Monitoring intracranial pressure (ICP) constitutes an important strategy for the management of various brain damage, including traumatic brain injury, intracranial hemorrhage, and severe cerebral infarction [1, 2]. While direct measurement of ICP requires invasive placement of a parenchymal sensor [3], accumulating evidence suggests that transcranial ultrasound examinations serve as an alternative noninvasive modality to evaluate ICP, which assesses the blood flow velocity of the middle cerebral artery (MCA) and the pulsatility index (PI) through the temporal window [4, 5]. Accurate insonation through the temporal window, however, requires substantial training because of the technical complexity [6–8]. Additionally, the acoustic window is absent in approximately 10 to 20% of patients due to skull thickness and variations in age and sex [9, 10]. Postoperative wounds or surgical drains may also hinder probe placement at the temporal window.

From a medicotechnical point of view, insonation of the extracranial internal carotid artery (ICA) appears feasible in all patients when applied through the submandibular window, and is unaffected by bone interference between the probe and the vessel [11]. Indeed, previous studies demonstrated that ICA-PI was correlated with MCA-PI in patients with ischemic stroke [12, 13]. Nevertheless, the relationship between the ICA-PI and MCA-PI has not been systematically validated across a broad range of acute intracranial diseases encountered in neurocritical care. By establishing a definite correlation between the ICA-PI and MCA-PI, the ICA-PI obtained via the submandibular window could offer a widely applicable and technically simpler method for non-invasive ICP estimation, particularly when the transtemporal window is unavailable.

We therefore aimed to evaluate whether the ICA-PI measured through the submandibular window could serve as a surrogate for the MCA-PI obtained through the temporal window. This approach may offer novel ICP management strategies and guide the timing of neuroimaging or surgical interventions [14].

Bedside TCCFI examinations were performed by Neuro ICU physicians and trained nurse practitioners using a 2-MHz sector probe (CX50, Philips, Netherlands). The MCA and ICA insonations were sequentially obtained during the same examination [6, 11]. For the submandibular ICA approach, the probe was placed beneath the mandibular angle and directed toward the external auditory canal, and color Doppler was used to differentiate the ICA from the external carotid artery (Fig. 1). The flow velocities were measured at depths of approximately 3–5 cm. The measurements were automatically calculated from the Doppler signals by the ultrasound system software. All recordings were digitally stored in electronic health records. The reported normal ranges for MCA mean flow velocity are 55–80 cm/s and 0.6-1.0 for PI [6, 15].

The following variables were extracted from the medical records: age, sex, height, weight, primary diagnosis, medical history (hypertension, diabetes mellitus, atrial fibrillation, and dyslipidemia), vital signs closest to the time of TCCFI (heart rate, mean arterial pressure, and body temperature), laboratory results on the day of TCCFI, Doppler flow parameters (velocities and PI), and ICP values if an invasive sensor was in place. Additional variables previously reported to influence the TCCFI measurements, such as arterial carbon dioxide tension (PaCO₂) and carotid artery calcification, were also collected [6].

There was observed a strong correlation between MCA-PI and ICA-PI (r = 0.894, p < 0.001, Fig. 2A). The Bland-Altman analysis demonstrated a mean difference of -0.005 (standard deviation, 0.177) without systematic bias across the PI values (Fig. 2B). The diagnostic performance of ICA-PI for detecting MCA-PI (> 1.3) demonstrated high sensitivity (88.6%), specificity (91.4%), PPV (81.6%), and NPV (94.9%) (Table 2B). When the regression line was divided at MCA-PI of 1.3, the slope coefficients did not differ between these two lines (p = 0.088, Supplementary Fig. 1). Likewise, the subgroup analysis showed strong correlations across diagnoses: SAH (r = 0.883), ICH (r = 0.903), cerebral infarction (r = 0.945), traumatic brain injury (r = 0.720), and hypoxic brain injury (r = 0.930) (Table 3).

Logistic regression analyses identified lower ICA-PI relative to MCA-PI in patients with hemoglobin ≥ 9 g/dL (odds ratio [OR] 2.44, 95% confidence interval [CI] 1.03–6.22, Fig. 3A) and in those with ICH (OR 3.14, 95% CI 1.19–8.96, Fig. 3B). No significant association was observed between ICA-PI and MCA-PI for any of other baseline characteristics (e.g., age, sex) or causes of brain damage (e.g., SAH, traumatic brain injury). When adjusted with hemoglobin as a potentially confounding factor, the impact of ICH on ICA-PI/MCA-PI association was diminished (OR = 2.44 [95%CI: 0.83–7.83], p = 0.104, Supplementary table 1).

The present study evaluated whether the ICA-PI could serve as a surrogate for the MCA-PI. Thus, the ICA-PI assessed via the submandibular window was found to closely mirror the MCA-PI measured via the transtemporal window across diverse neurocritical illnesses, with consistently strong correlations observed within their respective disease subgroups (Fig. 2, Table 3). These data not only coincide with prior observations in patient with ischemic stroke [12, 13], but also extend the premise that the ICA-PI may represent ICP, a dynamic marker for managing brain damage [19], as an alternative to the MCA-PI in various types of brain damage.

The close correlation between ICA-PI and MCA-PI can be attributed to the anatomical relationship between these vessels. Thus, both ICA and the MCA belong to the anterior circulation, with the perfusion territory of the ICA largely encompassing that of the MCA. Of course, in the acute phase of brain injury, individual transcranial ultrasound measurements can be influenced by multiple factors, including the insonated vessel, lesion location and severity, impairment of cerebral autoregulation, pre-existing small vessel disease, determinants of vascular resistance, and the vascular territory selected for measurement [6, 20–22]. Despite these potential confounders, the following two reasons may account for the observed correlation of PI. First, PI is calculated as the difference between peak systolic and end-diastolic velocities divided by the time-averaged maximum velocity [23]. Given that PI is expressed as a ratio, the impact of such diverse factors is relatively attenuated. Moreover, this property renders PI independent of probe angle, carrier frequency, and the velocity of sound in tissues [24], thereby mitigating several technical limitations compared with raw flow velocity measurements. Second, although the ICA follows a tortuous course through the skull base and cavernous sinus, and PI may vary slightly across its segments, previous reports have demonstrated no substantial differences in PI values between the extracranial ICA and the terminal intracranial ICA [20]. Taken together, these considerations provide a compelling anatomical and physiological explanation for the correlation of PI values between the ICA and MCA.

Of note, a tendency for PI values in the ICA to be lower than those in the MCA was observed among patients with blood hemoglobin ≥ 9 g/dL (Fig. 3A). We explored this threshold because prior studies have suggested a hemoglobin cut-off of 9 g/dL to be associated with neurological outcomes [16–18]. This finding is reasonably explained by blood viscosity- and vessel caliber-related differences. Higher blood viscosity is reportedly associated with increased PI values [6, 25]. As hemoglobin levels rise, blood viscosity increases; if the cerebral blood flow remains constant, the increase in viscosity elevates vascular resistance. As the MCA displays a smaller caliber than the ICA, these hemodynamic effects are more likely to be susceptible, as would be expected from Poiseuille’s law.

The causes of brain injury may also contribute to the difference in PI between ICA and MCA. Thus, we observed a lower ICA-PI in patients with ICH (Fig. 3B). Although this effect is attributed in part to the hemoglobin level (Supplementary table 1), a hematoma per se could increase local vascular resistance through mass effect. However, as the ICA supplies not only the MCA but also the anterior cerebral artery, compensatory flow might explain why ICA-PI values tended to be lower than those of the MCA. In contrast, this pattern was not evident in SAH despite both conditions involving intracranial bleeding. The discrepancy might be attributable to secondary pathophysiological processes unique to SAH, including delayed cerebral ischemia mediated by large-vessel vasospasm, microvascular constriction, microthrombosis, and cortical spreading depolarizations [19], which are not typically apparent in ICH.

When validated for clinical applications, our findings support the feasibility of ICA-PI estimation for ICP assessment, even in patients without an implanted ICP sensor or when the MCA is not accessible via the temporal bone window. This approach may have meaningful implications for the management of the patients with neurological examination-related difficulties and otherwise limited monitoring options. Contemporary paradigms in ICP management emphasize not only single threshold values but also the ICP intensity-time burden [26, 27]. Hence, repeated bedside noninvasive ICP monitoring using the submandibular window may be a reasonable and clinically practical strategy.

The MCA-PI and ICA-PI demonstrated strong correlations in patients requiring neurocritical care. ICA-PI measured using the submandibular window may serve as a practical surrogate for MCA-PI, enabling bedside noninvasive estimation of ICP and cerebral perfusion in critically ill patients. These findings suggest that the ICA-PI offers a feasible alternative to the MCA-PI even when a temporal window is unavailable. Further research is needed, however, to demonstrate the generalizability of these results and particularly to clarify whether similar outcomes can be achieved in settings where variability in measurement techniques is anticipated across multiple facilities.