RESEARCH ARTICLE

Ultrasound non-invasive measurement of

intracranial pressure in neurointensive care: A
prospective observational study
Chiara Robba1,2*, Danilo Cardim3, Tamara Tajsic4, Justine Pietersen5, Michael Bulman5,
Joseph Donnelly3, Andrea Lavinio1, Arun Gupta1, David K. Menon1, Peter J.
A. Hutchinson6, Marek Czosnyka3,7
1 Neurosciences Critical Care Unit, Addenbrooke’s Hospital, University of Cambridge, Cambridge, United
Kingdom, 2 Department of Neuroscience, University of Genoa, Genoa, Italy, 3 Brain Physics Laboratory,
Division of Neurosurgery, Department of Clinical Neurosciences, Addenbrooke’s Hospital, University of
a1111111111 Cambridge, Cambridge, United Kingdom, 4 Division of Neurosurgery, Department of Clinical Neurosciences,
a1111111111 Addenbrooke’s Hospital, University of Cambridge, Cambridge, United Kingdom, 5 Department of
a1111111111 Anaesthesia, Addenbrooke’s Hospital, University of Cambridge, Cambridge, United Kingdom, 6 Division of
a1111111111 Neurosurgery, University of Cambridge Hospitals, Cambridge, United Kingdom, 7 Institute of Electronic
a1111111111 Systems, Warsaw University of Technology, Warsaw, Poland

* [email protected]

OPEN ACCESS Abstract

Citation: Robba C, Cardim D, Tajsic T, Pietersen J,
Bulman M, Donnelly J, et al. (2017) Ultrasound
non-invasive measurement of intracranial pressure Background
in neurointensive care: A prospective observational
study. PLoS Med 14(7): e1002356. https://doi.org/ The invasive nature of the current methods for monitoring of intracranial pressure (ICP) has
10.1371/journal.pmed.1002356 prevented their use in many clinical situations. Several attempts have been made to develop
Academic Editor: Martin Schreiber, Oregon Health methods to monitor ICP non-invasively. The aim of this study is to assess the relationship
and Science University, UNITED STATES between ultrasound-based non-invasive ICP (nICP) and invasive ICP measurement in neu-
Received: February 10, 2017 rocritical care patients.

Accepted: June 14, 2017

Published: July 25, 2017

Methods and findings
This was a prospective, single-cohort observational study of patients admitted to a tertiary
Copyright: © 2017 Robba et al. This is an open
access article distributed under the terms of the neurocritical care unit. Patients with brain injury requiring invasive ICP monitoring were con-
Creative Commons Attribution License, which sidered for inclusion. nICP was assessed using optic nerve sheath diameter (ONSD),
permits unrestricted use, distribution, and
venous transcranial Doppler (vTCD) of straight sinus systolic flow velocity (FVsv), and meth-
reproduction in any medium, provided the original
author and source are credited.
ods derived from arterial transcranial Doppler (aTCD) on the middle cerebral artery (MCA):
MCA pulsatility index (PIa) and an estimator based on diastolic flow velocity (FVd). A total of
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
445 ultrasound examinations from 64 patients performed from 1 January to 1 November
files. 2016 were included. The median age of the patients was 53 years (range 37–64). Median
Funding: DC and MC are partially supported by
Glasgow Coma Scale at admission was 7 (range 3–14), and median Glasgow Outcome
NIHR Brain Injury Healthcare Technology Co- Scale was 3 (range 1–5). The mortality rate was 20%. ONSD and FVsv demonstrated the
operative, Cambridge, UK. JD is supported by a strongest correlation with ICP (R = 0.76 for ONSD versus ICP; R = 0.72 for FVsv versus
Woolf Fisher Scholarship (NZ). PJAH is supported
ICP), whereas PIa and the estimator based on FVd did not correlate with ICP significantly.
by the National Institute for Health Research
Cambridge BRC as a Research Professor of Combining the 2 strongest nICP predictors (ONSD and FVsv) resulted in an even stronger
Neurosurgery. The funders had no role in study correlation with ICP (R = 0.80). The ability to detect intracranial hypertension (ICP  20 mm

PLOS Medicine | https://doi.org/10.1371/journal.pmed.1002356 July 25, 2017 1 / 16

 Ultrasound based non-invasive ICP

design, data collection and analysis, decision to Hg) was highest for ONSD (area under the curve [AUC] 0.91, 95% CI 0.88–0.95). The com-
publish, or preparation of the manuscript. bination of ONSD and FVsv methods showed a statistically significant improvement of AUC
Competing interests: DKM is a member of the values compared with the ONSD method alone (0.93, 95% CI 0.90–0.97, p = 0.01). Major
Editorial Board of PLOS Medicine. MC is honorary limitations are the heterogeneity and small number of patients included in this study, the
director of Medicam Ltd, a company producing
‘cranial access device’, used for invasive
need for specialised training to perform and interpret the ultrasound tests, and the variability
multimodal brain monitoring, including ICP. in performance among different ultrasound operators.
Abbreviations: ABP, arterial blood pressure; ARDS,
acute respiratory distress syndrome; aTCD, arterial
transcranial Doppler; AUC, area under the curve; Conclusions
CPP, cerebral perfusion pressure; CSF, Of the studied ultrasound nICP methods, ONSD is the best estimator of ICP. The novel
cerebrospinal fluid; FVd, diastolic flow velocity;
combination of ONSD ultrasonography and vTCD of the straight sinus is a promising and
FVdv, straight sinus diastolic flow velocity; FVm,
mean flow velocity; FVmv, straight sinus mean flow easily available technique for identifying critically ill patients with intracranial hypertension.
velocity; FVs, systolic flow velocity; FVsv, straight
sinus systolic flow velocity; GCS, Glasgow Coma
Scale; GOS, Glasgow Outcome Scale; ICP,
intracranial pressure; IQR, interquartile range; Author summary
MCA, middle cerebral artery; nICP, non-invasive
intracranial pressure; nICPFVsv, non-invasive
intracranial pressure estimator based on straight
Why was this study done?
sinus systolic flow velocity; nICPONSD, non-invasive
• Intracranial pressure (ICP) monitoring is necessary in many clinical scenarios.
intracranial pressure estimator based on optic
nerve sheath diameter; nICPONSD+FVsv, non-invasive • Invasive ICP methods are the gold standard, but have many contraindications. Never-
intracranial pressure estimator based on the
theless, non-invasive ICP (nICP) measurement is a poorly developed technique.
combination of optic nerve sheath diameter and
straight sinus systolic flow velocity; ONSD, optic • We present a study comparing ultrasound-based nICP measurement techniques with
nerve sheath diameter; PIa, middle cerebral artery the gold standard.
pulsatility index; ROC, receiver operating
characteristic; vTCD, venous transcranial Doppler.
What did the researchers do and find?
• In a cohort of 64 patients with brain injury, we compared invasive ICP measurement
with 3 different ultrasound-based measurements of nICP: optic nerve sheath diameter
(ONSD) ultrasonography, arterial transcranial doppler (aTCD)–derived methods, and
straight sinus systolic flow velocity (FVsv).
• We found that both optic nerve sheath diameter ultrasonography (ONSD) and straight
sinus systolic flow velocity (FVsv) are strongly correlated with invasive ICP. In addition,
the combination of these 2 nICP parameters (ONSD and FVsv) resulted in stronger cor-
relation with ICP.

What do these findings mean?

• A novel nICP monitoring method based on combined ONSD ultrasonography and
venous transcranial Doppler has shown promising results for the measurement of intra-
cranial pressure in patients with brain injury.
• This ultrasound-based method is low-cost, quick, and based on technology widely
available.
• Future prospective studies will be needed to validate these results.

PLOS Medicine | https://doi.org/10.1371/journal.pmed.1002356 July 25, 2017 2 / 16

 Ultrasound based non-invasive ICP

Introduction
Intracranial hypertension is a frequent and harmful complication of brain injury; it is an
important contributing factor for secondary brain injury, and its severity and duration have
been correlated with a fatal outcome [1,2].
A recent trial comparing an invasive intracranial pressure (ICP) monitoring protocol with
a protocol based on imaging and clinical examination found no significant differences in
patient outcome [3]. However, the trial has been criticised for being underpowered and for the
methodology used to measure and treat ICP. Thus, invasive monitoring and treatment of
intracranial hypertension is still widely recommended in the management of severely brain-
injured patients despite a paucity of randomized evidence [4].
Invasive ICP monitoring through an intraventricular catheter or intraparenchymal micro-
transducer continues to be the standard of care after severe traumatic brain injury, and should
be performed when indications are met [5]. Because the use of invasive transducers can cause
complications including infection or haemorrhage [6–8], reliable non-invasive ICP (nICP)
estimation would be helpful, especially in clinical situations where the risk–benefit balance of
invasive ICP monitoring is unclear or when ICP monitoring is not immediately available or is
even contraindicated [4]. Several non-invasive methods based on transcranial Doppler and
optic nerve sheath diameter (ONSD) ultrasound are gaining clinical popularity due to their
safety, availability, and reliability [8–13]. At present, the best accuracy for a non-invasive
method reported in the literature [14,15] has been demonstrated by 2-depth high-resolution
transcranial Doppler insonation of the ophthalmic artery. This method does not need calibra-
tion and is based on the measurement of the balance point when the measured parameters of
blood flow velocity waveforms in the intracranial segment of the ophthalmic artery (which
reflect ICP) are identical to extracranial segments (which are mechanically compressed by an
externally applied pressure). Other authors [16,17] have proposed different methods for con-
tinuous nICP monitoring based on the waveform analysis of cerebral blood flow velocity from
the middle cerebral artery (MCA) and arterial pressure. However, despite these promising
results, non-invasive techniques remain of insufficient accuracy and temporal resolution to
replace invasive ICP monitoring [18,19].
The aim of this study was to compare the accuracy of different ultrasound-based methods
for nICP measurement in patients with severe traumatic brain injury undergoing invasive ICP
monitoring. Such methods included the ultrasound measurement of the ONSD, venous tran-
scranial Doppler (vTCD), and derived indices obtained from the straight sinus (such as
straight sinus systolic flow velocity [FVsv]), and arterial transcranial Doppler (aTCD)–derived
indices such as middle cerebral artery (MCA) pulsatility index (PIa) and diastolic flow velocity
(FVd).

Methods
This is a single-centre, prospective observational study conducted from 1 January 2016 to 1
November 2016. Recruited patients were admitted at the Neurosciences Critical Care Unit,
Addenbrooke’s Hospital, Cambridge, UK. The protocol was approved by the research ethics
boards at the University of Cambridge (REC 15/lo/1918), and written consent was obtained
from all participants’ next of kin. The article is reported as per Strengthening the Reporting of
Observational Studies in Epidemiology (STROBE) reporting guidelines (S1 Text). Patients
older than 18 years requiring sedation, mechanical ventilation, and ICP monitoring with an
admission diagnosis of severe traumatic brain injury, aneurysmal subarachnoid haemorrhage,
intraparenchymal haemorrhage, or stroke were considered for inclusion. Exclusion criteria

PLOS Medicine | https://doi.org/10.1371/journal.pmed.1002356 July 25, 2017 3 / 16

 Ultrasound based non-invasive ICP

were the following: the absence of an informed consent, a known history of ocular pathology
or optic nerve trauma, skull base fracture with a cerebrospinal fluid (CSF) leak, inaccessible
ultrasound windows (temporal for aTCD and occipital for vTCD), and clinical or radiological
suspicion of cerebral venous thrombosis or vasospasm.

Patient monitoring
Patients were sedated with a continuous infusion of propofol or midazolam, fentanyl, and,
when necessary, the muscle relaxant atracurium besylate. Mechanical ventilation was tar-
geted to maintain adequate oxygenation (SaO2 > 90%) and normocapnia (PaCO2 = 35–40
mm Hg). Intravenous fluids, vasopressors, and inotropic support (norepinephrine and/or
epinephrine) were administered to achieve and maintain an adequate cerebral perfusion
pressure (CPP > 60 mm Hg). Clinical management was in accordance to international
guidelines [20–22].
Treatment of intracranial hypertension was based on a protocol-driven strategy, which
included optimisation of arterial blood pressure (ABP) and volaemia, sedation, and infusion of
hyperosmolar fluids according to our institutional guidelines.
After a decision to place an ICP monitoring device (by the neurosurgical and intensive care
physician in charge), patients were enrolled in the study. ICP was measured via an intrapar-
enchymal probe (Codman & Shurtleff, Raynham, Massachusetts, US) or a catheter inserted
into the brain ventricles and connected to an external pressure transducer and drainage system
(Codman, Johnson & Johnson, Raynham, Massachusetts, US).
For each patient, we collected the following characteristics: admission Glasgow Coma Scale
(GCS), age, sex, height, weight, comorbidities, mechanism and severity of brain injury, and
discharge Glasgow Outcome Scale (GOS). The Rotterdam and Marshall scores as well as the
Fisher scale were calculated using the admission computer tomography head scan reports [22].

Ultrasound measurements
Ultrasound measurement was performed by a selected group of experienced operators (TT, JP,
MB) using a standardised insonation technique to reduce inter-operator variability. The oper-
ators were blinded to the patient’s admission diagnosis, demographics, baseline characteristics,
and clinical and physiological background. Operators were not blinded to the actual ICP, but
they were blinded to the final formulae to obtain a nICP estimation from the different mea-
surements. Mean arterial blood pressure, end-tidal carbon dioxide partial pressure (ETCO2),
MCA flow velocities (diastolic [FVd], mean [FVm], and systolic [FVs]), straight sinus flow
velocities (diastolic [FVdv], mean [FVmv], and systolic [FVsv]), and ONSD were recorded twice
daily from day 1 to day 5 after ICP insertion. Additional measurements were performed in
case of acute increases in ICP (above 20 mm Hg). In cases where ICP mean values changed
more than ±2 mm Hg during any of the 3 studies (ONSD ultrasound, vTCD, and aTCD), the
measurements were excluded from the analysis.
ONSD. Ultrasound examination of the ONSD was performed using a 7.5-MHz linear
ultrasound probe (11L4, Xario 200; Toshiba, Zoetermeer, The Netherlands) using the lowest
possible acoustic power that could measure the ONSD. The probe was oriented perpendicu-
larly in the vertical plane and at around 30 degrees in the horizontal plane on the closed eyelids
of both eyes of individuals in supine position with head elevated to 30 degrees. Ultrasound gel
was applied on the surface of each eyelid and the measurements were made in the axial and
sagittal planes of the widest diameter visible 3 mm behind the retina in both eyes. The final
ONSD value was calculated by averaging 4 measured values, as previously described [23,24].

PLOS Medicine | https://doi.org/10.1371/journal.pmed.1002356 July 25, 2017 4 / 16

 Ultrasound based non-invasive ICP

Transcranial Doppler. aTCD was performed bilaterally on the MCA through the tempo-
ral window using a traditional 2-MHz transducer (5S2, Xario 20; Toshiba, Zoetermeer, The
Netherlands) with head elevated to 30 degrees, as previously described [23,25]. The final values
of flow velocities were calculated by averaging the 2 measured values.
vTCD was performed on the straight sinus using a 2-MHz transducer (5S2, Xario 20;
Toshiba, Zoetermeer, The Netherlands) through an occipital and transforaminal bone window
at a depth of 50 to 80 mm for flow directed toward the probe, as previously described [25].

Statistical analysis
On the basis of previous reports [25–27], we hypothesized that ICP is linearly associated with
ONSD, systolic flow velocity on the straight sinus (FVsv), PIa, and ABP × (1 − FVd/FVm), and
verified this hypothesis in 64 patients. A multivariable linear regression model was obtained
from the relationship among ICP, ONSD, and FVsv to derive an nICP estimator based on the
combination of ONSD and FVsv (nICPONSD+FVsv).
Deviations from the initial statistical plan (S2 Text) were based on reviewers’ requests and
consisted of inclusion of linear mixed effects model analysis for the determination of the esti-
mation formulas for nICP and exclusion of Bland–Altman analysis.
Statistical analysis of the data was conducted with RStudio software (R version 3.1.2). Ini-
tially, multiple measurement points were averaged for each patient; therefore, every patient
was represented by one single value for all variables assessed. Then, the correlations between
ICP and the variables of interest—ONSD, PIa, ABP × (1 − FVd/FVm), and FVsv—were verified
using the Pearson correlation coefficient (R, with the level of significance set at 0.05).
To provide prediction models for ICP estimation, the relationships between ICP and the
correlated variables were expressed as linear mixed effects models (R package lme4 [28]). As
fixed effects, we entered ICP and the non-invasive estimators into the model. As random
effects, we had intercepts and slopes for the repeated measurement points for each patient
(N = 445 measurements). A mixed effects multiple regression between ICP and 2 correlated
variables, ONSD and FVsv, was also performed. Chi-square (χ2) values and p-values for the
comparison of the models were obtained by likelihood ratio tests of the full model with ran-
dom intercepts and slopes against the null model with random intercepts only.
The area under the curve (AUC) of the receiver operating characteristic (ROC) curve was
calculated to determine the ability of the non-invasive methods to detect raised ICP (using a
threshold of 20 mm Hg; N = 445 measurements). Moreover, we also performed an analysis to
determine the best ONSD and FVsv cutoff values for prediction of ICP  20 mm Hg. In ROC
analysis, these are the values presenting the best sensitivity and specificity for prediction of a
given threshold. The predicting ability is considered reasonable when the AUC is higher than
0.7 and strong when the AUC exceeds 0.8 [29]. Statistical differences between ROC curves
were verified using the DeLong’s test for 2 correlated ROC curves (R package pROC [30]).
An analysis of variance (ANOVA) was performed to verify whether any of the variables
assessed were associated with mortality in the patient cohort.

Results
In all, 80 patients with intracranial pathology requiring invasive ICP monitoring were initially
considered for enrolment in this study. Among these, 3 were excluded because of the absence
of written consent, 2 because it was not possible to find a temporal window, 6 because the
occipital window was inaccessible (cervical collar or patient position), 3 because the straight
sinus could not be insonated, and 2 because of ocular lesions that precluded the assessment of
ONSD.

PLOS Medicine | https://doi.org/10.1371/journal.pmed.1002356 July 25, 2017 5 / 16

 Ultrasound based non-invasive ICP

A total of 445 recordings from 64 patients (each one including ONSD ultrasound, aTCD,
and vTCD) were included in the final analysis. The percentage of measurements presenting
ICP  20 mm Hg was 19.3% (N = 86). The characteristics of the patients are shown in Table 1.
In Table 2, we present the median (interquartile range [IQR]) values of the variables analysed.

nICP measurement
The correlation analysis between patients revealed good correlation between ICP and
ONSD (R = 0.76) and between ICP and FVsv (R = 0.72), averaged per patient (N = 64)—
both were statistically significant (p < 0.001) and without any influential outliers. The p-val-
ues for the correlations between ICP and PIa and ABP × (1 − FVd/FVm) were both non-sig-
nificant (p = 0.63 and p = 0.36, respectively). Thus, the regression formulas adopted in this
work considered only ONSD and FVsv, and the combination of both in a multiple regres-
sion model (Table 3).
Considering the variability in slopes between patients, full models allowing for random
intercepts and slopes were significantly better at fitting the data than null models for a nICP
estimator based on FVsv (nICPFVsv) and for nICPONSD+FVsv (χ2 = 44.19, p < 0.001, and χ2 =
40.92, p < 0.001, respectively). The inclusion of random slopes in the model describing a nICP
estimator based on ONSD (nICPONSD) did not produce a significant difference in comparison
to the model with random intercepts only (χ2 = 2.41, p = 0.30). The formulas of the derived

Table 1. Baseline characteristics of the patient cohort.

Characteristic N (%) or median (IQR)
Total number 64
Male sex 49 (76.6%)
Age (years) 53 (37–64)
Height (cm) 175 (165–180)
Weight (kg) 78 (67–87)
Pathology
Traumatic brain injury 45 (70.3%)
Aneurysmal subarachnoid haemorrhage 15 (23.4%)
Intracranial haemorrhage 4 (6.3%)
Comorbidities
Hypertension 7 (10.9%)
Depression 12 (18.7%)
Asthma 6 (9.4%)
Alcohol abuse 23 (35.9%)
Smokers 21 (32.8%)
Previous myocardial infarction 1 (3.1%)
GCS at admission 7 (3–14)
GOS at discharge 3 (1–5)
Complications
Chest infection 12 (18.7%)
Sepsis 2 (3.1%)
Ventriculitis 1 (1.5%)
Post-traumatic ARDS 2 (3.1%)

ARDS, acute respiratory distress syndrome; GCS, Glasgow Coma Scale; GOS, Glasgow Outcome Scale;
IQR, interquartile range.

https://doi.org/10.1371/journal.pmed.1002356.t001

PLOS Medicine | https://doi.org/10.1371/journal.pmed.1002356 July 25, 2017 6 / 16

 Ultrasound based non-invasive ICP

Table 2. Median (IQR) values of the studied parameters.

Parameter Definition Median (IQR)
ICP (mm Hg) Intracranial pressure 10 (5–17)
CPP (mm Hg) Cerebral perfusion pressure 79 (70–87)
ABPm (mm Hg) Mean arterial blood pressure 90 (85–100)
PaCO2 (mm Hg) Partial pressure of CO2 5.20 (4.90–5.90)
ONSD (mm) Optic nerve sheath diameter 4.9 (4.2–6.0)
FVs (cm/s) MCA systolic flow velocity (aTCD) 105 (97–114)
FVd (cm/s) MCA diastolic flow velocity (aTCD) 48 (42–56)
FVm (cm/s) MCA mean flow velocity (aTCD) 68.67 (59.67–75.00)
FVsv (cm/s) Straight sinus systolic flow velocity (vTCD) 30 (22–39)
FVdv (cm/s) Straight sinus diastolic flow velocity (vTCD) 13 (12–18)
FVmv (cm/s) Straight sinus mean flow velocity (vTCD) 18.67 (15.67–24.67)
PIa MCA pulsatility index (aTCD) 0.86 (0.67–1.04)

aTCD, arterial transcranial Doppler; IQR, interquartile range; MCA, middle cerebral artery; vTCD, venous transcranial Doppler.

https://doi.org/10.1371/journal.pmed.1002356.t002

models that best fitted the data are described below:

nICPONSD ¼ 5:00  ONSD 13:92 ðmm HgÞ ð1Þ

nICPFVsv ¼ 0:38  FVsv þ 0:0005 ðmm HgÞ ð2Þ

nICPONSDþFVsv ¼ 4:23  ONSD þ 0:14  FVsv 14:51 ðmm HgÞ ð3Þ

Accuracy of the nICP methods

The correlation coefficient between the ONSD method and ICP averaged per patient (N = 64)
was R = 0.76; the FVsv method showed a correlation with ICP of R = 0.72. The combination of

Table 3. Summary of the linear mixed effects models of ICP and the non-invasive estimators across all measurement points (N = 445).
Full model Null model
Estimate 95% CI p-Value Estimate 95% CI p-Value
nICPONSD
ONSD 4.90 4.39 to 5.38 <0.001 5.00 4.64 to 5.36 <0.001
Intercept −13.47 −16.05 to −10.84 — −13.92 −15.99 to −11.85 —
nICPFVsv
FVsv 0.38 0.29 to 0.47 <0.001 0.34 0.29 to 0.40 <0.001
Intercept 0.0005 −2.87 to 2.83 — 1.16 −0.82 to 3.13 —
nICPONSD+FVsv
ONSD 4.23 3.63 to 4.79 <0.001 4.54 4.14 to 4.95 <0.001
FVsv 0.14 0.07 to 0.22 <0.001 0.10 0.06 to 0.15 <0.001
Intercept −14.51 −16.82 to −12.19 — −14.79 −16.83 to −12.75 —

Full model accounts for random intercepts and slopes; null model accounts for random intercepts only.
FVsv, straight sinus systolic flow velocity; nICPONSD, non-invasive intracranial pressure estimator based on optic nerve sheath diameter; nICPFVsv, non-
invasive intracranial pressure estimator based on straight sinus systolic flow velocity; nICPONSD+FVsv, non-invasive intracranial pressure estimator based
on a combination of ONSD and FVsv; ONSD, optic nerve sheath diameter.

https://doi.org/10.1371/journal.pmed.1002356.t003

PLOS Medicine | https://doi.org/10.1371/journal.pmed.1002356 July 25, 2017 7 / 16

 Ultrasound based non-invasive ICP

Table 4. Correlations between ICP and non-invasive estimators across all measurement points
(N = 445) and for average values between patients (N = 64).
Estimator R for measurement points (N = 445) R for patients (N = 64)
ONSD 0.76 0.76
FVsv 0.54 0.72
nICPONSD+FVsv 0.78 0.81

FVsv, straight sinus systolic flow velocity; nICPONSD+FVsv, non-invasive intracranial pressure estimator
based on the combination of ONSD and FVsv; ONSD, optic nerve sheath diameter; R, Pearson correlation
coefficient.

https://doi.org/10.1371/journal.pmed.1002356.t004

the 2 methods presented a correlation coefficient of 0.81 (Table 4; Fig 1). S1 Fig displays the
regression plots between ICP and the non-invasive estimators (ONSD and FVsv) for each
patient, demonstrating the slope variability between patients with multiple measurement
points.
Table 5 summarises the 95% prediction and confidence intervals for the linear regressions
between ICP and all non-invasive estimators. The 95% prediction interval for ONSD ranged
from 5.05 ± 4.04 to 19.32 ± 4.17 mm Hg; the 95% confidence interval ranged from 11.03 ± 3.95
to 13.34 ± 4.30 mm Hg. The 95% prediction interval for FVsv ranged from 4.57 ± 3.83 to
19.79 ± 3.96 mm Hg; the 95% confidence interval ranged from 10.94 ± 3.71 to 13.43 ± 4.12
mm Hg. For the combination of the 2 methods (nICPONSD+FVsv), the 95% prediction interval
ranged from 5.65 ± 4.30 to 18.72 ± 4.45 mm Hg; the 95% confidence interval ranged from
10.90 ± 4.19 to 13.47 ± 4.60 mm Hg.
Results of ROC analysis are showed in Table 6 and Fig 2. ONSD had the best AUC among
all methods for discriminating cases with intracranial hypertension (ICP  20 mm Hg) from
cases without it (AUC = 0.91, 95% CI 0.88–0.95). The best ONSD and FVsv cutoff values for
prediction of intracranial hypertension were 5.85 mm and 38.50 cm/s, respectively. The
method based on the combination of ONSD and FVsv showed a statistically significant
improvement of AUC values compared with the ONSD method alone (0.93, 95% CI 0.90–
0.97, p = 0.01 [DeLong’s test]).

Mortality and nICP

The outcome assessed at discharge revealed that 13 patients died (20%) and 51 survived. Mean
ICP showed a tendency to be greater in patients who died; mean ONSD was greater in patients
who died than in those who survived (Table 7; Fig 3). FVsv was not significantly different
between survivors and non-survivors (p = 0.28).

Discussion
In this study, we present and compare new models for ultrasound-based non-invasive estima-
tion of ICP, based on ONSD ultrasonography, aTCD, and vTCD. Our results show that nICP
derived from ONSD has the strongest correlation with invasive ICP. Moreover, ONSD mea-
sured through ultrasound was correlated with mortality at discharge. Finally, we demonstrated
that a method based on the combination of the 2 best correlated parameters in our cohort
(ONSD and FVsv—nICPONSD+FVsv) performed even better across all measurement points
(R = 0.78; AUC for prediction of ICP  20 mm Hg was 0.93).
Measuring ONSD and FVsv using a duplex Doppler machine is fast and does not require
probe fixation or specific dedicated hardware [13]. Furthermore, ultrasonography devices are

PLOS Medicine | https://doi.org/10.1371/journal.pmed.1002356 July 25, 2017 8 / 16

 Ultrasound based non-invasive ICP

Fig 1. Scatterplot of ICP (mm Hg) and different nICP estimators between patients (N = 64). (A) ONSD method (R = 0.76); (B) FVsv (R = 0.72); (C) nICP
estimator based on the combination of ONSD and FVsv (nICPONSD+FVsv, R = 0.80). Dark grey shaded areas on the plots represent 95% confidence
intervals for the linear regressions; light grey shaded areas on the plots represent the 95% prediction intervals for the linear regressions. ICP, intracranial
pressure; nICP, non-invasive intracranial pressure; ONSD, optic nerve sheath diameter.
https://doi.org/10.1371/journal.pmed.1002356.g001

available in most emergency departments and intensive care units, and are used for many
other purposes. Therefore, ultrasonography could be very useful for nICP assessment.

PLOS Medicine | https://doi.org/10.1371/journal.pmed.1002356 July 25, 2017 9 / 16

 Ultrasound based non-invasive ICP

Table 5. Summary of the 95% prediction and confidence intervals (± standard deviations) for the linear regression between intracranial pressure
and non-invasive estimators between patients (N = 64).
Estimator 95% prediction interval 95% confidence interval
Lower bound Upper bound Lower bound Upper bound
ONSD 5.05 ± 4.04 19.32 ± 4.17 11.03 ± 3.95 13.34 ± 4.30
FVsv 4.57 ± 3.83 19.79 ± 3.96 10.94 ± 3.71 13.43 ± 4.12
nICPONSD+FVsv 5.65 ± 4.30 18.72 ± 4.45 10.90 ± 4.19 13.47 ± 4.60

FVsv, straight sinus systolic flow velocity; nICPONSD+FVsv, non-invasive intracranial pressure estimator based on the combination of ONSD and FVsv;
ONSD, optic nerve sheath diameter.

https://doi.org/10.1371/journal.pmed.1002356.t005

The optic nerve is surrounded by subarachnoid space [10,11,31]; hence, the intraorbital
part of the subarachnoid space is distensible and can therefore expand if the CSF pressure
increases, with the maximum ONSD fluctuations occurring in the anterior compartment.
Although the diameter of the optic nerve is narrower in the anterior than in the posterior seg-
ment, increased ICP in the perioptic CSF causes a greater enlargement of the retrobulbar seg-
ment of the optic nerve sheath, 3 mm behind the globe, than of the posterior segment [32].
This is probably related to the asymmetrical distribution of the arachnoidal trabeculae and the
lower density of the arachnoidal trabeculae in the retrobulbar space. ONSD has been investi-
gated in different clinical scenarios [10,33–35], showing a good correlation with ICP measured
invasively and low inter- and intra-observer variability [10,11,27,36]. Our results agree with
these findings. Among the studied methods, ONSD was the most accurate in the assessment of
ICP; moreover, it is a safe and quick method, as the orbital window is easily available and has
no complications.
vTCD for the assessment of ICP is a poorly developed technique. It is known that increasing
ICP leads to venous haemodynamic changes, as the part of the cerebral vasculature most sensi-
tive to elevated ICP is the subarachnoid bridging veins. According to the Monro–Kellie doc-
trine, cerebral compliance strongly depends on the compressibility of the low-pressure venous
compartment, and stasis in the pial veins occurs early as a compensatory mechanism in case of
increased ICP [37,38]. Consequently, venous blood may be pooled toward larger venous ves-
sels (straight sinus and Rosenthal vein), causing an increase in venous flow velocity. An alter-
native explanation may be that straight sinus can be compressed by rising ICP, and, with
constant volume flow, flow velocity may increase.
Schoser et al. applied vTCD for the estimation of ICP in 30 control volunteers and 25
patients with elevated ICP and found a linear relationship, with strong correlation between
mean ICP and FVsv [25]. Similarly to Schoser et al., we found that FVsv is strongly correlated
with ICP, whereas other vTCD parameters (venous pulsatility index and FVdv) were not good
estimators of ICP.

Table 6. Results of receiver operating characteristic analysis for ICP  20 mm Hg considering all mea-
surement points (N = 445).
Estimator AUC (95% CI)
ONSD 0.91 (0.88–0.95)
FVsv 0.81 (0.74–0.87)
nICPONSD+FVsv 0.93 (0.90–0.97)

AUC, area under the curve; FVsv, straight sinus systolic flow velocity; nICPONSD+FVsv, non-invasive
intracranial pressure estimator based on the combination of ONSD and FVsv; ONSD, optic nerve sheath
diameter.

https://doi.org/10.1371/journal.pmed.1002356.t006

PLOS Medicine | https://doi.org/10.1371/journal.pmed.1002356 July 25, 2017 10 / 16

 Ultrasound based non-invasive ICP

Fig 2. Receiver operating characteristic analysis for different nICP predictors for a threshold of ICP  20 mm Hg. (A) ONSD method; (B)
FVsv; (C) nICP estimator based on the combination of ONSD and FVsv (nICPONSD+FVsv). The values shown on the curve in (A) and (B) represent
the best thresholds (cutoff values presenting the best sensitivity and specificity [in parentheses]) for prediction of intracranial hypertension (ICP  20
mm Hg), respectively, for ONSD and FVsv. AUC is presented followed by the 95% confidence interval. AUC, area under the curve; FVsv, straight
sinus systolic flow velocity; ICP, intracranial pressure; nICP, non-invasive intracranial pressure; ONSD, optic nerve sheath diameter.
https://doi.org/10.1371/journal.pmed.1002356.g002

Although the measurement of FVsv seems promising, this technique has some limitations:
it can be impossible in polytraumatic patients because of the presence of a cervical collar (6
cases in our cohort). Moreover, the insonation of the straight sinus is feasible in just 72% of

Table 7. Summary table describing the association between ICP, ONSD, and mortality.
Measure Survived (mean ± SD) Dead (mean ± SD) Difference
Mean p-Value
ICP 11.60 ± 4.65 14.45 ± 7.43 −2.85 0.088
ONSD 5.11 ± 0.66 5.71 ± 1.37 −0.61 0.023

ICP, intracranial pressure; ONSD, optic nerve sheath diameter.

https://doi.org/10.1371/journal.pmed.1002356.t007

PLOS Medicine | https://doi.org/10.1371/journal.pmed.1002356 July 25, 2017 11 / 16

 Ultrasound based non-invasive ICP

Fig 3. Boxplots of the analysis of variance of ICP and ONSD between patients who survived and those who died. (A) ICP; (B) ONSD. The mean
ONSD between patients who survived and died was significantly different. ICP, intracranial pressure; ONSD, optic nerve sheath diameter.
https://doi.org/10.1371/journal.pmed.1002356.g003

cases because of anatomical variations in cerebral veins and transcranial insonation difficulties
[25] (even though we had just 3 unsuccessful cases in our cohort, 3.7%).
Our method has several potential clinical applications: it could be useful when invasive
monitoring is contraindicated or unavailable, or in many “borderline” situations in which the
insertion of invasive monitoring is questioned but a nICP measurement could be useful
[20,21]. It can also be applied in patients at risk of intracranial hypertension for causes that are
not primarily neurosurgical (such as liver transplantation and intraoperative settings at risk of
intracranial hypertension [23,24]) or as screening tool in the emergency department in
patients where there is doubt about the need for invasive ICP monitoring.

Limitations
There are several limitations that deserve to be mentioned. First, transcranial Doppler (and
ONSD) measurements were intermittent, and continuous measurements remain more feasible
with invasive techniques. Second, the mixed cohort of enrolled patients, including different
types of acute brain injury, may represent a bias, as the ICP and cerebral perfusion pressure
thresholds for subarachnoid haemorrhage, intracerebral haemorrhage, and stroke are not as
well defined as for traumatic brain injury. However, this heterogeneity increases the applicabil-
ity of the study in many clinical scenarios. Other major limitations are the small number of
patients included in this study, the need for specialised training to perform and interpret the
ultrasound tests, and the variability in performance among different ultrasound operators.
Finally, most our measurements were obtained in patients with relatively well-controlled
ICP. Although a strong correlation between nICP and invasive ICP within the range investi-
gated supports the assumption of validity beyond the range investigated, larger validation stud-
ies will need to be performed before non-invasive techniques will be able to substitute for
invasive ICP monitoring. In addition, despite our findings showing that mortality has a stron-
ger association with ONSD than with ICP, this does not imply that it would be clinically better
to monitor and manage ONSD than ICP.

PLOS Medicine | https://doi.org/10.1371/journal.pmed.1002356 July 25, 2017 12 / 16

 Ultrasound based non-invasive ICP

Conclusion
A novel nICP monitoring method based on combined ONSD ultrasonography and vTCD was
shown to have promising value for the diagnosis of intracranial hypertension, and a strong
correlation with invasive ICP monitoring. This ultrasound-based method is quick, low-cost,
and based on technology widely available in emergency departments and intensive care units.
Whilst we still advocate the superiority of invasive ICP monitoring when this is clearly indi-
cated, the non-invasive methodology presented here may be of potential benefit for ICP assess-
ment in several clinical scenarios where invasive measurement is not immediately available or
is contraindicated. However, this method has several limitations, and further studies are
needed to confirm and validate our findings.

Supporting information
S1 Data. Database of the patients included in the study.
(CSV)
S1 Fig. Individual regression plots between ICP and ONSD and ICP and FVsv. (A) ICP and
ONSD; (B) ICP and FVsv. The numbers on the plots represent the patients (N = 64), with iden-
tities ranked from bottom left (3) to top right (60) on the basis of mean ICP value (in mm Hg).
It is noticeable from the individual plots that the association between ICP and each predictor
is variable across patients. FVsv, straight sinus systolic flow velocity; ICP, intracranial pressure;
ONSD, optic nerve sheath diameter.
(PNG)
S1 Text. STROBE checklist.
(DOC)
S2 Text. Protocol of the study and ethical committee form.
(PDF)

Acknowledgments
PJAH is a Research Professor of the National Institute for Health Research (UK). The authors
would like to acknowledge that Cambridge Healthcare Technology Cooperative supported
CR, DC, JD, and MC for participation in the 16th International Symposium of Intracranial
Pressure in Boston, where data on ONSD/ venous Doppler examination were preliminarily
presented.

Author Contributions
Conceptualization: Chiara Robba, Danilo Cardim, Tamara Tajsic, Justine Pietersen, Joseph
Donnelly, Andrea Lavinio, Arun Gupta, David K. Menon, Peter J. A. Hutchinson, Marek
Czosnyka.
Data curation: Chiara Robba, Danilo Cardim, Tamara Tajsic, Justine Pietersen, Michael Bul-
man, Joseph Donnelly, Andrea Lavinio, Arun Gupta, David K. Menon, Peter J. A. Hutchin-
son, Marek Czosnyka.
Formal analysis: Chiara Robba, Danilo Cardim, Tamara Tajsic, Justine Pietersen, Michael
Bulman, Joseph Donnelly, Andrea Lavinio, Arun Gupta, Peter J. A. Hutchinson, Marek
Czosnyka.

PLOS Medicine | https://doi.org/10.1371/journal.pmed.1002356 July 25, 2017 13 / 16

 Ultrasound based non-invasive ICP

Funding acquisition: Chiara Robba, Danilo Cardim, Tamara Tajsic, Justine Pietersen,
Michael Bulman, Joseph Donnelly, Andrea Lavinio, Arun Gupta, David K. Menon, Peter J.
A. Hutchinson, Marek Czosnyka.
Investigation: Chiara Robba, Danilo Cardim, Tamara Tajsic, Justine Pietersen, Michael Bul-
man, Joseph Donnelly, Andrea Lavinio, Arun Gupta, David K. Menon, Peter J. A. Hutchin-
son, Marek Czosnyka.
Methodology: Chiara Robba, Danilo Cardim, Tamara Tajsic, Justine Pietersen, Michael Bul-
man, Joseph Donnelly, Arun Gupta, David K. Menon, Peter J. A. Hutchinson, Marek
Czosnyka.
Project administration: Chiara Robba, Danilo Cardim, Tamara Tajsic, Justine Pietersen,
Michael Bulman, Joseph Donnelly, Andrea Lavinio, Arun Gupta, David K. Menon, Peter J.
A. Hutchinson, Marek Czosnyka.
Resources: Chiara Robba, Danilo Cardim, Tamara Tajsic, Justine Pietersen, Michael Bulman,
Andrea Lavinio, Arun Gupta, David K. Menon, Peter J. A. Hutchinson, Marek Czosnyka.
Software: Chiara Robba, Danilo Cardim, Tamara Tajsic, Michael Bulman, Joseph Donnelly,
Andrea Lavinio, Arun Gupta, David K. Menon, Peter J. A. Hutchinson, Marek Czosnyka.
Supervision: Chiara Robba, Danilo Cardim, Tamara Tajsic, Justine Pietersen, Michael Bul-
man, Joseph Donnelly, Andrea Lavinio, Arun Gupta, David K. Menon, Peter J. A. Hutchin-
son, Marek Czosnyka.
Validation: Chiara Robba, Danilo Cardim, Tamara Tajsic, Justine Pietersen, Michael Bulman,
Joseph Donnelly, Andrea Lavinio, Arun Gupta, David K. Menon, Peter J. A. Hutchinson,
Marek Czosnyka.
Visualization: Chiara Robba, Danilo Cardim, Tamara Tajsic, Justine Pietersen, Michael Bul-
man, Joseph Donnelly, Andrea Lavinio, Arun Gupta, David K. Menon, Peter J. A. Hutchin-
son, Marek Czosnyka.
Writing – original draft: Chiara Robba, Arun Gupta, Marek Czosnyka.
Writing – review & editing: Chiara Robba, Danilo Cardim, Tamara Tajsic, Justine Pietersen,
Michael Bulman, Joseph Donnelly, Andrea Lavinio, Arun Gupta, David K. Menon, Peter J.
A. Hutchinson, Marek Czosnyka.

References
1. Balestreri M, Czosnyka M, Hutchinson P, Steiner LA, Hiler M, Smielewski P, et al. Impact of intracranial
pressure and cerebral perfusion pressure on severe disability and mortality after head injury. Neurocrit
Care. 2006; 4:8–13. https://doi.org/10.1385/NCC:4:1:008 PMID: 16498188
2. Badri S, Chen J, Barber J, Temkin NR, Dikmen SS, Chesnut RM, et al. Mortality and long-term func-
tional outcome associated with intracranial pressure after traumatic brain injury. Intensive Care Med.
2012; 38:1800–9. https://doi.org/10.1007/s00134-012-2655-4 PMID: 23011528
3. Chesnut RM, Temkin N, Carney N, Dikmen S, Rondina C, Videtta W, et al. A trial of intracranial-pres-
sure monitoring in traumatic brain injury. N Engl J Med. 2012; 367:2471–81. https://doi.org/10.1056/
NEJMoa1207363 PMID: 23234472
4. Carney N, Totten AM, O’Reilly C, Ullman JS, Hawryluk GWJ, Bell MJ, et al. Guidelines for the manage-
ment of severe traumatic brain injury, fourth edition. Neurosurgery. 2017; 80(1):6–15. https://doi.org/10.
1227/NEU.0000000000001432 PMID: 27654000
5. Brain Trauma Foundation, American Association of Neurological Surgeons/Congress of Neurological
Surgeons Joint Section on Neurotrauma and Critical Care. Guidelines for the management of severe
traumatic brain injury, 3rd edition. J Neurosurg. 2007;24(Suppl 1):S1-106.

PLOS Medicine | https://doi.org/10.1371/journal.pmed.1002356 July 25, 2017 14 / 16

 Ultrasound based non-invasive ICP

6. Holloway KL, Barnes T, Choi S, Bullock R, Marshall LF, Eisenberg HM, et al. Ventriculostomy infec-
tions: the effect of monitoring duration and catheter exchange in 584 patients. J Neurosurg. 1996;
85:419–24. https://doi.org/10.3171/jns.1996.85.3.0419 PMID: 8751626
7. Binz DD, Toussaint LG, Friedman JA. Hemorrhagic complications of ventriculostomy placement: a
meta-analysis. Neurocrit Care. 2009; 10:253–6. https://doi.org/10.1007/s12028-009-9193-0 PMID:
19224404
8. Bauer DF, Razdan SN, Bartolucci AA, Markert JM. Meta-analysis of hemorrhagic complications from
ventriculostomy placement by neurosurgeons. Neurosurgery. 2011; 69:255–60. https://doi.org/10.
1227/NEU.0b013e31821a45ba PMID: 21471831
9. Strumwasser A, Kwan RO, Yeung L, Miraflor E, Ereso A, Castro-Moure F, et al. Sonographic optic
nerve sheath diameter as an estimate of intracranial pressure in adult trauma. J Surg Res. 2011;
170:265–71. https://doi.org/10.1016/j.jss.2011.03.009 PMID: 21550065
10. Dubourg J, Javouhey E, Geeraerts T, Messerer M, Kassai B. Ultrasonography of optic nerve sheath
diameter for detection of raised intracranial pressure: a systematic review and meta-analysis. Intensive
Care Med. 2011; 37:1059–68. https://doi.org/10.1007/s00134-011-2224-2 PMID: 21505900
11. Geeraerts T, Merceron S, Benhamou D, Vigué B, Duranteau J. Non-invasive assessment of intracranial
pressure using ocular sonography in neurocritical care patients. Intensive Care Med. 2008; 34:2062–7.
https://doi.org/10.1007/s00134-008-1149-x PMID: 18509619
12. De Riva N, Budohoski KP, Smielewski P, Kasprowicz M, Zweifel C, Steiner LA, et al. Transcranial dopp-
ler pulsatility index: what it is and what it isn’t. Neurocrit Care. 2012; 17:58–66. https://doi.org/10.1007/
s12028-012-9672-6 PMID: 22311229
13. Zweifel C, Czosnyka M, Carrera E, de Riva N, Pickard JD, Smielewski P. Reliability of the blood flow
velocity pulsatility index for assessment of intracranial and cerebral perfusion pressures in head-injured
patients. Neurosurgery. 2012; 71:853–61. https://doi.org/10.1227/NEU.0b013e3182675b42 PMID:
22791038
14. Ragauskas A, Bartusis L, Piper I, Zakelis R, Matijosaitis V, Petrikonis K, et al. Improved diagnostic
value of a TCD-based non-invasive ICP measurement method compared with the sonographic ONSD
method for detecting elevated intracranial pressure. Neurol Res. 2014; 36:607–14. https://doi.org/10.
1179/1743132813Y.0000000308 PMID: 24620972
15. Ragauskas A, Matijosaitis V, Zakelis R, Petrikonis K, Rastenyte D, Piper I, et al. Clinical assessment of
noninvasive intracranial pressure absolute value measurement method. Neurology. 2012; 78
(21):1684–91. https://doi.org/10.1212/WNL.0b013e3182574f50 PMID: 22573638
16. Schmidt B, Klingelhöfer J, Schwarze JJ, Sander D, Wittich I. Noninvasive prediction of intracranial pres-
sure curves using transcranial Doppler ultrasonography and blood pressure curves. Stroke. 1997;
28:2465–72. https://doi.org/10.1161/01.STR.28.12.2465 PMID: 9412634
17. Kashif FM, Verghese GC, Novak V, Czosnyka M, Heldt T. Model-based noninvasive estimation of intra-
cranial pressure from cerebral blood flow velocity and arterial pressure. Sci Transl Med. 2012;
4:129ra44. https://doi.org/10.1126/scitranslmed.3003249 PMID: 22496546
18. Robba C, Bacigaluppi S, Cardim D, Donnelly J, Bertuccio A, Czosnyka M. Non-invasive assessment of
intracranial pressure. Acta Neurol Scand. 2016; 134(1):4–21. https://doi.org/10.1111/ane.12527 PMID:
26515159
19. Budohoski KP, Schmidt B, Smielewski P, Kasprowicz M, Plontke R, Pickard JD, et al. Non-invasively
estimated ICP pulse amplitude strongly correlates with outcome after TBI. Acta Neurochir Suppl. 2012;
114:121–5. https://doi.org/10.1007/978-3-7091-0956-4_22 PMID: 22327676
20. Steiner T, Juvela S, Unterberg A, Jung C, Forsting M, Rinkel G. European Stroke Organization guide-
lines for the management of intracranial aneurysms and subarachnoid haemorrhage. Cerebrovasc Dis.
2013; 35:93–112. https://doi.org/10.1159/000346087 PMID: 23406828
21. Connolly ES, Rabinstein AA, Carhuapoma JR, Derdeyn CP, Dion J, Higashida RT, et al. Guidelines for
the management of aneurysmal subarachnoid hemorrhage: a guideline for healthcare professionals
from the American Heart Association/American Stroke Association. Stroke. 2012; 43(6):1711–37.
https://doi.org/10.1161/STR.0b013e3182587839 PMID: 22556195
22. Maas AIR, Hukkelhoven CWPM, Marshall LF, Steyerberg EW. Prediction of outcome in traumatic brain
injury with computed tomographic characteristics: a comparison between the computed tomographic
classification and combinations of computed tomographic predictors. Neurosurgery. 2005; 57:1173–81.
https://doi.org/10.1227/01.NEU.0000186013.63046.6B PMID: 16331165
23. Robba C, Bragazzi NL, Bertuccio A, Cardim D, Donnelly J, Sekhon M, et al. Effects of Prone position
and positive end-expiratory pressure on noninvasive estimators of ICP: a pilot study. J Neurosurg
Anesthesiol. 2017; 29(3):243–250. https://doi.org/10.1097/ANA.0000000000000295 PMID: 26998650

PLOS Medicine | https://doi.org/10.1371/journal.pmed.1002356 July 25, 2017 15 / 16

 Ultrasound based non-invasive ICP

24. Robba C, Cardim D, Donnelly J, Bertuccio A, Bacigaluppi S, Bragazzi N, et al. Effects of pneumoperito-
neum and Trendelenburg position on intracranial pressure assessed using different non-invasive meth-
ods. Br J Anaesth. 2016; 117:783–91. https://doi.org/10.1093/bja/aew356 PMID: 27956677
25. Schoser BG, Riemenschneider N, Hansen HC. The impact of raised intracranial pressure on cerebral
venous hemodynamics: a prospective venous transcranial Doppler ultrasonography study. J Neuro-
surg. 1999; 91:744–9. https://doi.org/10.3171/jns.1999.91.5.0744 PMID: 10541230
26. Cardim D, Robba C, Donnelly J, Bohdanowicz M, Schmidt B, Damian M, et al. Prospective study on
non-invasive assessment of ICP in head injured patients: comparison of four methods. J Neurotrauma.
2016; 33(8):792–802. https://doi.org/10.1089/neu.2015.4134 PMID: 26414916
27. Rajajee V, Vanaman M, Fletcher JJ, Jacobs TL. Optic nerve ultrasound for the detection of raised intra-
cranial pressure. Neurocrit Care. 2011; 15:506–15. https://doi.org/10.1007/s12028-011-9606-8 PMID:
21769456
28. Bates D, Maechler M, Bolker B, Walker S. Fitting linear mixed-effects models using lme4. J Stat Softw.
2015; 67:1–48. https://doi.org/10.18637/jss.v067.i01
29. Hosmer D, Lameshow S. Applied logistic regression. New York: John Wiley & Sons; 1989.
30. Robin AX, Turck N, Hainard A, Lisacek F, Sanchez J, Müller M. pROC: an open-source package for R
and S+ to analyze and compare ROC curves. BMC Bioinformatics. 2011; 12:77. https://doi.org/10.
1186/1471-2105-12-77 PMID: 21414208
31. Geeraerts T, Newcombe VFJ, Coles JP, Abate MG, Perkes IE, Hutchinson PJ, et al. Use of T2-
weighted magnetic resonance imaging of the optic nerve sheath to detect raised intracranial pressure.
Crit Care. 2008; 12:R114. https://doi.org/10.1186/cc7006 PMID: 18786243
32. Helmke K, Hansen HC. Fundamentals of transorbital sonographic evaluation of optic nerve sheath
expansion under intracranial hypertension. I. Experimental study. Pediatr Radiol. 1996; 26:701–5.
https://doi.org/10.1007/BF01383383 PMID: 8805599
33. Helmke K, Burdelski M, Hansen HC. Detection and monitoring of intracranial pressure dysregulation in
liver failure by ultrasound. Transplantation. 2000; 70:392–5. https://doi.org/10.1097/00007890-
200007270-00029 PMID: 10933171
34. Komut E, Kozaci N, Sönmez BM, Yilmaz F, Komut S, Yildirim ZN, et al. Bedside sonographic measure-
ment of optic nerve sheath diameter as a predictor of intracranial pressure in ED. Am J Emerg Med.
2016; 34:963–7. https://doi.org/10.1016/j.ajem.2016.02.012 PMID: 26944107
35. Moretti R, Pizzi B. Optic nerve ultrasound for detection of intracranial hypertension in intracranial hemor-
rhage patients confirmation of previous findings in a different patient population. 2009; 21:16–20.
https://doi.org/10.1097/ANA.0b013e318185996a PMID: 19098619
36. Amini A, Kariman H, Arhami Dolatabadi A, Hatamabadi HR, Derakhshanfar H, Mansouri B, et al. Use of
the sonographic diameter of optic nerve sheath to estimate intracranial pressure. Am J Emerg Med.
2013; 31:236–9. https://doi.org/10.1016/j.ajem.2012.06.025 PMID: 22944553
37. Valdueza JM, Schultz M, Harms L, Einhaupl KM. Venous transcranial Doppler ultrasound monitoring in
acute dural sinus thrombosis. Report of two cases. Stroke. 1995; 26:1196–9. PMID: 7604413
38. Wardlaw JM, Vaughan GT, Steers AJ, Sellar RJ. Transcranial Doppler ultrasound findings in cerebral
venous sinus thrombosis. Case report. J Neurosurg. 1994; 80:332–5. https://doi.org/10.3171/jns.1994.
80.2.0332 PMID: 8283275

PLOS Medicine | https://doi.org/10.1371/journal.pmed.1002356 July 25, 2017 16 / 16