.000\u0026ndash;1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.923(0.909\u0026ndash;0.923)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eClinical.G3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.944(0.823-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.933(0.681\u0026ndash;0.998)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.000(0.000\u0026ndash;1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.917(0.750-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.750(0.000-0.750)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e1.000(1.000\u0026ndash;1.000)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOMB.G3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.944(0.823-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.867(0.595\u0026ndash;0.983)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.667(0.000\u0026ndash;1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.917(0.750-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.667(0.000-0.750)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.917(0.900-0.923)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eOur study developed and validated a radiomics model for predicting 90-day functional outcomes in patients with ICH. Not only can the model accurately predict functional outcomes for ICH patients with OIT\u0026thinsp;\u0026lt;\u0026thinsp;6 h, but by exploring its feasibility across different timescales. This fills a gap in the literature regarding ICH prognosis.\u003c/p\u003e\u003cp\u003eThe results showed that Group 3 patients were the latest to be admitted to hospital in groups 1\u0026ndash;3, but had the lowest rate of poor prognosis in the three groups. These findings contradict the principle of \u0026ldquo;Time is brain.\u0026rdquo; Further analysis revealed that the age and hematoma volume of patients in Group 3 were the lowest among the three groups. The incidence of high GCS scores upon admission was the highest. Due to the relatively small amount of bleeding, these patients exhibited mild clinical symptoms with minimal neurological dysfunction. These patients may present only with persistent headaches. Moreover, due to an insufficient understanding of stroke symptoms by patients or their families, stroke attacks are often not identified in a timely manner, leading to delayed admission. Nevertheless, it is important to note that up to 40% of ICH patients in this group still have a poor prognosis. Therefore, accurate stratification of prognosis in this subset of patients is crucial for reducing the disease burden of ICH.\u003c/p\u003e\u003cp\u003eThe radiomics model constructed from the first NCCT scan of patients with OIT\u0026thinsp;\u0026lt;\u0026thinsp;6 h was directly applied to follow-up patients and achieved good performance in prognosis prediction. This result indicates that a review of NCCT data does not provide additional value for predicting the prognosis of patients with ICH. This suggests that the COMB.G1 model for hyperacute phase ICH, developed in previous studies, can accurately predict the 90-day functional outcomes in ICH patients. The key radiomic features of Rad.G1 reflect critical pathophysiological changes that impact prognosis, key molecular expressions related to outcomes, and inflammatory mechanisms.\u003c/p\u003e\u003cp\u003eBy comparing clinical and laboratory outcome-independent risk factors, it was found that elevated WBC count in Group 2 was an independent risk factor affecting prognosis. Increasing evidence [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e] indicates that inflammatory injury plays a key role in ICH-induced secondary brain injury and that the inflammatory response is an important defense mechanism following cerebral hemorrhage. When intracerebral hemorrhage occurs, blood components, including red and WBC, macrophages, and plasma proteins such as thrombin, enter the brain parenchyma, triggering an inflammatory response[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Leukocyte activation is most active within the first 24 h after ICH onset, marked by the accumulation and activation of inflammatory cells. The exudation of white blood cells is central to the inflammatory process, followed by the lysis of red blood cells, which release cytotoxic substances like Hb, heme, and iron, further exacerbating brain damage. The inflammatory response following a hemorrhage aggravates cerebral hemorrhage-induced brain injury, leading to neuronal injury, destruction of the blood-brain barrier, and widespread brain cell death [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. This may be the main reason why an elevated WBC count was an independent risk factor for poor prognosis in Group 2.\u003c/p\u003e\u003cp\u003eIn group 2, after combining clinical and laboratory factors, the AUC values for COMB.G1.G2 and COMB.G2 exceeded 0.85 in all three cohorts, with no statistical difference between the two models. This indicates that the comprehensive model with OIT\u0026thinsp;\u0026lt;\u0026thinsp;6 h is also applicable to patients with 6\u0026thinsp;\u0026le;\u0026thinsp;OIT\u0026lt;24 h. Calibration and DCA showed that the COMB.G1.G2 model had good accuracy and clinical utility in Group 2 patients. This result demonstrates that applying the Group 1 model to predict the functional prognosis of patients with ICH in Group 2 can achieve high predictive accuracy. Specifically, the 90-day outcome prediction model constructed for ICH patients with OIT\u0026thinsp;\u0026lt;\u0026thinsp;6 h can also be applied to those with OIT\u0026thinsp;\u0026lt;\u0026thinsp;24 h with good accuracy and stability.\u003c/p\u003e\u003cp\u003eWhen comparing the applicability of the comprehensive model, COMB.G1 to Group 3, it was found that COMB.G1.G3 and COMB.G3 achieved similar predictive accuracies. In the training cohort, the AUC values were 0.811 for COMB.G1.G3 and 0.817 for COMB.G3. Notably, COMB.G1.G3 demonstrated higher accuracy and specificity than COMB.G3. In both Groups 1 and 3, the same clinical and laboratory risk factors were independently associated with outcome. Calibration and DCA showed that COMB.G1.G3 had good accuracy and clinical utility in Group 3 patients. These results suggest that the Group 1 model can also be used to predict the functional prognosis of ICH patients in Group 3. In other words, the 90-day outcome prediction model constructed for ICH patients with OIT\u0026thinsp;\u0026lt;\u0026thinsp;6 h can also be applied to 24\u0026thinsp;\u0026le;\u0026thinsp;OIT\u0026lt;72h ICH patients with good accuracy and stability.\u003c/p\u003e\u003cp\u003eThis study had several limitations. First, it was based on retrospective data, and outcomes were obtained via telephone follow-up, introducing inevitable recall bias. To minimize this bias, patient prognosis was divided into two categories according to the MRS, using independent walking as the classification criterion, which could reduce recall bias. Second, while multi-dimensional comparisons confirmed the applicability of COMB.G1 to multiple timescales, the pathophysiological changes reflected by key radiomic features in each model could not be fully elucidated. Finally, the large sample size and manual segmentation were time-consuming, resulting in a substantial workload. In future research, we plan to develop precise deep-learning segmentation models to achieve faster and more accurate region of interest segmentation.\u003c/p\u003e\u003cp\u003eIn conclusion, the results of this study suggest that the 90-day outcome prediction model for hyperacute ICH patients can be applied to outcome classification across different time windows. We constructed a universal model, COMB.G1, capable of predicting outcomes in all ICH patients with OIT\u0026thinsp;\u0026lt;\u0026thinsp;72 h. This model provides a foundation for future research to quantify key radiomic features and explore underlying pathological mechanisms.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eICH\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eIntracerebral hemorrhage\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eNCCT\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eNon-contrast computed\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003etomography\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eOIT\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eOnset-to-imaging time\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eROC\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eReceiver operating characteristic\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eAUC\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eArea under curve\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eACC\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eSBP\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eSystolic blood pressure\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eGCS\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eGlasgow coma scale\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eGLU\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eGlucose\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eTG\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eTriglyceride\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eWBC\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eWhite blood cell\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eNE\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eNeutrophil\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eLY\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eLymphocyte\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eNLR\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eNeutrophil-lymphocyte ratio\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eHGB\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eHemoglobin\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eINR\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eInternational normalized ratio\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eSEN\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eSensitivity\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eSPE\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eSpecificity\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003ePPV\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003ePositive predictive value\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eNPV\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eNegative predictive value\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eDCA\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eDecision curve analysis\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eIVH\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eIntraventricular hemorrhage\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eSAH\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eSubarachnoid hemorrhage\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eCI\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eConfidence interval\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eOR\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eOdds ratio\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eRad-score\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eRadiomics score\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgments:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe extend our deepest appreciation to the patient volunteers and their families. We thank, Gansu Provincial Hospital, Lanzhou University Second Hospital and Xi\u0026rsquo;an Central Hospital (Department of Radiology) for freely providing data. We would like to thank Editage (www.editage.cn) for English language editing.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data that support the findings of this study are available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by the Guizhou Provincial Basic Research (Natural Science) Program (grant number QKHJC [2024] youth 319; recipient: Xiaoyu Huang).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions:\u003c/strong\u003e All authors attest that they meet the current International Committee of Medical Journal Editors (ICMJE) criteria for Authorship.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAl-Kawaz MN, Hanley DF, Ziai W (2020) Advances in Therapeutic Approaches for Spontaneous Intracerebral Hemorrhage. Neurotherapeutics 17:1757\u0026ndash;1767. https://doi.org/10.1007/s13311-020-00902-w\u003c/li\u003e\n\u003cli\u003ePuy L, Parry-Jones AR, Sandset EC, et al (2023) Intracerebral haemorrhage. Nat Rev Dis Primers 9:14. https://doi.org/10.1038/s41572-023-00424-7\u003c/li\u003e\n\u003cli\u003eTsao CW, Aday AW, Almarzooq ZI, et al (2023) Heart Disease and Stroke Statistics-2023 Update: A Report From the American Heart Association. Circulation 147:e93\u0026ndash;e621. https://doi.org/10.1161/CIR.0000000000001123\u003c/li\u003e\n\u003cli\u003eO\u0026rsquo;Carroll CB, Brown BL, Freeman WD (2021) Intracerebral Hemorrhage: A Common yet Disproportionately Deadly Stroke Subtype. Mayo Clin Proc. https://doi.org/10.1016/j.mayocp.2020.10.034\u003c/li\u003e\n\u003cli\u003eGreenberg SM, Ziai WC, Cordonnier C, et al (2022) 2022 Guideline for the Management of Patients With Spontaneous Intracerebral Hemorrhage: A Guideline From the American Heart Association/American Stroke Association. Stroke 53:e282\u0026ndash;e361. https://doi.org/10.1161/STR.0000000000000407\u003c/li\u003e\n\u003cli\u003eMorotti A, Arba F, Boulouis G, Charidimou A (2020) Noncontrast CT markers of intracerebral hemorrhage expansion and poor outcome: A meta-analysis. Neurology 95:632\u0026ndash;643. https://doi.org/10.1212/WNL.0000000000010660\u003c/li\u003e\n\u003cli\u003eSong Z, Tang Z, Liu H, et al (2021) A clinical-radiomics nomogram may provide a personalized 90-day functional outcome assessment for spontaneous intracerebral hemorrhage. Eur Radiol 31:4949\u0026ndash;4959. https://doi.org/10.1007/s00330-021-07828-7\u003c/li\u003e\n\u003cli\u003eDavis SM, Broderick J, Hennerici M, et al (2006) Hematoma growth is a determinant of mortality and poor outcome after intracerebral hemorrhage. Neurology 66:1175\u0026ndash;1181. https://doi.org/10.1212/01.wnl.0000208408.98482.99\u003c/li\u003e\n\u003cli\u003eBrott T, Broderick J, Kothari R, et al (1997) Early hemorrhage growth in patients with intracerebral hemorrhage. Stroke 28:1\u0026ndash;5. https://doi.org/10.1161/01.str.28.1.1\u003c/li\u003e\n\u003cli\u003eDavis SM, Broderick J, Hennerici M, et al (2006) Hematoma growth is a determinant of mortality and poor outcome after intracerebral hemorrhage. Neurology 66:1175\u0026ndash;1181. https://doi.org/10.1212/01.wnl.0000208408.98482.99\u003c/li\u003e\n\u003cli\u003eLi S, Cui L-Y, Anderson C, et al (2019) Public Awareness of Stroke and the Appropriate Responses in China: A Cross-Sectional Community-Based Study (FAST-RIGHT). Stroke 50:455\u0026ndash;462. https://doi.org/10.1161/STROKEAHA.118.023317\u003c/li\u003e\n\u003cli\u003eZhang J (2024) Harnessing radiomics to predict hematoma expansion in intracerebral hemorrhage: a step towards personalized care. Eur Radiol. https://doi.org/10.1007/s00330-024-11014-w\u003c/li\u003e\n\u003cli\u003eFeng C, Ding Z, Lao Q, et al (2024) Prediction of early hematoma expansion of spontaneous intracerebral hemorrhage based on deep learning radiomics features of noncontrast computed tomography. Eur Radiol 34:2908\u0026ndash;2920. https://doi.org/10.1007/s00330-023-10410-y\u003c/li\u003e\n\u003cli\u003eHuang X, Wang D, Zhang Q, et al (2022) Radiomics for prediction of intracerebral hemorrhage outcomes: A retrospective multicenter study. Neuroimage Clin 36:103242. https://doi.org/10.1016/j.nicl.2022.103242\u003c/li\u003e\n\u003cli\u003eZhou Y, Wang Y, Wang J, et al (2014) Inflammation in intracerebral hemorrhage: from mechanisms to clinical translation. Prog Neurobiol 115:25\u0026ndash;44. https://doi.org/10.1016/j.pneurobio.2013.11.003\u003c/li\u003e\n\u003cli\u003eIglesias-Rey R, Rodr\u0026iacute;guez-Y\u0026aacute;\u0026ntilde;ez M, Arias S, et al (2018) Inflammation, edema and poor outcome are associated with hyperthermia in hypertensive intracerebral hemorrhages. Eur J Neurol 25:1161\u0026ndash;1168. https://doi.org/10.1111/ene.13677\u003c/li\u003e\n\u003cli\u003ed-Prostaglandin J2 activates peroxisome proliferator-activated receptor-gamma, promotes expression of catalase, and reduces inflammation, behavioral dysfunction, and neuronal loss after intracerebral hemorrhage in rats - PubMed. https://pubmed.ncbi.nlm.nih.gov/16208315/. Accessed 15 Feb 2023\u003c/li\u003e\n\u003cli\u003eWang J, Tsirka SE (2005) Contribution of extracellular proteolysis and microglia to intracerebral hemorrhage. Neurocrit Care 3:77\u0026ndash;85. https://doi.org/10.1385/NCC:3:1:077\u003c/li\u003e\n\u003cli\u003eWang J, Dor\u0026eacute; S (2007) Inflammation after intracerebral hemorrhage. J Cereb Blood Flow Metab 27:894\u0026ndash;908. https://doi.org/10.1038/sj.jcbfm.9600403\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"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":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Intracerebral Hemorrhage, Radiomics, Outcome Prediction, Timescales, Non-contrast Computed Tomography","lastPublishedDoi":"10.21203/rs.3.rs-7191002/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7191002/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003ePurpose\u003c/h2\u003e\u003cp\u003eTo establish an outcome prediction model for multiple timescales itracerebral hemorrhage (ICH).\u003c/p\u003e\u003ch2\u003eMaterials and methods\u003c/h2\u003e\u003cp\u003eICH patients with an onset-to-imaging time (OIT) of less than 72 h were retrospectively collected. Patients were divided into three groups according to their OIT. Group 1\u0026ndash;3: OIT\u0026thinsp;\u0026lt;\u0026thinsp;6h, 6\u0026thinsp;\u0026le;\u0026thinsp;OIT\u0026lt;24h, 24\u0026thinsp;\u0026le;\u0026thinsp;OIT\u0026lt;72h. The first preoperative review in Group 1 within 72 h were recorded for Group 4. A binary logistic regression classifier was used to construct outcome prediction models for each group. The predictive performance of each group\u0026rsquo;s model was compared with the application of the Group 1 model across different groups to explore its accuracy and applicability in predicting the outcomes for patients in various groups.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eA total of 2,136 patients with ICH were included in the study. In the training set, the AUC value obtained by directly applying the group 1 radiomics model to group 2 patients was 0.773. The AUC value for the direct application of the Group 1 combined model to Group 3 patients was 0.811. The AUC for applying the Group 1 radiomics model directly to Group 4 patients was 0.779. The AUC of the radiomics model, constructed by combining the key radiomics features of Groups 1 and 4, was 0.788. The AUC of the independent radiomics model for Group 4 was 0.815.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e\u003cp\u003eThe radiomics and combined models for predicting outcomes in ICH patients with OIT\u0026thinsp;\u0026lt;\u0026thinsp;6 h can be applied to all patients with OIT\u0026thinsp;\u0026lt;\u0026thinsp;72 h, allowing for early and accurate outcome prediction across multiple timescales.\u003c/p\u003e","manuscriptTitle":"Functional Outcome Prediction Across Multiple Timescales Intracerebral Hemorrhage Using a Radiomics Model","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-03 08:44:38","doi":"10.21203/rs.3.rs-7191002/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"1c164968-db6c-4efd-9ef6-ceac9148847b","owner":[],"postedDate":"September 3rd, 2025","published":true,"recentEditorialEvents":[{"type":"decision","content":"Withdrawn","date":"2026-05-06T12:08:24+00:00","index":"","fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":54013836,"name":"Health sciences/Medical research"},{"id":54013837,"name":"Health sciences/Neurology"}],"tags":[],"updatedAt":"2026-05-06T12:26:24+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-03 08:44:38","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7191002","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7191002","identity":"rs-7191002","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}