Supporting Novice Prehospital Transcranial Ultrasound Scanning

for Brain Haemorrhage

Leila Eadie

, Luke Regan

, Ashish MacAden

and Philip Wilson

Centre for Rural Health, University of Aberdeen, Centre for Health Science, Old Perth Road, Inverness, U.K.

NHS Highland, Raigmore Hospital, Old Perth Road, Inverness, U.K.

Keywords: Ultrasound, Prehospital, Remote, Support, Transcranial, 3-Dimensional, Haemorrhage.

Abstract: Traumatic brain injury is a significant problem due to difficulties in early diagnosis in the field. Computed

tomography is the gold standard for detecting brain haemorrhage, but scanners are bulky and expensive. A

cheap, portable scanner such as transcranial ultrasound (TCUS) could allow early triage and intervention.

Transmitting images to remote experts for diagnosis means TCUS could be used by any minimally trained

person in the field. We propose a virtual 3-dimensional model of the head which shows which areas of the

brain have been imaged already, where the probe currently is, and where still needs to be covered in order to

generate a complete scan. Using sensors to measure the position and rotation of the TCUS transducer, we

can link this to the 3D model of the head and visually display which areas have been imaged. The images

can be analysed and composited to form a personalised 3D scan with maximal coverage of the brain, which

can be transmitted for diagnostic review, reducing data loss compared with streaming ongoing images.

Initial testing of the software has been performed in healthy volunteers and further testing is planned in

patients with brain haemorrhage.

1 INTRODUCTION

Traumatic brain injury (TBI) is a significant

problem, with challenges in diagnosis, especially

early diagnosis in the field. Closed head injuries are

of particular concern and far outnumber the

penetrative head injuries on which official statistics

are based. Currently, there are no well-accepted

diagnostic tests for use in standard medical practice

to diagnose TBI (Centers for Disease Control and

Prevention [CDC] et al., 2013). Ideally, brain

imaging should be performed as soon as possible

because posttraumatic bleeding within the skull is

associated with worse prognosis and can be life

threatening (CDC et al., 2013). Prompt diagnosis

and improved prehospital care can mean that

secondary (non-immediate) brain injury can be

prevented or limited by good early care (maintaining

blood and oxygen flow to the brain, controlling

blood pressure and haemorrhage and potentially

drilling of surgical burr holes), training and

organisation of trauma services, potentially leading

to significant reductions in both mortality and long-

term disability (Gentleman, 2008; CDC et al., 2013).

Computed tomography (CT) scanning is the gold

standard for detecting haemorrhage in the brain, but

is not feasible in the field: the scanners are heavy,

bulky, expensive and currently not developed in any

ruggedized form. A cheap, portable scanner for

bleeding in the brain (plus other conditions such as

skull fractures, indicators of intracranial pressure,

etc.) could alert medics to problems and allow early

triage and intervention. We believe transcranial

ultrasound (TCUS) has potential in this area and, if

used with a communications system to transmit the

images to a remote expert for diagnosis, could be

used to assess the injured by any minimally trained

person in the field. These scanners can fit into cases

only a little larger than a laptop, and can therefore be

used in ambulances and other prehospital situations.

Ultrasound is a useful tool for many diagnostic

purposes in trauma and beyond, and it is hoped that

this work will add to its utility rather than requiring

a different tool for head injuries.

Currently ultrasound is not routinely used

clinically for identifying brain structures and

abnormalities, although transcranial Doppler,

measuring blood flow velocity within cerebral

arteries, is a more commonly performed scan

(Sarkar et al., 2007). However, following previous

work by our group – the Satellite Ultrasound for

118

Eadie, L., Regan, L., MacAden, A. and Wilson, P.

Supporting Novice Prehospital Transcranial Ultrasound Scanning for Brain Haemorrhage.

DOI: 10.5220/0005789901180123

In Proceedings of the 9th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2016) - Volume 2: BIOIMAGING, pages 118-123

ISBN: 978-989-758-170-0

Rural Stroke project (Mort et al., 2015) – we believe

that TCUS could provide useful prehospital

information when imaging brains looking for

haemorrhage. In this project, prehospital ultrasound

images were recorded by novice users and streamed

live using mobile or satellite networks from an

ambulance in remote and rural areas of the Scottish

Highlands to hospital-based experts for diagnosis.

The quality of the real-time streamed images was

rated and found to be good enough for diagnostic

purposes in ~93% of cases, although connectivity

was variable. This preliminary work identified that

novice scanners could, after brief training, record

images of the brain’s midline, located using the third

ventricle. The next logical step was to focus on

facilitating the TCUS scanning by novices to ensure

that the appropriate information is captured and

easily and efficiently transmitted.

If TCUS is to be used to provide information

about bleeding in the brain (or its absence), then it is

important to ensure that as much of the brain has

been scanned as possible so the accuracy of the

diagnosis can be quantified. Novice scanners in

particular may find this difficult, without the expert

knowledge to interpret the images and direct their

scanning. It would be useful to provide them with a

guide showing what they have scanned and what has

not yet been imaged.

The current research is developing a software

package to assist the operator in achieving

comprehensive ultrasound scanning of the head. We

aim to make TCUS scanning simpler for non-

medically trained users by providing them with a

visual 3D model of the head which clearly shows

which areas of the brain have been imaged already,

where the probe currently is, and where still needs to

be covered in order to generate a maximal scan –

something that is important for diagnosing and

treating brain injury. By measuring the position and

rotation of the ultrasound transducer using

movement sensors, we can determine where the

probe is pointing at all times during the scanning

session. Linking this to a 3D model of the head

allows us to determine which areas have been

imaged and which have not. The images recorded

will then be composited to form a complete 3D scan

of the subject’s head, using image processing

techniques to locate the skull in order to fit the

model to the individual’s specific head size and

shape. This means a personalised 3D TCUS scan

with maximum coverage of the brain is created,

offering the ability to view the brain images from

any point on the head, as if the user possessed a

virtual probe that can be placed anywhere on the 3D

model. This 3D scan is also a single file that can be

transmitted for diagnostic review with less chance of

data loss than streaming images from an ongoing

scan.

This position paper describes the initial tests in

healthy volunteers, and discusses plans for further

development and testing.

2 METHODS

The software was written using MATLAB R2015a

(MathWorks, Massachusetts USA) and produces a

real-time display of where the TCUS transducer is

currently pointing and has already scanned; plus

offline image analysis to locate the skull in the

recorded TCUS images and use this to deform a

standard 3D skull mesh. The scan model is displayed

and interacted with on a laptop, but in future work

could potentially be integrated into manufacturers’

ultrasound software.

A Philips CX50 portable ultrasound machine

(Philips Healthcare, Amsterdam, Netherlands) was

used in the testing, with an inertial measurement unit

sensor (3-Space Sensor, YEI Technology, Ohio

USA) attached to the transducer throughout

scanning. One temporary experimental set-up is

shown in Figure 1; ideally the sensor chip would be

integrated into the probe for easier handling. This

sensor records quaternion measurements which are

then used by the new software to calculate the plane

of the scan, which is then displayed within the skull

mesh. There are currently no dedicated TCUS

transducers available, so volunteers used a cardiac

probe with an appropriately small footprint, which

can be used at lower frequencies (1-5 MHz) to

penetrate through the skull and image to the opposite

side of the cranial vault; software settings of the

Figure 1: The inertial measurement unit sensor and the

ultrasound probe.

Supporting Novice Prehospital Transcranial Ultrasound Scanning for Brain Haemorrhage

119

Figure 2: The head measurements made in the healthy volunteer study.

ultrasound machine were also optimised as much as

possible for head scanning.

The TCUS images were recorded as video clips

to capture both the images and any changes in depth,

power and sensitivity required to image the various

depths of brain for each volunteer. The images

recorded will be composited to form a complete 3D

scan of the subject’s head, using image processing

techniques (segmentation) to locate the skull in each

image in order to deform and fit the virtual 3D mesh

model to the individual’s specific head size. This

creates a personalised 3D TCUS scan.

Initial testing of the software was performed in

healthy volunteers to gather data about the scanning

process and visibility of brain structures on

ultrasound, plus transducer position data to calculate

the brain coverage of scans. Previous studies have

shown that between 5–20% of subjects have acoustic

windows offering reduced TCUS penetration, which

makes imaging more difficult or impossible (Seidel

et al., 1993; Sarkar et al., 2007).

Tests were done under laboratory conditions.

The dimensions of each volunteer’s head were

measured using standard neurophysiological

landmarks (nasion, inion, preauricular points; see

Figure 2). Structural imaging involved scanning

through the left transtemporal acoustic window (an

area of often thinner bone used to allow greater

penetration of the ultrasound waves) to the opposite

side of the head with the transducer in the vertical

plane, then determining the lateral and vertical range

of imaging achievable through this window.

Visualisation of landmarks such as the skull base,

brainstem, sphenoid bone and choroid plexus was

attempted. These steps were repeated in the

horizontal transducer plane, and from the opposite

transtemporal acoustic window. Imaging then

focussed on the midline of the brain, represented by

the third ventricle. The distance from skull to third

ventricle was measured from both transtemporal

windows. Volunteers then scanned a member of the

study team with the help of the new head scanning

software, after which they were given a short

questionnaire about their experience using the

program. Feedback was used to improve the

program’s usability and features.

3 RESULTS

The initial testing recruited 12 healthy volunteers: 9

female, 3 male, none of whom had used ultrasound

before. Table 1 shows the average head

measurements recorded in the study, grouped by

gender. This shows that there are differences in head

size that support the use of a personalised head

model rather than simply using a standard ‘average’

mesh model.

Table 1: The mean head measurements ± standard

deviation (in cm) recorded in the healthy volunteer study.

Subject type Nasion to

inion

Pre-auricular

point to point

Circum-

ference

Average

37.6 ± 1.7 35.7 ± 2.5 57.8 ± 1.6

Average male

38.3 ± 0.5 37.3 ± 2.6 58.7 ± 0.9

Average

female

37.3 ± 1.9 35.1 ± 2.1 57.6 ± 1.7

Two of the volunteers had transtemporal acoustic

windows that allowed more limited views of the

brain than others. This proportion is similar to what

has been reported in the literature. Imaging was still

possible, but of reduced clarity.

All twelve volunteers used the new software to

help them scan a team member’s head. Figure 3

shows a screenshot of the software from such a scan,

illustrating the cumulative coverage of the head

scans. The average scan time was 8 minutes, 47

seconds (range: 6 minutes, 1 second to 13 minutes,

52 seconds). The computational time of the program

BIOIMAGING 2016 - 3rd International Conference on Bioimaging

120

Figure 3: A screenshot of the 3D head model display program showing where scans have been performed from the different

acoustic windows (red for the patient’s right temporal window, green for their left temporal window and blue for the

occipital window).

was not an issue: it averaged approximately 0.3

seconds, with a maximum of 1 second seen during

the testing.

All volunteers said they felt they could use the

3D head model program, with 92% saying it was

easy to use. Some commented that they found using

the probe difficult initially, but this was because they

had no prior experience with ultrasound: more than

half (58%) said they thought they needed more

instructions or help, such as knowing exactly where

and how to place the probe (e.g., finding the

optimum window, which is actually done by

exploring the approximate area and locating the spot

that produces the best images; it is difficult to give

more than a general guide to location); and being

reminded to make sure the probe didn’t slide away

from the window due to the ultrasound gel used.

All of the volunteers reported that they thought

the 3D model accurately represented the direction of

the probe and that the model updated quickly

enough as they moved. All also thought the model

correctly showed areas they had scanned. Only 36%

said they were able to scan the areas the model

showed were unscanned. This is most likely because

there will be areas, such as the very top of the head,

where the probe cannot scan because it loses contact

with the head when tilted to the necessary angle. The

volunteers did not scan from the occipital window

(due to the poor images seen from this site when it

was attempted), which might have allowed them to

scan some areas otherwise unreachable.

Finally, all of the volunteers reported they

thought the program included all the functions they

might expect to find, given its aims, and all also

were satisfied that the program achieved these aims.

The volunteers were asked to rate the program as

a whole on a scale of 1 (useless) to 5 (useful), and

the mean rating was 4.3, the median was 4 (25

–

percentiles: 4–4.5).

Suggestions from the volunteers as to how the

program could be improved included:

• Providing a more in-depth tutorial on ultrasound,

the machine and the probe, which is the sort of

training that would be provided to end users.

Supporting Novice Prehospital Transcranial Ultrasound Scanning for Brain Haemorrhage

121

• Having a probe designed specifically for head

imaging, with a better shape for reaching all areas;

and incorporating the 3D-model into the ultrasound

screen so it is easier to see both at once. Addressing

both of these points could involve working with

ultrasound manufacturers for further development.

• Indicating the places on the model where they are

unlikely to be able to image.

• Showing a different coloured plane to indicate

where the probe is currently imaging, which has

since been implemented.

• Highlighting any abnormalities imaged, which

would involve computer-aided diagnosis analysis,

discussed in the Future Work section of this paper.

• Automatically detecting when the probe has been

removed from the patient’s head (currently this is

done manually by pressing a button). This is a point

for future development, but a partial solution has

been implemented, detecting when the probe is held

upright, a position in which it would not usually be

during usual TCUS scanning.

4 FUTURE WORK

The analysis of the ultrasound data is ongoing,

looking at forming 3D scans from the 2D images

collected, and calculating the variation between the

scans of the research team member that were taken

by different users.

The patient scanning study is currently starting

and involves hospital-based testing of the software

package in up to 10 patients with bleeding in the

brain. Images of the bleeding are collected for use in

two ways: to test the scanning support software in a

controlled clinical environment, and to compile a set

of TCUS scans featuring haemorrhage, to help

investigate the appearance of bleeding in the brain

on ultrasound. This appearance will change as the

time from the bleeding event increases and the blood

clots, so it is important to image several different

patients with brain haemorrhage (ideally as a result

of different causes), to capture a range of

haemorrhage images. These will also be used to

explore the potential for automated computer-

assisted diagnosis to support TCUS assessment of

patients with brain injury.

Details of the diagnosis and timing of the brain

haemorrhage will be recorded, as will CT images

that form part of the patient’s usual care and

diagnosis, for comparison. Scanning will take place

at the patient’s bedside and involve the same

structural scanning as in the healthy volunteer study,

plus the area of the haemorrhage will be

comprehensively imaged. The expert ultrasound

operators will provide feedback via a short

questionnaire; they will also be asked to assess the

resulting 3D image models for utility and quality at a

later time point.

5 CONCLUSIONS

The software described in this paper is specifically

designed to support non-medically trained TCUS

users in taking diagnostically useful images, so that

no expertise in interpreting ultrasound is required.

Brief training on the use of the ultrasound machine

and 3D imaging program is all that is needed. It has

been initially tested in healthy volunteers, with

further testing planned in patients with brain

bleeding.

This study recruited volunteers for whom it was

their first time using ultrasound, and they took a

little time to become comfortable with holding the

probe and finding the optimum acoustic windows

through which to image. The average scan time of

almost 9 minutes would no doubt improve with

practice, experience and confidence. The

computational time of the program did not produce

any appreciable lag and the feedback provides

essentially updates in ‘real time’. The scan model

used in this work was displayed and interacted with

on a laptop, but in future work could potentially be

integrated into manufacturers’ ultrasound software,

so that users do not have to view two different

screens.

The volunteers reported that the support system

fulfilled its aims and appeared to be working

correctly. They also made suggestions for

improvements, some of which have already been

implemented and will be ready for testing in further

trials. This study was the first feasibility test and

highlighted some problems with the imaging system,

such as the existence of areas of the brain that are

extremely difficult to scan with the currently

available hardware. Using additional acoustic

windows could help provide solutions to this

problem, as could development of a TCUS-specific

probe.

TBI is a significant problem, especially for the

military, and can lead to neuropsychiatric and

neurological sequelae. The benefits of a field-based

lay-user TCUS system are applicable to both

military and civilian situations (e.g., prehospital,

ambulance-based diagnosis for head injury and

stroke). It builds on existing (currently experimental)

ultrasound transmission technology to improve its

BIOIMAGING 2016 - 3rd International Conference on Bioimaging

122

usability and efficiency for use by non-expert

medics. The proposed system will make it easy for

any minimally trained personnel to collect

diagnostically relevant head images, which can then

be transmitted in a single package to a remote site

for interpretation. In this way, early diagnosis of

brain injury in the field – specifically looking for

haemorrhage in closed traumatic head injury - can

be improved without requiring major training for

novice users, because, thanks to modern

communication technologies, the images can be

transmitted and diagnosis performed by experts at a

remote site. The proposed system will also decrease

the effects of unstable transmission and packet loss

in sending the images to experts, by compositing the

data into a single file to be transmitted rather than

streaming the ultrasound video, where frames will

frequently be lost.

The impact of earlier diagnosis of TBI using

such a system as described here could be potentially

huge (preventing/minimising sequelae, long term

health effects of early/any treatment). It does rely on

TCUS being able to reliably detect haemorrhage in

the brain, but previous studies have shown there is a

strong possibility that the sensitivity will be of a

useful level (e.g., Mäurer et al., 1998). These studies

were performed some time ago so, with the benefit

of today’s improved ultrasound technology and

ongoing transducer optimisation, we are optimistic

that TCUS will prove worthwhile and useful for this

situation. We are planning concurrent validity

studies with TCUS and CT in patients with stroke in

order to provide updated evidence that modern US

systems can be used to reliably detect haemorrhage.

Although TCUS may be less sensitive than CT for

detecting haemorrhage, its portability and low cost

make it an attractive technology for battlefield and

transit use. Ruggedised systems are already available

for use in the field, and are used by air ambulance

services around the world.

We believe TCUS has potential and, if used with

a communications system to transmit the images to a

remote expert for diagnosis, could be used to assess

the injured by any minimally trained person in the

field.

ACKNOWLEDGEMENTS

This project is funded by the UK Ministry of

Defence, Defence Science and Technology

Laboratory. The ultrasound machine used was

loaned to the Centre for Rural Health by Philips

Healthcare.

REFERENCES

Centers for Disease Control and Prevention, the National

Institutes of Health, the Department of Defense, and

the Department of Veterans Affairs Leadership Panel.

2013. Report to Congress on traumatic brain injury in

the United States: Understanding the public health

problem among current and former military

personnel.

Cronk, T.M. 2012. Military Leads in Treating Traumatic

Brain Injury, Expert Says. Department of Defense

News, August 31 2012. Available from:

http://www.defense.gov/news/newsarticle.aspx?id=11

7724 [Accessed 05/11/2015].

Defense Centers of Excellence for Psychological Health

and Traumatic Brain Injury. 2010. Portable, field-

based devices for the early diagnosis of mild traumatic

brain injury. 2010

Gentleman, D. 2008. Synopsis of causation. Head injury.

Ministry of Defence. Available from:

http://www.veterans-uk.info/publications/

head_injury.pdf [Accessed 05/11/2015].

Mäurer, M., Shambal, S., Berg, D., Woydt, M., Hofmann,

E., Georgiadis, D., Lindner, A., Becker, G. 1998.

Differentiation between intracerebral hemorrhage and

ischemic stroke by transcranial color-coded duplex-

sonography. Stroke. Vol. 29, p.2563-2567.

Mort, A., Eadie, L., Regan, L., Macaden, A., Heaney, D.,

Mouley-Bouamrane, M., Rushworth, G., Wilson, P.

Combining transcranial ultrasound with intelligent

communications methods to enhance the remote

assessment and management of stroke patients –

Framework for a technology demonstrator. Health

Informatics Journal, in press.

Sarkar, S., Ghosh, S., Ghosh, S.K., Collier, A. 2007. Role

of transcranial Doppler ultrasonography in stroke.

Postgraduate Medical Journal. Vol. 83, p.683-689.

Seidel, G., Kaps, M., Dorndorf, W. 1993. Transcranial

color-coded duplex sonography of intracerebral

hematomas in adults. Stroke. Vol. 24, p.1519-1527.

Supporting Novice Prehospital Transcranial Ultrasound Scanning for Brain Haemorrhage

123