Volumetric blood cutoff for outcome prediction in aneurysmal subarachnoid hemorrhage using semiautomated quantification Authors: Gema Bravo-Garrido, Antonio José Vargas-Lopez, Miriam Fernández-Gómez, Mario Gomar-Alba, Gaizka Urreta-Juárez, Laura Amaya-Pascasio, Patricia Martínez-Sánchez Published: 2025 ABSTRACT AbstractPurposeHemorrhage volume is a major prognostic factor in aneurysmal subarachnoid hemorrhage (aSAH). This study investigates the association between semiautomatically measured hemorrhage volumes and functional outcomes at discharge.MethodsA retrospective cohort study was conducted including patients with aSAH admitted to a tertiary hospital between 2016 and 2021. Hemorrhage volumes—including subarachnoid (SAH), intraventricular (IVH), intraparenchymal (IPH), and total hemorrhage (TH)—were quantified from admission CT scans using semiautomated segmentation software (AW Server). Functional outcome was assessed with the Glasgow Outcome Scale (GOS) at discharge, dichotomized as favorable (GOS 4–5) or unfavorable (GOS ≤ 3). ROC curves were used to identify optimal volume thresholds, and multivariate logistic regression assessed independent associations.ResultsA total of 170 patients were included (65.3% women, mean age 58.5 years). Unfavorable outcomes were observed in 80 patients (47.1%). ROC analysis identified a TH volume > 29 cm3 as the optimal threshold for predicting poor outcomes, with an AUC of 0.791 (95% CI 0.721–0.861), sensitivity of 75.0%, and specificity of 73.3%. Other volume parameters (SAH, IVH, IPH) showed lower discriminatory power. In multivariate analysis, TH > 29 cm3 remained independently associated with unfavorable outcomes (OR: 3.872; 95% CI 1.704–8.799; p = 0.001), after adjusting for potential confounders.ConclusionsSemiautomated volumetric analysis of hemorrhage in aSAH identifies TH volume as a key predictor of unfavorable outcomes. TH volume could serve as a useful metric for early risk stratification and support clinical decision-making. Further prospective studies are needed to validate these findings and assess their broader applicability.Keywords: Aneurysmal subarachnoid hemorrhage, Hemorrhage volume, Glasgow outcome scale, Prognosis, Semiautomated measurement Introduction Aneurysmal subarachnoid hemorrhage (aSAH) is one of the most devastating forms of hemorrhagic stroke, with early mortality rates ranging from 30 to 40% [1–5]. Among survivors, functional outcomes remain a major clinical challenge, as secondary complications, particularly delayed cerebral ischemia (DCI), substantially increase morbidity and long-term disability [5–10]. Despite the widespread use of grading scales like the Fisher scale, their qualitative nature and interobserver variability limit predictive accuracy [11–14]. Recent evidence supports a direct and quantifiable relationship between hemorrhage volume and neurological outcomes in aSAH [3, 4, 6, 9–11]. Larger hemorrhage volumes may worsen secondary brain injury by increasing intracranial pressure, promoting vasospasm, and disrupting microvascular autoregulation [3, 6, 9, 12]. Despite this, precise volumetric thresholds that predict functional outcomes are still not well established. Semiautomated volumetric analysis of aSAH has shown promising correlations with neurological outcomes and mortality rates [15–27]. These tools utilize Hounsfield unit thresholds and three-dimensional segmentation algorithms to delineate blood from surrounding parenchyma across multiple compartments including subarachnoid, intraventricular, and intraparenchymal spaces [15–17, 20, 21, 24, 26, 27]. Compared to visual grading, semiautomated methods are more reproducible and may better capture the total hemorrhage burden, which has been associated with increased risk of seizures, hydrocephalus, and DCI [22, 27]. Despite these advances, there is no consensus on the prognostic thresholds for hemorrhage volume in aSAH. This study hypothesizes that larger hemorrhage volumes across all intracranial compartments are associated with poorer functional outcomes, as assessed by the Glasgow Outcome Scale (GOS) at discharge [28]. We aim to identify volumetric cutoffs, particularly for total hemorrhage (TH) volume, that predict unfavorable functional outcomes (GOS ≤ 3), with the goal of supporting early risk stratification and improving clinical decision-making. Methods Patient population A retrospective cohort study was conducted including patients with aSAH admitted to a tertiary referral university hospital (from January 2016 to December 2021) serving a population of 761,379 inhabitants [26]. Inclusion criteria were age ≥ 18 years, aneurysmal hemorrhage confirmed by brain computed tomography (CT) angiography or digital subtraction angiography (DSA). Exclusion criteria included non-aneurysmal hemorrhages, incomplete imaging data, or missing GOS assessments at discharge. Medical assessment and variables All patients were managed according to current international guidelines [5] as previously described [26]. On admission to the emergency department, neurological status was assessed using the Glasgow Coma Scale (GCS), the World Federation of Neurosurgical Societies (WFNS) grading system, and the Fisher scale. Systolic and diastolic blood pressure, as well as blood glucose levels, were documented at presentation. Additional baseline data included demographics and medical history: age, sex, arterial hypertension, dyslipidemia, diabetes mellitus, and substance use (alcohol and smoking). The Charlson Comorbidity Index and length of hospital stay (in days) were also recorded. Treatment-related variables included endovascular coiling, microsurgical clipping, external ventricular drainage (EVD), ventriculoperitoneal (VP) shunting, and decompressive craniectomy. The primary outcome was functional status at discharge, assessed using the Glasgow Outcome Scale (GOS) [8]. This five-point scale classifies outcome from best to worst as follows: 5—good recovery; 4—moderate disability; 3—severe disability; 2—persistent vegetative state; and 1—death. In accordance with international recommendations [29], GOS was dichotomized into favorable (scores 4–5) and unfavorable (scores ≤ 3) outcomes. Neuroimaging analysis The presence, size, quantity, and location of the aneurysm (anterior and posterior circulation) were determined using admission CT angiography (General Electric, Siemens, or Philips) or digital subtraction angiography (DSA) (Artis Zee biplane, Siemens) conducted within the first 72 h after symptom onset (Fig. 1). In patients with multiple CT scans within this time window, the earliest scan was used for the primary volumetric analysis. Furthermore, the bicaudate index (BCI), was measured on the admission brain-CT on the axial plane [26]. Intracranial blood volumes were analyzed on admission brain CT using the Advantage Workstation (AW) Server 3.2 (GE Healthcare, Chicago, IL), which enabled semiautomated supervised measurements, as previously described [26]. Briefly, DICOM-format CT images were loaded into the software, which used image tools to outline and quantify hemorrhage dimensions and density. The method involved selecting a region of interest (ROI) within the hemorrhage; adjacent pixels with similar intensity were automatically included. The observer adjusted thresholds and used contour tools to refine the ROI on each slice. In anatomically complex areas—such as near the skull base—windowing and slice-by-slice criteria adjustment improved delineation of subarachnoid hemorrhage. The software then summed the segmented regions across slices to calculate final volumes for subarachnoid (SAH), intraventricular (IVH), intraparenchymal (IPH), and total hemorrhage (TH) (Fig. 2a–f). Image analysis was performed by a blinded author. Interobserver agreement was high, as previously reported [26]. Computed Tomography Angiography demonstrating the ruptured anterior communicating artery aneurysm. Sagittal (A) and axial (B) CTA slices, along with an oblique 3D volume-rendered reconstruction (C), identify the anterior communicating artery aneurysm (yellow arrowheads) in a 44-year-old male with aneurysmal subarachnoid hemorrhage. The aneurysm is located at the midline, projecting slightly anteriorly and superiorly. CTA: Computed Tomography Angiography Comparison between baseline non-contrast CT and semiautomated segmentation of aneurysmal subarachnoid hemorrhage. Axial non-contrast CT slices (A–C) and their corresponding semiautomated segmentation overlays (D–F) illustrate hemorrhage distribution in a patient with aSAH. Subarachnoid hemorrhage is visualized as hyperdense material in the basal cisterns and sulci (A–C), and is segmented in red using a Hounsfield-based region-growing algorithm (D–F). This comparison highlights the capacity of the method to objectively delineate cisternal and cortical sulcal blood for volumetric analysis. aSAH: Aneurysmal Subarachnoid Hemorrhage Statistical analysis Statistical analyses were performed using IBM SPSS Statistics for Windows, Version 27.0 (IBM Corp., Armonk, NY, USA). Continuous variables were expressed as mean ± SD or median (IQR), and categorical variables as frequencies and percentages. Patients were stratified by discharge outcome (favorable vs. unfavorable). Between-group comparisons used the χ2 or Fisher’s exact test for categorical data and the Student’s t-test or Mann–Whitney U test for continuous data. Associations between hemorrhage volumes—subarachnoid (SAH), intraventricular (IVH), intraparenchymal (IPH), and total hemorrhage (TH)—and outcome were evaluated in three steps. First, ROC curves assessed discriminative ability (AUC) and optimal thresholds based on the Youden index. Model performance was also measured using the SPSS “Overall Model Quality” statistic (analogous to pseudo-R2). Second, multivariate logistic regression was used to identify independent predictors of unfavorable outcomes. Variables with p ≤ 0.20 and ≥ 10 cases per group were included. A backward stepwise approach was applied, with likelihood ratio tests to assess model fit. Variables whose exclusion altered the OR by ≥ 15% were treated as confounders. Third, a forward stepwise regression (likelihood ratio method) confirmed model robustness. Results are presented with 95% confidence intervals (CI). Statistical significance was set at p < 0.05. The study protocol was approved by the Almería Provincial Research Ethics Committee (approval number: 87/2022). Informed consent was waived due to the retrospective design and use of fully anonymized data. Results A total of 170 patients were included (65.3% women, mean age 58.5 years). Unfavorable functional outcomes were observed in 80 patients (47.1%) (Fig. 3). Baseline characteristics and treatment-related variables are shown in Table 1. Patients with unfavorable outcomes were older and had a higher prevalence of hypertension. They also presented with lower GCS scores and higher WFNS and Fisher grades on admission. Endovascular treatment was more frequent in the favorable outcome group, whereas decompressive craniectomy, external ventricular drainage (EVD), and ventriculoperitoneal (VP) shunting were more common in patients with unfavorable outcomes. Flow diagram illustrating the inclusion of patients in the study. GOS, Glasgow Outcome Scale; SAH, subarachnoid hemorrhage Table 1Basal data and neurosurgical treatmentsVariableTotal (N = 170)Favorable outcomes (N = 90)Unfavorable outcomes (N = 80)PDemographic data and comorbiditiesFemale, n (%)111 (65.3)59 (65.6)52 (65)0.939Age, mean (SD), years58.5 (14.6)55.19 (14.1)62.3 (14.3) < 0.001Charlson Comorbidity Index, median (IQR)2 (2)1 (2)3 (3) < 0.001Smoking, n (%)64 (37.6)40 (44.4)24 (30)0.052Alcoholism, n (%)21 (12.4)11 (12.2)10 (12.5)0.956Drugs, n (%)5 (2.9)3 (3.3)2 (2.5)0.748Hypertension, n (%)70 (41.2)30 (33.3)40 (50)0.028Dyslipidemia, n (%)47 (27.6)25 (31.3)22 (24.4)0.322Diabetes mellitus, n (%)27 (15.9)13 (14.4)14 (17.5)0.586Previous stroke, n (%)6 (3.5)2 (2.2)4 (5)0.422Atrial fibrillation, n (%)10 (5.9)3 (3.3)7 (8.8)0.193Coronary artery disease, n (%)8 (4.7)2 (2.2)4 (5)0.422Congestive heart failure, n (%)6 (3.5)1 (1.1)4 (5)0.189COPD, n (%)6 (3.5)3 (3.3)3 (3.8)0.999Renal failure, n (%)5 (2.9)2 (2.2)3 (3.8)0.667Liver disease, n (%)6 (3.5)2 (2.2)4 (5)0.422Severity and risk of vasospasm on admissionGCS on admission, median (IQR)14 (6)15 (1)10 (9) < 0.001GCS on admission ≤ 8, n (%)39 (22.9)7 (7.8)32 (40) < 0.001Fisher scale, median (IQR)4 (1)3 (2)4 (1) < 0.001Fisher grade 3–4, n (%)125 (73.5)52 (57.8)73 (91.3) < 0.001WFNS scale, median (IQR)2 (3)1 (1)4 (2) < 0.001WFNS grade 4–5, n (%)66 (38.8)14 (15.6)52 (65) < 0.001Blood Pressure on admissionSBP, median (IQR)157.5 (40.3)152 (29)161.5 (52)0.042DBP, median (IQR)88 (24.3)86.5 (28)90 (28)0.318Glycemia, median (IQR), mg/dL146.5 (62)127 (46)166.5 (63) < 0.001Prior treatmentsStatins, n (%)38 (22.4)17 (18.9)21 (26.3)0.250Antiplatelet agents, n (%)21 (12.4)7 (7.8)14 (17.5)0.054Anticoagulants, n (%)10 (5.9)3 (3.3)7 (8.8)0.193SSRIs, n (%)26 (15.3)15 (16.7)11 (13.8)0.598Neurosurgical treatmentEndovascular Coiling, n (%)146 (85.9)86 (95.6)60 (75) < 0.001Neurosurgical clipping, n (%)0 (0)0 (0)(0)–Decompressive craniectomy, n (%)14 (8.2)0 (0)14 (17.5) < 0.001EVD and/or VP shunt, n (%)65 (7.6)22 (24.4)43 (53.8) < 0.001*Glasgow Outcomes Scale Score 4–5. **Glasgow Outcomes Scale Score 1–3COPD Chronic Obstructive Pulmonary Disease; EVD External Ventricular Drainage; GCS Glasgow Coma Scale; IQR Interquartile Range; SD Standard Deviation; SDAHC shunt-dependent acute hydrocephalus; SIADH Syndrome of Inappropriate Antidiuretic Hormone Secretion; SSRIs Selective Serotonin Reuptake Inhibitors; VP Ventriculoperitoneal, WFNS World Federation of Neurosurgical Societies Scale Basal data and neurosurgical treatments *Glasgow Outcomes Scale Score 4–5. **Glasgow Outcomes Scale Score 1–3 COPD Chronic Obstructive Pulmonary Disease; EVD External Ventricular Drainage; GCS Glasgow Coma Scale; IQR Interquartile Range; SD Standard Deviation; SDAHC shunt-dependent acute hydrocephalus; SIADH Syndrome of Inappropriate Antidiuretic Hormone Secretion; SSRIs Selective Serotonin Reuptake Inhibitors; VP Ventriculoperitoneal, WFNS World Federation of Neurosurgical Societies Scale Table 2 summarizes in-hospital complications. Delayed cerebral ischemia, cerebral edema, and rebleeding were significantly more frequent in the unfavorable group. Systemic complications such as pneumonia, sepsis, and cardiac arrhythmias also occurred more frequently in this group. Median length of hospital stay was longer in patients with unfavorable outcomes, although the difference did not reach statistical significance. Table 2In-hospital complications and length of stayVariableTotal (N = 170)Favorable outcomes (N = 90)Unfavorable outcomes (N = 80)PNeurological complications DCI, n (%)51 (30.4)21 (23.3)30 (38.5)0.033 Brain infarction, n (%)31 (18.2)9 (10)22 (27.5)0.003 Brain edema, n (%)34 (20)5 (5.6)29 (36.3) < 0.001 SDAHC, n (%)55 (32.4)21 (23.3)34 (42.5)0.008 Re-bleeding, n (%)17 (10)3 (3.3)14 (17.5)0.002 Seizures, n (%)8 (4.7)1 (1.1)7 (8.8)0.027Systemic complications Respiratory tract infection, n (%)49 (28.8)12 (13.3)37 (46.3) < 0.001 Urinary tract infection, n (%)29 (17.1)12 (13.3)17 (21.3)0.171 Sepsis, n (%)19 (11.2)3 (3.3)16 (20) < 0.001 Anemia, n (%)56 (32.9)25 (27.8)31 (38.8)0.129 Arrhythmia, n (%)18 (10.6)3 (3.3)15 (18.8)0.001 Acute pulmonary edema, n (%)9 (5.3)2 (2.2)7 (8.8)0.058 SIADH, n (%)24 (14.1)12 (13.3)12 (15)0.755Length of stay, (median, IQR) days18 (19)13 (14)22.5 (38)0.095DCI Delayed Cerebral ischemia; SDAHC Shunt-Dependent Acute Hydrocephalus; SIADH Syndrome of Inappropriate Antidiuretic Hormone Secretion In-hospital complications and length of stay DCI Delayed Cerebral ischemia; SDAHC Shunt-Dependent Acute Hydrocephalus; SIADH Syndrome of Inappropriate Antidiuretic Hormone Secretion Neuroimaging findings are presented in Table 3. Total hemorrhage volume, as well as subarachnoid (SAH), intraventricular (IVH), and intraparenchymal (IPH) components, were significantly higher in patients with unfavorable outcomes. Aneurysm size was also greater in this group. Table 3Neuroimaging data and haemorrhage volumesVariableTotal (N = 170)Favorable outcomes (N = 90)Unfavorable outcomes (N = 80)PVascular territory* Anterior (vs. posterior) circulation, n (%)145 (85.3)74 (82.2)71 (88.8)0.230Aneurysm characteristics* Aneurysm size, median (IQR), mm6 (5.1)5 (4)7 (7)0.019 Multiple aneurysms, n (%)52 (30.6)28 (31.1)24 (30)0.875Bicaudate index** Bicaudate index, median (IQR)0.19 (0.08)0.2 (0.1)0.2 (0.1)0.052 Bicaudate index > 0.273 (43.5)34 (38.2)40 (50)0.123Volumetry of the haemorrhage** TH, median (IQR), cm328.5 (33.6)18.9 (21.1)44.6 (35.8) < 0.001 SAH, median (IQR), cm319.7 (27.8)16.5 (21.5)27.65 (39.8)0.001 IPH, median (IQR), cm30 (0)0 (0)0 (2.9) < 0.001 IVH, median (IQR), cm30 (7.2)0 (3.6)3.62 (14.6) < 0.001*Measured in CT angiography and digital subtraction angiography (DSA)**Measured in brain-CTIQR Interquartile Range; SAH subarachnoid hemorrhage; SDAHC shunt-dependent acute hydrocephalus; IPH intraparenchymal hemorrhage; IVH, intraventricular hemorrhage; TH total hemorrhage Neuroimaging data and haemorrhage volumes *Measured in CT angiography and digital subtraction angiography (DSA) **Measured in brain-CT IQR Interquartile Range; SAH subarachnoid hemorrhage; SDAHC shunt-dependent acute hydrocephalus; IPH intraparenchymal hemorrhage; IVH, intraventricular hemorrhage; TH total hemorrhage ROC curve analysis demonstrated that TH volume had the highest discriminatory capacity for predicting unfavorable functional outcomes at discharge, with an AUC of 0.791 (95% CI 0.721–0.861; p < 0.001). A total hemorrhage (TH) volume > 29 cm3 was identified as the optimal cutoff, with a sensitivity of 73.8% and a specificity of 73.3% for predicting unfavorable functional outcomes. The positive predictive value (PPV) was 71.1%, and the negative predictive value (NPV) was 75.9%. In comparison, subarachnoid (SAH), intraparenchymal (IPH), and intraventricular (IVH) hemorrhage volumes yielded lower AUCs of 0.646 (95% CI 0.561–0.731; p = 0.001), 0.617 (95% CI 0.532–0.703; p = 0.008), and 0.677 (95% CI 0.596–0.759; p < 0.001), respectively (Fig. 4a). COR curves for the prediction of unfavorable outcomes (Glasgow Outcome Scale Score ≤ 3) based on hemorrhage volumes. A Receiver operating characteristic (ROC) curves for subarachnoid (SAH), intraventricular (IVH), intraparenchymal (IPH), and total hemorrhage (TH) volumes in predicting unfavorable functional outcomes at discharge (GOS ≤ 3). TH volume showed the highest discriminatory capacity, with an area under the curve (AUC) of 0.791 (95% CI 0.721–0.861; p < 0.001). AUCs for SAH, IPH, and IVH volumes were 0.646 (95% CI 0.561–0.731; p = 0.001), 0.617 (95% CI 0.532–0.703; p = 0.008), and 0.677 (95% CI 0.596–0.759; p < 0.001), respectively. B Overall model quality (pseudo-R2) from univariate logistic regression. TH volume achieved the highest predictive value (0.72), followed by IVH (0.60), SAH (0.56), and IPH (0.53). The vertical dashed line indicates the reference threshold (0.50). IVH: intraventricular hemorrhage volume (cm3); IPH: intraparenchymal hemorrhage volume (cm3); LR: logistic regression model; SAH: subarachnoid hemorrhage volume (cm3); TH: total hemorrhage volume (cm3) Model performance assessed using pseudo-R2 metrics supported the superior predictive performance of TH volume compared to individual hemorrhage compartments (Fig. 4b). Multivariate logistic regression confirmed TH > 29 cm3 as an independent predictor of unfavorable outcomes (OR: 3.872; 95% CI 1.670–9.774; p = 0.001), after adjusting for potential confounders (Table 4). Notably, in patients with 6-month follow-up, total hemorrhage volume > 29 cm3 was strongly associated with unfavorable functional outcome (70.1% vs. 29.9%, p < 0.001). Table 4Multivariate analysis of factors related to unfavorable outcomes at dischargeVariablesCrude ORMultivariate model 1 Adjusted OR*Mutivariate model 2 Adjusted OR*OR (IC 95%)pOR (IC 95%)pOR (IC 95%)pDemographic data and comorbidities Age, years1.036 (1.013–1.060)0.002–––– Female sex0.976 (0.518–1.837)0.9393.451 (1.132–10.520)0.029–– Charlson Index1.407 (1.169–1.694) < 0.0011.319 (1.038–1.676)0.0241.432 (1.139–1.800)0.002 Smoking0.536 (0.284–1.010)0.054–––– Hypertension2.000 (1.076–3.717)0.028––––Severity and risk of vasospasm on admission Fisher grade 3–47.621 (3.157–18.394) < 0.001–––– WFNS grade 4–510.082 (4.848–20.965) < 0.0012.751 (1.041–7.265)0.0415.938 (2.522–13.981) < 0.001SBP on admission, mmHg1.009 (1.000–1.018)0.059––––Glycemia, mg/dl1.012 (1.005–1.019)0.001––––Neurosurgical treatment Endovascular Coiling0.140 (0.045–0.419) < 0.0010.137 (0.031–0.600)0.008–– EVD and/or VP shunt3.592 (1.873–6.890) < 0.001––––Neurological complications DCI2.054 (1.053–4.007)0.0352.757 (0.985–7.713)0.053–– SDAHC2.429 (1.256–4.697)0.008––––Systemic complications Respiratory tract infection5.593 (2.642–11.841) < 0.0013.587 (1.207–10.666)0.0224.040 (1.670–9.774)0.002 Urinary tract infection1.754 (0.780–3.943)0.174–––– Anemia1.645 (0.864–3.133)0.130––––Length of stay, days1.027 (1.011–1.043) < 0.0011.024 (1.003–1.045)0.027––Neuroimaging data Aneurysm size, mm1.073 (1.004–1.147)0.038–––– Bicaudate index > 0.21.618 (0.877–2.984)0.124––––Hemorrhage volumes TH, cm31.057 (1.036–1.077) < 0.0011.047 (1.020–1.076) < 0.001 TH > 29 cm37.726 (3.903–15.294) < 0.0013.872 (1.704–8.799)0.001 SAH, median (IQR), cm31.032 (1.015–1.049) < 0.001–––– IPH, median (IQR), cm31.094 (1.023–1.169)0.008–––– IVH, median (IQR), cm31.116 (1.058–1.178) < 0.001––––*Backward stepwise logistic regression. Multivariate model 1: adjusted for volumetry of the total hemorrhage as continuous variables. The goodness-of-fit of the final multivariable model, measured by the C-test, was 0.906. Multivariate model 2: adjusted for volumetry of the total hemorrhage as dichotomized variable. The goodness-of-fit of the final multivariable model, measured by the C-test, was 0.870DCI Delayed Cerebral ischemia; EVD External Ventricular Drainage; IPH intraparenchymal hemorrhage; IVH intraventricular hemorrhage; SAH Subarachnoid hemorrhage; SDAHC Shunt-Dependent Acute Hydrocephalus; TH total hemorrhage; VP Ventriculoperitoneal; WFNS World Federation of Neurosurgical Societies Scale Multivariate analysis of factors related to unfavorable outcomes at discharge *Backward stepwise logistic regression. Multivariate model 1: adjusted for volumetry of the total hemorrhage as continuous variables. The goodness-of-fit of the final multivariable model, measured by the C-test, was 0.906. Multivariate model 2: adjusted for volumetry of the total hemorrhage as dichotomized variable. The goodness-of-fit of the final multivariable model, measured by the C-test, was 0.870 DCI Delayed Cerebral ischemia; EVD External Ventricular Drainage; IPH intraparenchymal hemorrhage; IVH intraventricular hemorrhage; SAH Subarachnoid hemorrhage; SDAHC Shunt-Dependent Acute Hydrocephalus; TH total hemorrhage; VP Ventriculoperitoneal; WFNS World Federation of Neurosurgical Societies Scale Discussion In this study, we identified a TH volume threshold > 29 cm3 that demonstrated good sensitivity and specificity in predicting short-term unfavorable outcomes in patients with aSAH. This quantitative marker may serve as a practical metric for early risk stratification in clinical settings. To our knowledge, this is the first study to report a specific TH threshold associated with early outcome using a standardized semiautomated segmentation method integrated into routine clinical software. This enhances both the external validity and the translational applicability of our findings [26]. Our results are consistent with previous volumetric studies. Lagares et al. identified a TH cutoff > 20 cm3 as an independent predictor of poor functional outcome and mortality in a prospective cohort [21]. Kramer et al. showed that IVH volume > 20 cm3 was associated with increased risk of complications and poor prognosis [19]. More recently, Yuan et al. reported that total bleeding volume assessed by automated methods correlated with delayed cerebral ischemia and unfavorable outcomes, with an optimal cutoff near 26 cm3 [20]. The threshold identified in our study aligns with these prior values and further supports the clinical utility of volumetric quantification in aSAH. Semiautomated volumetric tools offer a reproducible and time-efficient alternative to manual segmentation. While manual methods remain the gold standard in terms of accuracy, they are resource-intensive and impractical in emergency contexts [15, 16]. In contrast, semiautomated methods reduce interobserver variability and enable rapid volumetric estimation with minimal operator input [22, 26]. Traditional grading scales such as the Fisher and Hijdra scores are widely used in clinical practice but are limited by their qualitative nature, interrater variability, and inability to capture complex spatial blood distribution [11, 17]. Recent comparative studies confirm that volumetric methods provide superior prognostic performance, especially when assessing risk of complications such as delayed cerebral ischemia, hydrocephalus, and poor neurological outcome [18, 22, 23]. In our cohort, semi-automated quantification of TH volume showed strong prognostic accuracy for unfavorable outcome (AUC 0.791, 95% CI 0.721–0.861), with higher discriminative ability than conventional qualitative grading. In line with this, Eagles et al. reported modest AUCs for the modified Fisher and modified Graeb scores in predicting delayed cerebral ischemia (0.608 and 0.618, respectively), with only a slight improvement when both were combined (0.647) [30]. Furthermore, when assessed against functional outcome, the modified Graeb score achieved AUCs of 0.65–0.68 for early poor outcome [31]. Overall, these findings suggest that semi-automated volumetry may provide improved prognostic discrimination compared with qualitative scales and could be considered for gradual integration into clinical workflows as tools become more widely available. Although the prognostic role of SAH and IVH volumes has been separately explored [19, 21], our findings support the integration of all bleeding compartments into a single volumetric metric. This unified TH measurement may better reflect the global burden of early brain injury, including effects related to raised intracranial pressure, parenchymal compression, and cerebrospinal fluid circulation impairment [6, 7, 10]. Previous work by our group also demonstrated an association between TH volume and shunt-dependent hydrocephalus, further supporting its clinical relevance [26]. Importantly, volumetric thresholds such as the one proposed in our study could contribute to standardized classification systems for aSAH and be incorporated into predictive models for outcome or decision-support tools [20, 23, 24]. Early identification of high-volume bleeds could help tailor monitoring intensity and inform discussions with patients and families about prognosis [5, 8, 9]. This study has several limitations. First, it is a retrospective single-center analysis, which may limit the generalizability of our findings. Second, treatment modality was not controlled for in the analysis. Third, although segmentation was performed using a standardized method, some degree of observer-dependent variation may persist. Future prospective multicenter studies are warranted to validate these findings, assess long-term predictive value, and evaluate the performance of volumetric analysis against emerging deep learning–based approaches [23]. Conclusions This study demonstrates that semiautomated volumetric quantification of aSAH using clinically integrated software is a feasible and objective method to assess hemorrhage burden. Our findings support the role of total intracranial blood volume as an early prognostic biomarker. A total hemorrhage volume above 29 cm3 was independently associated with unfavorable outcome, suggesting that this threshold may assist in early risk stratification and clinical decision-making. Abbreviations Aneurysmal subarachnoid hemorrhage Advantage workstation Computed tomography Computed tomography angiography Cerebrospinal fluid Delayed cerebral ischemia Digital subtraction angiography External ventricular drainage Glasgow Coma Scale Glasgow Outcome Scale Intraclass correlation coefficient Intraparenchymal hemorrhage volume Interquartile range Intraventricular hemorrhage volume Magnetic resonance imaging Odds ratio Receiver operating characteristic Subarachnoid hemorrhage Statistical Package for the Social Sciences Total hemorrhage volume Ventriculoperitoneal shunt World Federation of Neurosurgical Societies Scale Author contributions G.B.G., A.J.V.L., and P.M.S. wrote the main manuscript text. G.B.G., M.F.G., and P.M.S. prepared Figs. 1–4. A.J.V.L. conceived the original idea for the study. G.U.J., P.M.S., L.A.P. and G.B.G. performed the hemorrhage volume measurements. G.B.G., M.G.A., and M.F.G. verified the measurement accuracy. A.J.V.L., P.M.S., and M.G.A. conducted the statistical analysis and created the tables. All authors reviewed and approved the final manuscript. Funding RICORS-ICTUS network, Grant Number RD24/0009/0030, Spanish Ministry of Health-Carlos III Health Institute (ISCIII) and the Next Generation EU funds (Recovery and Resilience Plan). Data availability The dataset used in this study may be available upon request. Request should be made to the corresponding authors. Declarations Ethics approval and consent to participate This study protocol was reviewed and approved by Ethics Committee of Almeria (Almeria Provincial Research Ethics Committee, ethics approval number 87/2022). It was not necessary for written informed consent because of the nature of the study, in accordance with the national legislation and the institutional requirements. Informed consent Not applicable. Guarantor PMS. Statement on the use of generative artificial intelligence The authors declare that they have used generative artificial intelligence, specifically ChatGPT 4o, to improve the grammar, style, and clarity of the English language in the manuscript. No AI-based tools were employed to generate scientific content, analyze data, interpret results, or contribute to the conceptual development of the study. Competing interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Antonio José Vargas-Lopez, Email: ajvargas@ual.es. Patricia Martínez-Sánchez, Email: patrinda@ual.es. References Associated Data Data Availability Statement The dataset used in this study may be available upon request. Request should be made to the corresponding authors.