Introduction

Hypertensive cerebral hemorrhage is a common acute and critical condition in neurosurgery, and surgery is one of the main means of treating this disease1. Intracranial infection is a common postoperative complication of hypertensive cerebral hemorrhage that not only increases the difficulty of treatment and medical costs for patients but also poses a safety concern. Since the occurrence of postoperative intracranial infection is related to a variety of factors, such as surgical trauma, blood‒brain barrier disruption, indwelling drainage tubes, and a wide variety of causative agents (e.g., bacteria, fungi, and viruses), it poses a challenge in terms of clinical diagnosis and treatment2. Traditional diagnostic methods for intracranial infections, such as cerebrospinal fluid examination, have high accuracy for confirming intracranial infections. However, these examinations are complicated, and they are associated with risks such as the spread of skin infections, hemorrhage induced by damage to blood vessels, and damage to the cauda equina. Therefore, it is important to identify a noninvasive and repeatable method for predicting the occurrence of intracranial infections. Modern medicine has developed a variety of tests to achieve early prediction of intracranial infections, including cerebrospinal fluid examination and imaging. Among these methods, transcranial Doppler ultrasound (TCD) is a noninvasive, convenient and noninvasive ultrasound technique that has unique advantages for assessing intracranial blood flow3. Transcranial Doppler (TCD) utilizes ultrasound waves at specific frequencies to penetrate thinner cranial regions such as the temporal and occipital windows, measuring blood flow velocity and direction in major intracranial vessels including the middle cerebral artery, anterior cerebral artery, and posterior cerebral artery. Changes in parameters like blood flow velocity, directional patterns, and vascular resistance in these critical vessels reflect cerebral hemodynamics, thereby revealing cerebrovascular function. Research has demonstrated that TCD-derived blood flow parameters can predict intracranial pressure changes in critically ill neurological patients4.

Therefore, This study proposes a hypothesis: Patients with postoperative intracranial infections following hypertensive cerebral hemorrhage exhibit significantly elevated hemodynamic parameters of the middle cerebral artery (MCA) measured by TCD—specifically, systolic peak velocity (Vs), end-diastolic velocity (Vd), and mean velocity (Vm)—which may serve as effective predictors of postoperative intracranial infections. The primary objective of this study is to evaluate the clinical value of TCD parameters (Vs, Vd, Vm) in predicting postoperative intracranial infections in hypertensive cerebral hemorrhage patients. Secondary objectives include: (1) analyzing the distribution characteristics of pathogenic bacteria in postoperative intracranial infection patients; (2) exploring key factors influencing the occurrence of postoperative intracranial infections. The research findings are presented as follows.

Patients and methods

General information

In this study, 127 hypertensive cerebral hemorrhage surgical patients admitted to a hospital from April 2021 to March 2024 were selected as the study subjects and were divided into the infection group (26 patients) and the noninfection group (101 patients) according to whether they had intracranial infections after surgery. The inclusion criteria for patients were as follows: (1) clearly diagnosed with hypertensive cerebral hemorrhage4 on the basis of clinical manifestations, laboratory tests, imaging tests, etc.; (2) received surgical treatment and did not have other types of infections or death in the 7-day postoperative period; (3) did not have intracranial infections before the operation; (4) aged ≥ 18 years; (5) complete clinical data; and (6) patients and their families signed an informed consent form and agreed to participate in this study. The exclusion criteria were as follows: (1) other neurological infectious diseases; (2) combined malignant tumors; (3) contraindications to lumbar puncture; (4) severe cranial defects and hydrocephalus; or (5) other serious complications, such as cerebral hemorrhage and cerebral infarction. This study was reviewed and approved by the Medical Ethics Committee of the hospital.This study was approved by the Medical Ethics Committee of Changzhi People’s Hospital. All participants or their legal guardians provided written informed consent prior to enrollment. All methods were performed in accordance with the relevant guidelines and regulations.

Methodologies

The following clinical data were collected from all patients: sex, age, body mass index (BMI), smoking history, hemorrhage site, comorbid diabetes mellitus, preoperative Glasgow Coma Scale (GCS) score5, volume of cerebral hemorrhage, time spent in surgery, type of surgery, postoperative cerebrospinal fluid leakage, hypo-proteinemia, duration of drain retention, and bed rest time.The patient was placed in the lying position to complete the examination, and the peak systolic flow velocity (Vs), end-diastolic flow velocity (Vd), and mean flow velocity (Vm) of the bilateral middle cerebral arteries (MCAs) were recorded. After the TCD examination, lumbar puncture was carried out, a cerebrospinal fluid sampling device was used to collect approximately 1.5 ml of cerebrospinal fluid. This procedure was carried out in strict compliance with aseptic methods. The bacterial cultures were maintained under standardized conditions at 37 °C in a humidified 5% CO₂ incubator for 48–72 h. After the colonies had grown to a suitable state, the samples were purified, and an automated bioidentification system (Phoenix100, BD, USA) was used to identify the pathogenic bacteria.

Statistical processing

Statistical analyses were conducted using SPSS software. Continuous variables are presented as the means ± standard deviations (SDs) and were compared using Student’s t test. Categorical variables are reported as frequencies (%) and were compared using the χ2 test. Independent predictors were identified through univariate analysis followed by multivariate logistic regression modeling. The diagnostic performance of the TCD parameters (Vs, Vd, Vm) was evaluated using receiver operating characteristic (ROC) curve analysis, with the area under the curve (AUC) and 95% confidence intervals calculated. Statistical significance was defined as a two-tailed P < 0.05. In ROC curve analysis, the best cut-off value (Cut-off value) for each TCD blood flow parameter to predict intracranial infection was determined by calculating the Youden index (J = sensitivity + specificity − 1).

During the multivariate logistic regression analysis, we conducted correlation analyses and variance inflation factor (VIF) tests to address potential collinearity or interaction effects among Vs, Vd, and Vm. To further confirm that the combined use of Vs, Vd, and Vm in the model was not affected by multicollinearity, we performed collinearity diagnostics. As shown in the coefficient table (now included as Supplementary Table 1), all three TCD parameters exhibited variance inflation factors (VIFs) < 1.1, and tolerance values > 0.94, indicating negligible collinearity. Therefore, it was statistically appropriate to retain these parameters in the model simultaneously without the need for variable exclusion due to interaction.Clinical covariates such as preoperative GCS scores and drainage tube duration were also incorporated to control for potential confounding factors.

Results

Single-factor analysis of intracranial infections

A total of 127 patients were included in the analysis, 26 (20.47%) of whom developed postoperative intracranial infections. There were no statistically significant differences between the infection and noninfection groups in terms of sex (P = 0.612), age (P = 0.839), BMI (P = 0.473), location of hemorrhage (P = 0.862), smoking history (P = 0.645), presence of diabetes (P = 0.177), hemorrhage volume (P = 0.401), surgical duration (P = 0.169), type of surgery (P = 0.251), or hypoproteinemia (P = 0.465).

However, several variables were significantly different between the groups. The preoperative GCS score was significantly lower in the infection group, with a greater proportion of patients having a GCS score < 8 (53.85% vs. 12.87%, P < 0.001). The incidence of postoperative cerebrospinal fluid (CSF) leakage was also significantly greater in the infection group (50.00% vs. 19.80%, P = 0.002). Additionally, patients in the infection group were more likely to have a drainage tube retention time of ≥ 3 days (76.92% vs. 45.54%, P = 0.004) and bed rest duration of > 7 days (69.23% vs. 44.55%, P = 0.025). The TCD parameters also significantly differed between groups. The infection group had significantly elevated flow velocities, including (1) Vs: 141.29 ± 33.24 cm/s vs. 101.95 ± 34.85 cm/s, P < 0.001; (2) Vd: 66.70 ± 26.51 cm/s vs. 42.51 ± 20.51 cm/s, P < 0.001; and (3) Vm: 93.59 ± 29.63 cm/s vs. 68.52 ± 24.75 cm/s, P < 0.001 (Table 1).

Table 1 Single factor analysis of intracranial infection.
Full size table

These results suggest that lower preoperative neurological status, prolonged drainage and bed rest, postoperative CSF leakage, and altered cerebral hemodynamics are significantly associated with the development of postoperative intracranial infections.

Multivariate analysis of intracranial infections

The results indicated that the multivariate logistic regression analysis (Table 2; Fig. 1) demonstrated a significant association: For every 1-point increase in preoperative GCS score, the risk of postoperative intracranial infection decreased (OR 0.096, 95% CI 0.020–0.452, P = 0.003), confirming higher GCS scores as an independent protective factor against intracranial infections. Additionally, the duration of indwelling drainage tubes ≥ 3 days (OR 5.454, 95% CI 1.012–29.384, P = 0.048) and mean blood flow velocity Vm (OR  1.045, 95% CI 1.013–1.078, P = 0.006) were identified as independent risk factors for intracranial infections.Multivariate analysis identified a drainage tube retention duration ≥ 3 days as a strong risk factor for intracranial infection (OR 5.454), consistent with previous studies and underscoring the importance of early removal of unnecessary drainage tubes for infection prevention. A lower preoperative GCS score (as a continuous variable, with each point reduction increasing risk) was a significant risk factor, reflecting the close correlation between level of consciousness impairment and susceptibility to infection. Although mean blood flow velocity (Vm) was an independent risk factor (OR 1.045), its relatively small OR value suggests that each 1 cm/s increase in Vm corresponds to approximately a 4.5% higher infection risk. While this association is statistically significant, achieving clinically meaningful absolute risk changes at the individual patient level requires substantial variations in Vm.

Table 2 Multivariate logistic regression analysis results of intracranial infection.
Full size table
Fig. 1
figure 1

Multifactorial logistic regression analysis of intracranial infections.

Full size image

Predictive value of TCD for intracranial infections

The ROC curve (Fig. 2; Table 3) shows that the AUC values for predicting intracranial infection using individual blood flow parameters (Vs, Vd, Vm) were 0.808, 0.752, and 0.729 respectively. The combined prediction model achieved an AUC of 0.901. Furthermore, through multivariate logistic regression analysis, we identified independent predictors (preoperative GCS score [continuous variable], drainage tube retention time [≥ 3 days vs. < 3 days], and mean blood flow velocity Vm [continuous variable]) to establish a clinical prediction model. This model demonstrated superior predictive power (AUC = 0.942, 95% CI 0.896–0.988), with sensitivity of 80.8%, specificity of 93.1%, and a Jaccard index of 0.739.

Fig. 2
figure 2

ROC curve of TCD blood flow parameters and prediction model for intracranial infection.

Full size image
Table 3 Prediction efficacy of TCD blood flow parameters and predictive model for intracranial infection.
Full size table

Combining Vs, Vd, and Vm significantly improved the diagnostic accuracy. The combination model yielded an AUC of 0.901 (95% CI 0.827–0.974), with a sensitivity of 84.6% and specificity of 88.1%. Furthermore, the final predictive model, which integrated TCD indicators and clinical risk factors, achieved the highest predictive value, with an AUC of 0.942 (95% CI 0.896–0.988), a sensitivity of 80.8%, and a specificity of 93.1%.

These findings suggest that TCD parameters, especially when they are used in combination or when they are integrated into a predictive model, can serve as reliable noninvasive indicators for the early prediction of postoperative intracranial infections.

Distribution of causative agents of intracranial infections

Among the 34 pathogenic strains isolated from patients with postoperative intracranial infections, gram-positive bacteria accounted for the majority, representing 64.71% (22/34) of all the isolates. Among them, Staphylococcus aureus was the most common pathogen, isolated in 10 cases (29.41%), followed by S. epidermidis in 6 cases (17.65%), coagulase-negative staphylococci in 4 cases (11.76%), and other gram-positive organisms in 2 cases (5.88%).

Gram-negative bacteria accounted for 35.29% (12/34) of the isolates, with Pseudomonas aeruginosa identified in 4 cases (11.76%) (Table 4.)

Table 4 Composition ratio of pathogenic bacteria in patients with intracranial infections.
Full size table

These results indicate that gram-positive cocci, particularly Staphylococcus species, are the predominant pathogens responsible for postoperative intracranial infections, thus highlighting the importance of early identification and targeted antimicrobial therapy.

Discussion

Intracranial infection is a common postoperative complication among patients with hypertensive cerebral hemorrhage, with an incidence rate ranging from 3% to 25%6,7,8. The clinical manifestations of these infections include fever, headache, impaired consciousness, and increased intracranial pressure, which seriously affect the quality of life of patients. In this study, the incidence of intracranial infections was 20.47%, highlighting their prevalence and severity. In view of the serious threat of intracranial infections to patients’ health, this study focused on the predictive value of TCD parameters for intracranial infections. An in-depth analysis of the changes in TCD parameters is expected to provide new ideas for the diagnosis and treatment of intracranial infections and thus reduce the occurrence of related endpoint events.

In this study, the risk factors for postoperative intracranial infections in patients with hypertensive cerebral hemorrhage were thoroughly investigated. The results revealed that the preoperative GCS score was an independent protective factor against postoperative intracranial infection, while prolonger drain retention time and higher cerebral blood flow rates were independent risk factors for postoperative intracranial infection. When the preoperative GCS score, which is an important indicator of a patient’s state of consciousness, is lower than 8, the patient’s consciousness is typically severely impaired, which may lead to a decrease in the patient’s resistance to infection during and after surgery; in such cases, it is easier for pathogens to invade the intracranium and cause infection9,10. Prolonged drain retention increases the chance of bacterial growth around the drain, leading to retrograde entry of bacteria into the skull through the drain and causing infection. A study by Huang et al. also revealed that extraventricular drainage is more likely to increase the risk of pathogenic bacteria entering the skull and causing infection11. Previous studies have shown that cerebrospinal fluid leakage may lead to the communication of brain tissue with the external environment, which provides a pathway for bacterial invasion and is an independent risk factor for intracranial infection in neurosurgery patients12,13. The results of this study differ from those of previous studies, which may be explained by the fact that cerebrospinal fluid leakage is repaired in time after surgery, thus reducing the contact time between the cerebrospinal fluid and the outside world and thereby reducing the risk of infection. Hence, the risk of infection needs to be strictly controlled during surgery, with a focus on patients with low preoperative GCS scores. Furthermore, postoperatively, the drainage retention time should be reasonably set, and adequate evaluation and treatment should be carried out before removing the drains.

In recent years, several relevant studies have confirmed14,15,16,17 that TCD ultrasound has good predictive value for the prognosis of brain injury, suggesting that hemodynamic parameters can reflect the degree of cerebrovascular disorders in patients to a certain extent. The results of this study revealed that the levels of Vs, Vd, and Vm were significantly greater in patients in the infection group than in those in the noninfection group, suggesting that increased levels of these hemodynamic parameters are manifestations of abnormal cerebrovascular status in patients with intracranial infection. This may be the result of a combination of an inflammatory response, vasospasm, increased intracranial pressure, and a systemic inflammatory response18. First, in intracranial infections, pathogen attack leads to an inflammatory response in the intracranial blood vessels, resulting in swelling and thickening of the vessel walls and narrowing of the lumen, thus increasing resistance to blood flow. To overcome this resistance, the heart needs to pump more vigorously, thus increasing Vs. Additionally, owing to the increase in vascular resistance, Vd and Vm are correspondingly increased to ensure that enough blood can reach the brain tissue. Second, intracranial infections may lead to vasospasm, which causes temporary constriction or dilation of blood vessels. When vasospasm leads to vasoconstriction, the heart needs to pump faster to maintain the blood supply to the brain tissue, resulting in increased blood flow rate parameters such as Vs, Vd, and Vm. The inflammatory response and vasospasm increase intracranial pressure in patients, leading to impaired cerebrospinal fluid circulation and brain tissue edema. To maintain normal perfusion pressure to the brain tissue, the pumping effort of the heart increases, thereby increasing the blood flow rate. Additionally, intracranial infections may cause a systemic inflammatory response, resulting in a hyperdynamic state of the systemic vascular system. This hyperdynamic state increases the rate of blood flow throughout the body, including the rate of blood flow in the brain. Therefore, blood flow velocity parameters, such as Vs, Vd and Vm, increase accordingly in patients with intracranial infections. The results of the ROC curve analysis revealed that Vs, Vd, Vm and the combination of the three have high sensitivity and specificity for the early prediction of intracranial infections, suggesting that the TCD assay has a certain application value in the early diagnosis of intracranial infections. The TCD ultrasound can provide an objective diagnosis of intracranial infection through timely understanding of the changes in intracranial vascular resistance and blood flow. This study confirms that elevated MCA blood flow velocity parameters (Vs, Vd, Vm) in TCD tests are significantly associated with increased post-hypertensive cerebral hemorrhage infection risks. The combined or integrated application of these three parameters in predictive models incorporating clinical risk factors (low GCS scores, prolonged drainage tube duration) demonstrates high predictive value (AUC > 0.9), offering a non-invasive and convenient early warning tool for clinicians. However, it should be emphasized that elevated cerebral blood flow velocity is not a specific indicator of intracranial infection. Multivariate analysis results (e.g., OR = 1.045 for Vm) suggest that while statistically significant, the magnitude of absolute velocity changes requires cautious interpretation for individual risk assessment. Clinically, elevated TCD velocities should not be relied upon alone for diagnosing intracranial infection. TCD findings should serve as supplementary information, requiring comprehensive evaluation through clinical manifestations (e.g., fever, altered consciousness, meningeal irritation signs), inflammatory markers (e.g., blood/cerebrospinal fluid leukocytes, CRP, PCT), and definitive cerebrospinal fluid microbiological examination results. When TCD indicates a significant increase in blood flow velocity, clinicians should heighten vigilance for intracranial infections and actively pursue diagnostic evidence (e.g., lumbar puncture). It is noteworthy that the elevated MCA blood flow velocity observed in this study may also be influenced by non-infectious factors. For instance, common postoperative systemic inflammatory response syndrome (SIRS) or confirmed sepsis (even when the infection source is not intracranial), postoperative fever, pain stimulation, use of vasoactive drugs (especially pressor agents), and changes in arterial carbon dioxide partial pressure (PaCO2) (hypocapnia causing vasoconstriction and hypercapnia inducing vasodilation) may all significantly affect cerebral blood flow velocity. Although this study recorded relevant variables (such as systemic inflammatory markers, vasoactive drug use, and PaCO2) during data collection, these confounding factors were not fully eliminated in the multivariate model. Therefore, interpretation of TCD flow velocity results should fully consider the patient’s overall clinical status and physiological parameters.

In addition, through cerebrospinal fluid pathogen culture, this study revealed that the causative organisms of intracranial infections were mainly gram-positive cocci, which accounted for as many as 64.71% of the total. The detection rate of S. aureus was particularly prominent. This finding is consistent with previous reports in the literature19,20,21,22, suggesting that S. aureus may be one of the main causative agents of intracranial infections. S. aureus is highly pathogenic and capable of producing a variety of toxins and enzymes that disrupt the body’s defense mechanisms and lead to severe intracranial infections. Therefore, monitoring, prevention and control of S. aureus should be the focus of intracranial infection prevention and control. This study also revealed that among the gram-negative bacteria, A. baumannii was predominant, accounting for 17.65% of the bacteria. A. baumannii is a conditionally pathogenic bacterium that is widely found in the natural environment but is particularly common in hospital infections23,24. It is extremely viable and can cause a variety of serious infections, including bacteremia, pneumonia, and meningitis. In intracranial infections, the presence of A. baumannii may further increase the difficulty of treatment. Therefore, hospitals need to strengthen infection control and preventive measures to minimize the spread and infection of other pathogenic bacteria, such as A. baumannii. Clinics can develop individualized treatment plans for intracranial infections on the basis of the distribution characteristics of the causative organisms, considering the pathogenic bacterial species and drug resistance.

In summary, the TCD hemodynamic parameters Vs, Vd and Vm were significantly greater among hypertensive cerebral hemorrhage patients with postoperative intracranial infection than among patients without infection. Furthermore, the combination of these parameters can be used to predict the occurrence of intracranial infection. Moreover, the pathogenic bacteria associated with intracranial infections are mainly gram-positive cocci, among which S. epidermidis accounts for a relatively high percentage. The distribution characteristics of pathogenic bacteria have important guiding significance for the clinical treatment and prevention of intracranial infections.

Limitations

It should be noted that due to the inherent limitations of TCD technology, it cannot directly visualize infection sites. However, vascular changes caused by infections (such as vasospasm and increased intracranial pressure) can be captured by TCD, which introduces limitations in its application for intracranial infections—specifically, non-specific diagnosis, etiology-independent assessment, operator-dependent interpretation, and primary focus on major arteries. The study has the following limitations: First, TCD’s detection of cerebral blood flow velocity changes lacks specificity, as mentioned earlier, and is susceptible to interference from various physiological and pathological states (systemic inflammation, fever, PaCO₂, vasoactive drugs, etc.). Although we attempted to account for some confounding factors, residual heterogeneity may persist. Second, this single-center retrospective observational study had a relatively small sample size (n=127,26 infections), potentially introducing selection bias. While the established predictive models (particularly those incorporating TCD parameters) demonstrate good internal validity, their generalizability and stability require further validation through larger, multicenter prospective cohort studies. Third, our analysis focused solely on MCA-derived hemodynamic parameters (Vs, Vd, Vm) without considering additional indices such as pulsation index (PI) or resistance index (RI)—factors that may reflect vascular resistance and intracranial pressure. Future research could incorporate more comprehensive TCD metrics. Fourth, the positive rate of cerebrospinal fluid pathogen culture is influenced by multiple factors (such as antibiotic use), which may lead to biased estimates of pathogen distribution and composition. Finally, this study did not systematically evaluate all potential and known variables affecting intracranial infection risk (such as operating room environment, specific surgical procedures, perioperative antibiotic prophylaxis regimens). Given the limitations of retrospective studies, future prospective cohort designs should systematically collect TCD data, detailed clinical variables, laboratory parameters (including PaCO₂), and microbiological results at different pre-and postoperative time points. Advanced statistical methods (such as time-dependent analysis and machine learning modeling) should be employed for dynamic evaluation and prediction. This will help more accurately establish TCD’s value in predicting intracranial infections while controlling potential confounding factors. Future multicenter studies are warranted to validate these findings and further investigate the mechanistic associations among cerebral hemodynamics, blood-brain barrier disruption, and infection susceptibility.

Conclusions

This prospective cohort study demonstrated that TCD hemodynamic parameters are reliable and noninvasive predictors of postoperative intracranial infections in patients with hypertensive intracerebral hemorrhage. The combined evaluation of peak Vs, Vd, and Vm showed the highest diagnostic performance (AUC = 0.901), surpassing the predictive value of individual parameters. Clinically, prolonged drainage duration (> 72 h) and a preoperative GCS score < 8 were identified as independent and potentially modifiable risk factors for postoperative intracranial infection, thus warranting enhanced perioperative monitoring.

Microbiological analysis revealed a predominance of gram-positive bacteria (64.7%), with S. aureus accounting for the largest proportion (29.4%) of isolates. The notable detection of A. baumannii (17.6%), a multidrug-resistant gram-negative organism, underscores the growing challenge of nosocomial infections in neurosurgical settings.

Collectively, these findings support two key clinical implications: the integration of TCD-based hemodynamic monitoring into routine postoperative surveillance protocols to facilitate early, preclinical detection of intracranial infections and the development of empiric antibiotic strategies on the basis of local pathogen distribution, emphasizing anti-staphylococcal coverage and vigilant monitoring for resistant gram-negative organisms.