PROTOCOL

https://doi.org/10.1038/s41596-021-00548-8

Whole-brain functional ultrasound imaging in

awake head-fixed mice
Clément Brunner 1,2,3,4,9, Micheline Grillet 1,2,3,4,9
, Alan Urban1,2,3,4, Botond Roska 5,6,7
,
Gabriel Montaldo1,2,3,4,10 ✉ and Emilie Macé 8,10 ✉

Most brain functions engage a network of distributed regions. Full investigation of these functions thus requires
assessment of whole brains; however, whole-brain functional imaging of behaving animals remains challenging. This
protocol describes how to follow brain-wide activity in awake head-fixed mice using functional ultrasound imaging, a
method that tracks cerebral blood volume dynamics. We describe how to set up a functional ultrasound imaging system
with a provided acquisition software (miniScan), establish a chronic cranial window (timing surgery: ~3–4 h) and image
brain-wide activity associated with a stimulus at high resolution (100 × 110 × 300 µm and 10 Hz per brain slice, which
takes ~45 min per imaging session). We include codes that enable data to be registered to a reference atlas, production of
3D activity maps, extraction of the activity traces of ~250 brain regions and, finally, combination of data from multiple
sessions (timing analysis averages ~2 h). This protocol enables neuroscientists to observe global brain processes in mice.
1234567890():,;
1234567890():,;

Introduction
Systems neuroscience aims to understand the circuits and computations underlying brain functions
and behaviors. However, one major challenge for systems neuroscience is that circuits span multiple
spatial and temporal scales, from the level of the synapse to the entire brain. In past decades, mice
emerged as a popular model for systems neuroscience because of the variety of genetic and probing
tools available to interrogate local neuronal circuits. Neuroscientists can now target, manipulate and
record specific neuronal populations in the context of behavior in mice1–3. Despite these advances,
there is a limited number of options for capturing whole-brain activity, particularly under awake
conditions. Large-scale imaging techniques exist but have specific limitations, such as depth of field,
brain coverage, resolution or compatibility with behavior (see ‘Comparison with other methods’).
Difficulty imaging at the whole-brain scale introduces a bias toward investigating a handful of
regions at a time, resulting in important whole-brain aspects of the function studied potentially
being omitted.
Functional ultrasound imaging (fUS)4 is an emerging technology that has the potential to fill this
gap in the context of mouse research. fUS reports relative changes in cerebral blood volume (CBV) as
a proxy for neuronal activity5 and offers a set of unique features and advantages. First, ultrasound
penetrates deep into tissue: a ~15 MHz ultrasound frequency produces images with ~1 cm pene-
tration, ideal for imaging the entire depth of the mouse brain. Second, fUS offers a high spatio-
temporal resolution (100 × 110 × 300 µm, 10 Hz), which enables the dynamics of brain activity to be
captured through the CBV signal. Finally, the portability of the ultrasound scanner and the compact
size of the probe used for fUS make it easy to combine with other methods and to integrate into
routine neuroscience experiments.
Nonetheless, adapting fUS to chronic imaging of whole-brain activity in awake mice presents
specific challenges. First, because ultrasound is strongly attenuated by the skull, a cranial window
must be implanted. Second, because ultrasound high-resolution probes are limited to a single brain
slice, the imaging procedure must integrate technical constraints (how to scan the brain efficiently)
with constraints linked to the task in which the mouse is engaged. Finally, it is necessary to have an

1
Neuro-Electronics Research Flanders, Leuven, Belgium. 2VIB, Leuven, Belgium. 3Imec, Leuven, Belgium. 4Department of Neurosciences, KU Leuven,
Leuven, Belgium. 5Institute of Molecular and Clinical Ophthalmology Basel, Basel, Switzerland. 6University of Basel, Basel, Switzerland. 7NCCR
Molecular Systems Engineering, Basel, Switzerland. 8Brain-Wide Circuits for Behavior Lab, Max Planck Institute of Neurobiology, Martinsried,
Germany. 9These authors contributed equally: Clément Brunner, Micheline Grillet. 10These authors jointly supervised this work: Gabriel Montaldo,
Emilie Macé. ✉e-mail: [email protected]; [email protected]

NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot 3547

 PROTOCOL NATURE PROTOCOLS

understanding of how to analyze the data, how to combine data from different imaging sessions and
how to localize brain structures in this new type of image.
We recently published a study in which we successfully overcame these challenges6. We used fUS
to reveal all the brain regions activated in the mouse brain during a visuomotor reflex, and how these
regions are functionally affected during various perturbations of the reflex. This protocol facilitates
application of fUS and provides neuroscientists with a standard approach for capturing in an
unbiased way the whole-brain activity elicited by a behavior or brain function in head-fixed mice.
It outlines how to set up an fUS experiment and provides a standard pipeline for data acquisition and
statistical analysis. We believe that this protocol is essential to facilitate comparisons across labora-
tories, and for this method to be widely adopted in systems neuroscience.

Development of the protocol

fUS is a versatile method that has been implemented in different animal models. To date, fUS has
been successfully used in rats7–10, birds11, rabbits12, ferrets13 and nonhuman primates14,15. fUS has
also been used for intraoperative functional neurosurgery in adult humans16,17 and for transfonta-
nelar neuromonitoring in neonates18. On the technological front, the development of miniature
probes has broadened the applications of fUS to image rodents under freely moving conditions8,19.
More recently, the development of matrix probes allowed for volumetric imaging (vfUSI), although at
a reduced spatial resolution20,21. In addition to being applicable to a wide range of species and the
various technical possibilities, fUS has also been successfully applied in different fields of neu-
roscience, from fundamental neuroscience, to preclinical research9,10,22 and in the clinic16,17.
The surgical protocol, imaging parameters, data processing and behavioral context have evolved
since we published our first study using rodents 10 years ago4. This has resulted in various experi-
mental procedures being used, with differing procedures being optimal depending on the application.
Based on our extensive experience of both the challenges and the most up-to-date capabilities of the
method, we provide here a comprehensive protocol to set up the equipment and software, perform
the surgery, image the brain and process the data. Until our recent publication that outlined a general
unbiased whole-brain approach6, experiments were targeted at a specific part of the brain to answer a
specific question. We now present a detailed procedure that can be used to obtain a whole-brain view
of the activity evoked by a stimulus or event. We provide standardized analysis outputs to ensure that
different animals or sessions can be compared. Finally, we explain how to conduct these experiments
in awake head-fixed mice, a context that hopefully suits the needs of many neuroscientists.

Applications
The main use of this protocol we envisage is to identify the brain-wide networks activated by an
‘event’ and to extract the time course of activity in these regions. Here we give the specific example of
a visual stimulus, but this can conceptually be replaced by any other sensory stimuli, a cued task or
any other triggered behavioral event without changing the procedure.
In its current form, particularly the analysis, the protocol is primarily designed for events that can
be repeated multiple times. Localizing the network of regions active during an event, and determining
how this network is dynamically engaged in the event, can provide valuable information to under-
stand the large-scale interactions involved in the underlying neural processes. The ability to image
whole-brain activity in awake mice, and repeatedly for months, allows for a large repertoire of tasks or
events to be studied with fUS. For example, it is possible to follow the engagement of the whole brain
across the learning of a task, test the effect of a drug over time or study the effect of progression of a
disease on brain activity.
One key asset of this protocol is its compatibility with other technologies. Combining fUS with
electrophysiology4,6,7 is important for confirming activation at the neuronal level in deep brain
regions, or for identifying the role of specific neuronal populations in the activity detected. The
combination of fUS with optogenetic control of genetically defined neuronal populations can help
identify cell-type-specific connectivity across the entire brain, and also be used to examine the
behavioral consequences of manipulating targeted populations21,23. While it is beyond the scope of
this protocol to provide examples for all combinations with other methods, this protocol should serve
as a starting point for conducting such experiments. In the case of electrophysiology and optoge-
netics, the main modification would consist of implanting an electrode or an optical fiber during the
cranial window surgery in a position that allows space for positioning the ultrasound probe during
the imaging session. For optogenetics, viral vectors can also easily be injected into the brain during

3548 NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot

 NATURE PROTOCOLS PROTOCOL
the same surgical preparation, the sensory stimulus is replaced by light stimulation and the data
analysis is conceptually the same. Using a chronic window for fUS is also compatible with wide-
field24–27 or two-photon calcium imaging28,29, therefore extending the range of its application.

Comparison with other methods

Large-scale methods are commonly used for mouse brain imaging, but suffer from specific limita-
tions. For example, functional magnetic resonance imaging (fMRI) can capture whole-brain activity
via hemodynamic responses (BOLD-fMRI, CBV-fMRI) in a noninvasive way, but remains mostly
limited to anesthetized animals. While challenging, awake fMRI strategies are emerging30,31; none-
theless, intrinsic MRI scanner specifications strongly limit the repertoire of behaviors that are
compatible with fMRI. Moreover, an MRI scanner is a considerable investment that is not available to
all laboratories. Wide-field calcium imaging is widely used by neuroscientists to efficiently capture
neuronal activity of genetically defined populations of neurons on a large scale, and with good
spatiotemporal resolution during behavior, but remains mostly limited to the dorsal part of the
cortex32. The recent development of the neuropixels probe33 for electrophysiology now allows
thousands of neurons from multiple brain regions of awake behaving rodents to be recorded34.
However, the brain coverage is intrinsically limited given that the location of the probe determines
which regions can be recorded and because only a few of these probes can be implanted simulta-
neously. Finally, multifiber photometry simultaneously targets cortical and subcortical regions of the
mouse brain35, but it remains constrained to tens of recording sites.
Beyond the established methods discussed above, recent developments have also shown the
potential of combining different modalities. For example, the combination of wide-field calcium
imaging and fMRI36 gives access to neuronal activity in the cortex while benefiting from the whole-
brain view offered by fMRI. Similarly, whole-brain optoacoustic imaging of neuronal activity is
promising as it allows detection of the signal from genetically encoded calcium indicators even deep
in the brain37. However, both these approaches are still limited to anesthetized mice to date.
Therefore, compared with other methods, fUS is a practical way to assess the brain-wide activity
associated with a stimulus or an ‘event’ in awake head-fixed mice.
In terms of affordability, the ultrafast ultrasound scanner used in this protocol currently costs
~€100,000 (excluding the high-performance computing workstation required to enable extended
acquisition and processing capabilities) and the probe costs ~€7,000. Therefore, the acquisition of a
functional ultrasound system is in the same price range as a commercial two-photon microscope, but
one order of magnitude cheaper than a small-animal MRI scanner. While this is a considerable
investment that is not available to all laboratories, the fUS system does have the advantage of not
requiring additional maintenance costs.

Limitations
One practical limitation of this protocol is that only one cross-section of the brain can be imaged by
the linear ultrasound probe; therefore, it must be motorized to scan the brain volume. The scanning
approach takes time and requires repetitions of the stimulus or ‘event’ studied. Repetition can lead to
adaptation, and should be partially mitigated by choosing an appropriate intertrial interval or time
between imaging sessions. As mentioned before, the development of matrix ultrasound probes will
likely alleviate this limitation in the future21. However, volumetric imaging (vfUSI) is technologically
challenging, and the resolution is currently limited to ~300 µm3. Note that the surgical and data
analysis steps presented in this protocol are nonetheless directly applicable to vfUSI.
A second practical limitation of this protocol is motion artifacts. Filtering strategies—namely high-
pass and singular value decomposition (SVD) filters38—can manage tissue motion induced by
breathing or small movements, but fail for strong and sudden motion. Head-fixed mice can move
during the imaging sessions, and abrupt movements induce noise in the fUS data. To date, the signal
from the noisy frames cannot be recovered, and the corresponding frames are lost and must be
discarded from the analysis. Nevertheless, as presented in this protocol, the animal’s careful habi-
tuation to the setup and head fixation is sufficient to drastically reduce the proportion of rejected
frames to below 10%, which corresponds to our tolerance rating.
Some limitations are more general to imaging techniques requiring a chronic cranial window and
head fixation, such as wide-field calcium imaging25,39. Although invasive, a cranial window is crucial
to achieve high-quality imaging with fUS. An alternative strategy is thinning the skull7,10, but bone
regrowth usually degrades imaging quality for deep brain structures over time, making this approach

NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot 3549

 PROTOCOL NATURE PROTOCOLS

Animal preparation Imaging session Offline data analysis

30 min
5d Mouse head fixation Anatomical scan
Handling Steps 5 min
Box 1A and scanner setup
26–33 Steps
Registration with 39–43
Allen mouse CCF

3h 15 min/
Skull-bone
Steps Functional ultrasound session
replacement Single-session
1–24 imaging Steps
analysis 1h
34–37 Steps
Stability and rejection 44–45

Session average 46
5d
Surgical recovery Activity map 47–48
Step 25
Post-imaging >1 d
Region–time traces 49–50
care and recovery Step 38

Multiple-sessions
Head-fixation 3d analysis 1h
habituation Box 1B
Steps
Activity map 51–52
Region–time trace 53–55

Fig. 1 | Flowchart of the procedure. Description of the three main stages encountered during chronic fUS of brain-wide activity in awake head-fixed
mice. Animal preparation (green boxes), imaging session (blue boxes) and offline data analysis (orange boxes) protocol described in the Procedure.
Timing and numbers refer to corresponding steps of the Procedure. CCF, Common Coordinate Framework, v3.

challenging for longitudinal imaging. Head fixation also adds constraints to the repertoire of beha-
viors that can be explored, and it can stress the animal. Freely moving imaging with fUS is possible
but with a reduced field of view. This approach is not yet sufficiently developed to replace head
fixation when looking at whole-brain activity8,19.
Finally, this protocol using fUS monitors the vascular signal as a proxy of neuronal activity. Unlike
two-photon or wide-field calcium imaging, fUS cannot follow the activity of individual cells or of
genetically defined cell types. fUS depends on neurovascular coupling, which may vary across brain
regions or across experimental conditions. The possible effects of neurovascular coupling on imaging
results have been reviewed previously40. However, the compatibility of this protocol with other tools
such as electrophysiology offers the opportunity to verify the whole-brain results locally at the
neuronal level.

Overview of the procedure

The procedure describes how to undertake chronic brain-wide fUS in head-fixed mice. The entire
workflow is shown in Fig. 1. The procedure was optimized on adult mice (>8 weeks) and can be
performed similarly on both male and female. Sex is an important biological variable that should be
taken into consideration; therefore we recommend to use mice from both sexes whenever possible.
The first part of the procedure describes how to prepare the animal. fUS requires a cranial window
to be in place as the skull strongly attenuates ultrasound; thus establishment of a chronic cranial
window that gives access to a large fraction of the brain volume is a critical step of the protocol. We
also provide photos and diagrams to facilitate preparation of such a window (Fig. 2). It is important
to handle the mouse appropriately and habituate it to the imaging setup prior to imaging. Habi-
tuation reduces the stress of the mouse caused by head fixation or new environments, and thus limits
strong motion artifacts and improves the reproducibility of the results (Box 1). Although animal
habituation helps to reduce variability, we recommend that behavior is always monitored using body
and facial cameras whenever possible. Because of this variability, it is beneficial to combine a large
number of sessions per and across mice to obtain a strong statistical result.
The second part of the protocol outlines how to run an imaging session. This includes how to set
up an fUS system, design the experiment, acquire data and handle the animal during the imaging
session (Fig. 3). We describe the hardware requirements, but more importantly, we offer a custom-
made graphical user interface to assist data acquisition and synchronization of the ultrasound system
with the motor, which is essential for scanning the brain. The exact design of the experiment (e.g.,
stimuli and number of brain slices) depends on the application. Here, we provide a detailed example
for acquiring the whole-brain activity elicited by a 1 s short visual stimulus.

3550 NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot

 NATURE PROTOCOLS PROTOCOL
a Step 9 d Step 17
Sagittal Superior sagittal
suture sinus

Skull bone Pial vessels

A
Cortex Dura mater
R

b Step 14 e Step 21
Dental
cement Dental
Burr drill cement

Thinned skull
Plastic polymer
ultrasound
Headpost compatible

c Step 16 f Step 22
Dumont fine
forceps

Protective
Piece of skull silicone rubber

Dura mater

g Step 25
h Step 26 Step 27 Step 28 Step 29

US probe

US gel

Headpost

Mouse
holder

i Narrow Cortical bulge Loss of perfusion Signal attenuation Signal saturation

cranial window after dura damage after cortical damage from air bubble from skull

D
R

Fig. 2 | Surgical procedure for chronic whole-brain imaging of awake head-fixed mice. a–f, Pictures and detailed illustrations of the surgical procedure
for headpost cementing, cranial window and bone replacement. Step 9: the skin and periosteum are removed, temporal muscles retracted and skull
bone cleaned (a). Step 14: the headpost is cemented to the skull. Squares of skull are thinned with a hand drill (b). Steps 16–17: the cranial window is
progressively opened by gently removing the square skull plaques with fine forceps (dotted black region). The dura mater and the surface of the brain
remain intact (c,d). Step 21: the cranial window is covered with a 50 µm plastic sheet (polymethylpentene, plastic polymer ultrasound compatible)
cemented to the headpost (e). Step 22: apply the silicone rubber over the plastic cover to protect the preparation and ease access for imaging (f). Step
25: picture of the mouse in its homecage before imaging session (g). Steps 26–29 from left to right: pictures of the mouse preparation for the imaging
session from head-fixation, removal of the silicone plug, ultrasound gel application and adjustment of the probe over the brain window. US, ultrasound
(h). Examples of problematic µDoppler images (from left to right) due to signal attenuation from a narrow cranial window (red dotted areas), surgical
damage provoking a bulge (red arrow) or some cortical damage (red dotted areas). Signal attenuation caused by an air bubble (red dotted area)
trapped in the ultrasound gel; and signal saturation from the skull (i). Troubleshooting guidance can be found in Table 1. A, anterior; R, right; D, dorsal;
R, right. Scale bars, 2 mm. All animal experiments were approved and conducted according to the Committee on Animal Care of the Catholic
University of Leuven, Belgium, the Veterinary Department of the Canton of Basel-Stadt, Switzerland, and European guidelines.

The third part of the protocol discusses the data analysis. We supply software for the complete
analysis of fUS data (Figs. 4 and 5; Supplementary Software (also available at https://github.com/nerf-
common/whole-brain-fUS)). Registration and segmentation processes are particularly important for
comparing datasets across sessions, animals and conditions, as for other whole-brain methods such as
fMRI41,42. The software includes a graphical user interface to assist with the registration of the
ultrasound data on a reference atlas: the Allen Mouse Common Coordinate Framework (CCF)43.

NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot 3551

 PROTOCOL NATURE PROTOCOLS

Box 1 | Mouse handling and head-fixed habituation ● Timing ~1 h

Presurgery handling and postsurgery habituation are time-consuming but dramatically reduce animal anxiety and
consequently increase both the quality and the repeatability of the acquisition.

(A) Presurgery mouse handling

Five days before surgery, handle mice daily following behavioral expert recommendations as described in
refs. 47,48. We progressively habituate mice to be nonaversively hand-cupped, and, once accustomed to the
experimenter (e.g., exhibiting grooming behaviors while hand-cupped), mice are progressively acclimatized to
walking through the body tube of the experimental fUS setup.

(B) Postsurgery head-fixation habituation

Once they have recovered from the surgical procedure (Step 25 of the Procedure), mice should be daily
habituated to head fixation49 after walking through the body tube. Fixation periods should be progressively
increased, from 10 min the first day up to 1 h after 3 d. In the case of signs of discomfort, such as excessive
movement and/or vocalization, the experimenter should stop the head-fixation session.

We also present a method for creating region-specific activity curves, based on a standardized list of ~250
regions from the Allen Mouse CCF (Fig. 6). Finally, we detail how to produce voxel-based and region-
based activity maps in response to a given stimulus, and present examples in the case of the visual
stimulation (Fig. 6). The outputs of this protocol (whole-brain activity maps) are given with a statistical
threshold, which assesses the significance of the stimulus-induced increase or decrease of activity com-
pared with baseline periods. As such, there is generally no need for a specific control group of animals.

Materials

Animals
●
Wild-type mice (we used 8-week-old C57BL/6 mice; Janvier Labs; the results presented here are from
male mice, but this protocol can be used with both males and females) ! CAUTION Any experiments
involving live mice must conform to institutional guidelines, and local and national regulations. All
experiments presented in this protocol were approved by the Committee on Animal Care of the
Catholic University of Leuven, Belgium, and the Veterinary Department of the Canton of Basel-Stadt,
Switzerland, and were in accordance with standard ethical guidelines (European Communities
Guidelines on the Care and Use of Laboratory Animals, 86/609/EEC). All efforts were made to
minimize the number of animals used.

Reagents
Drugs
! CAUTION All drugs listed are controlled substances and must be used according to institutional and
governmental guidelines.
● Isoflurane 1000 mg/g (Iso-Vet, Dechra, cat. no. 2909737) ! CAUTION Isoflurane is an anesthetic gas
with known adverse effects in humans and therefore must be handled in accordance with institutional
guidelines. Isoflurane should only be used in a designated area with appropriate ventilation and a
charcoal-based scavenging system. Isoflurane is suspected to reduce fertility and harm an unborn child.
Pregnant or breastfeeding personnel should avoid exposure to it.
● Fentanyl 0.05 mg/ml (Curamed, cat. no. GTIN 7680534840018)

● Medetomidine 1 mg/ml (Domitor, Orion Pharma, cat. no. 1070499)

● Midazolam 5 mg/ml (Ratiopharm, cat. no. PZN-04921530)

● Ketamine 100 mg/ml (Nimatek, Dechra, cat. no. 3120060)

●
Atipamezole 5 mg/ml (Revertor, Vibrac, cat. no. GTIN 04042668304324)
● Flumazenil 0.1 mg/ml (Sintetica, cat. no. GTIN 7680594000018)

● Naloxone 0.4 mg/ml (Orpha, cat. no. GTIN 7680569520015)

● Buprenorphine 0.3 mg/ml (Vetergesic, Ecuphar, cat. no. 263627)

● Dexamethasone 2 mg/ml (Rapidexon, Dechra, cat. no. 1421825)

● Emdotrim 10% sol (Ecuphar, cat. no. 1633163)

● Cefazolin 1 g (Sandoz, cat. no. 1676691)

● Enrofloxacin 22.7 mg/ml (Bayer Healthcare, cat. no. 08713254)

Reagents for surgical procedure

3552 NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot

 NATURE PROTOCOLS PROTOCOL
a b

y x Primary computer
Linear ultrasound probe Linear motor
Visual z Acquisition GUI
stimulus Manual
Receive data
Ultrasound electronics manipulator
(128 E/R channels) Generate Beamforming
US sequence
US probe
Linear motorized stage Motor control Filtering
Mouse holder
Stimulation computer Trigger out fUS images
Headpost

Suspended
table

c d One-pixel spectrum
6° One-pixel signal
4° –6° 1 1 Tissue
2.5 mm 0.1 mm Blood
12.8 mm 0 0.5
500 ms
+ +...+ –1 0
SVD + HP filters –250 0 250
Probe
Frequency (Hz)
element
1 One-pixel filtered spectrum
Acoustic Σ 1 One-pixel filtered signal
lens 0.5
0.2 0
mm
10 mm 500 ms 0
–1 –250 0 250
Compound image Frequency (Hz)
y x (2 ms) μDoppler image
z
2
I = ∫s dt Intensity
0.5 mm
130
Slow mode: 250 images/500 ms
Fast mode: 50 images/100 ms
70

e Dead time
f
E R Emission/reception
block
3.9 16.2 95.2 μs
Plane wave
95.2 190.4 285.6 μs

6° 4° 2° 0° –2° –4° –6° Compound image

0.2856 2 ms

1 2 49 50 μDoppler image
2 100 ms

1 2 3 4 35 69 70 Trial
0.1 Trig 7s

Trial 1 Trial 2 Trial20 Scan

0 Motor 2.5 min

Fig. 3 | Real-time whole-brain fUS of stimulus-evoked activity. a, Connectivity diagram of the fUS setup, illustrating the computers, hardware and
software required to perform real-time whole-brain imaging. The acquisition GUI generates the ultrasound sequence, receives the data acquired via
the ultrasound electronics and the ultrasound probe, and then beamforms, filters and processes the dataset to produce µDoppler fUS images in real
time. The GUI also controls the linear motorized stage, and the stimulation computer to move the ultrasound probe and trigger the visual stimulus,
respectively. Required tools and equipment are listed in ‘Equipment’ and the manual of the Supplementary Software. b, CAD schematic of the
experimental setup for performing whole-brain fUS in awake head-fixed mice (see Supplementary Data). c, Schematic representation of the
dimensions of the linear ultrasound probe and of a single coronal plane acquired using this probe. Top right: schematic of a whole-brain scan by
moving the probe along the antero-posterior axis. d, Diagram of the steps required to generate a single µDoppler image. Seven tilted plane waves are
transmitted in 2 ms to build one compound image, and 250 compound images are accumulated in 500 ms (or 50 images in 100 ms for the fast mode),
coherently added, decomposed into frequencies by a fast Fourier transform, filtered with high-pass and SVD filters to finally extract the intensity I of
the signal and then generate a µDoppler image. Plots presented here are one-pixel signals. Scale bar, 1 mm. e, Timeline of the experimental sequence
from ultrasound emission/reception block to functional scan. Box 2 provides a detailed description of the experimental sequence. f, Screenshot of the
miniScan software for acquisition of functional ultrasound images. The ‘Settings’ panel (left) allows the user to log the experiment, load the imaging
sequence, set motor parameters and control the slave computer to trigger the visual stimulus. The ‘Acquisition’ panel allows the user to generate
individual Bmode and µDoppler images, perform functional scans and check the status of the acquisition. All experimental information, saved images
and scans are logged in the ‘Journal’. More details can be found in the manual of the Supplementary Software. All animal experiments were approved
and conducted according to the Committee on Animal Care of the Catholic University of Leuven, Belgium, the Veterinary Department of the Canton of
Basel-Stadt, Switzerland, and European guidelines.

NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot 3553

 PROTOCOL NATURE PROTOCOLS

●
Iso-betadine solution (Meda, cat. no 82-217)
● Lubricant eye gel (Duratears, Novartis Pharma, cat. no. RVG 10187)
● Bone wax (CP Medical, cat. no. CPB31A)

● Ethanol 70% (VWR, cat. no. 83805) ! CAUTION Ethanol solution is highly volatile and flammable.

Careful handling is recommended.

a Registration software b Coronal planes

SSS SSS

LV
Select plane position
and adjust scale

Sagittal planes L
RTS RTS

A
Transverse planes

Apply modification RTS

RRRV

Save the transformation

matrix
LRRV
LTS R

Select the atlas to compare with Adjust colormap and range A

c Registration
d Analysis of a single session

Anatomical scan Imaging session

Functional Scan

Registration_ccf.m Steps Scan 1 Scan 2 Scan N

create affine transformation 39–43
Image_stability.m
Step
check the stability of the
Transformation matrix 44
signal

fUS files image_rejection.m

Step
Remove frames affected by
45
Functions.m animal movements

∑(Scans) Step
Processed data
46

Session average

Step Activity.m Segmentation_ccf.m Step

47 correlation or GLM registration + segmentation 49
Transformation
matrix
Step Register_data.m Select_brain_regions.m Step
48 50

Significant
Activity map Threshold Region–time
voxels

e Analysis of multiple sessions

Activity map 1 Activity map n Region–time 1 Region–time n

Steps Map_statistics.m Region_statistics.m Steps

51–52 averaging and t-test averaging and t-test 53–55

Mean Significant Mean Significant

activity map voxels Region–time regions

3554 NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot

 NATURE PROTOCOLS PROTOCOL
●
Glucose (Sigma-Aldrich, cat. no. G7021)
● Vetbond tissue adhesive (3M, cat. no. 70200742529)
● Resin cement kit, including catalyst V, monomer, polymer L-type radiopaque, red activator, adjustable

precision applicator brushes and sponge spear (Sun Medical Co., Ltd, cat. no. Super-Bond C&B)
! CAUTION Resin cement is highly volatile and flammable. Avoid skin contact, and limit vapor
inhalation. Careful handling is recommended.
● Polymethylpentene TPX plastic sheet, 50 µm (Goodfellow, cat. no. ME311051)

● Silicone rubber (Smooth-on, cat. no. Body Double-Fast Set)

● DietGel Recovery (ClearH2O, cat. no. 72-06-5022)

● Sterile saline (any)

● Compressed air duster (any) ! CAUTION Compressed air bottle is highly flammable. Careful handling
is recommended.

Reagents for imaging

●
Sterile saline (any) or artificial cerebrospinal fluid (Diacleanshop, cat. no. LERE-S-LSG-1000-1)
●
Ultrasound gel (Aquasonic Clear, Parker Laboratories, cat. no.03-08).

Equipment
Surgical setup, device and instruments
● Anesthesia system with induction chamber (TemSega, cat. no. MiniHUB V3)

● Bright-field stereomicroscope (Leica Microsystems, cat. no. A60)

● Stereotaxic frame and accessories (ear bars, nose clamp) (Stoelting, cat. no. 51600)

●
Physiosuite system for temperature control (Kent Scientific, cat. no. PS-CO2-RT)
●
Surgical trimmer (Stoelting Co., cat. no. 51470V)
● Scalpel handle and blades (Fine Science Tools, cat. no. 10003-12 and 10011)

● Fine straight scissors CeramaCut (Fine Science Tools, cat. no. 14958)

● Dumont fine and 45°-angled forceps (Fine Science Tools, cat. no. 11252-30 and 11251-35)

● 0.5 mm burrs (Fine Science Tools, cat. no. 19007-05)

● High-speed rotary micromotor (Foredom Electric Co., cat. no. K.1070)

● Hot bead sterilizer (Fine Science Tools, cat. no. 18000-45)

● Unmounted absorption triangles (Fine Science Tools, cat. no. 18105-03)

● Spatula (Fine Science Tools, cat. no. 10092-12)

● Sterile gel foam (any)

●
Sterile 0.5 ml syringes (any)
●
Sterile pipette (any)
● Sterile cotton tip applicators (any)

● Ruler and fine-point surgical marker (any)

fUS system
● Ultrasound scanner (Verasonics, cat. no. Vantage 128)
● Linear ultrasound transducer (Verasonics, cat. no. L22-14Vx or Vermon, cat. no. L15-Xtech)

● Host computer (AUTC cat. no. HPC-fUSI-2 or Verasonics, cat. no. host computer or custom

computer). The miniScan demo acquisition graphic user interface (GUI) can be operated on any
computer equipped with a video card (Nvidia, GTX-Series, or superior). Note that a high-performance
computing workstation (AUTC, cat. no. HPC-fUSI-2) is required to enable extended acquisition and
processing capabilities. This workstation includes a dual CPU (Intel Xeon), 4 GPUs (Nvidia,

Fig. 4 | Registration and processing of whole-brain fUS data. a, Screenshot of the GUI for registering the dataset with the Allen Mouse CCF.
b, Coronal, sagittal and transverse views of the affine transformation of the anatomical scan with the Allen Mouse CCF (white outlines). Red lines
delineate edges of the brain volume; red arrows give the position of the main brain vessels. Lines and arrows provide a good template for registering
the dataset and for further transformations. c, Registration of the anatomical scan with the Allen Mouse CCF generates a transformation matrix
applied in d. d, Workflow for a single imaging session analysis with steps for noise rejection and session averaging. Two distinct paths of computing
(including a registration step using the transformation matrix generated in a) enable either an activity map or region–time traces to be extracted.
e, Activity maps and region–time traces from multiple sessions are averaged to test whether voxels or regions are significantly activated in response to
the stimulus. Numbers refer to corresponding steps in the Procedure. A, anterior; D, dorsal; R, right; L, left; SSS, superior sagittal sinus; L/R-TS, left/
right transverse sinus; L/R-RRV, left or right rostral rhinal vein. LV, lateral ventricles. Scale bars: 1 mm. All animal experiments were approved
and conducted according to the Committee on Animal Care of the Catholic University of Leuven, Belgium, the Veterinary Department of the Canton of
Basel-Stadt, Switzerland, and European guidelines.

NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot 3555

 PROTOCOL NATURE PROTOCOLS

Sparse and randomly Frequent and randomly

distributed motion artifacts distributed motion artifacts Motion artifacts aligned with the trial

Intensity distribution Intensity distribution Intensity distribution

80 200 200
Threshold: 1.3 Threshold: 1.5 Threshold: 1.3

Images
Images

Images
40 Rejection: 1.9% 100 Rejection: 14.8% 100 Rejection: 4.8%

0 0 0
0.8 1.0 1.2 1.4 1.6 1 2 3 4 5 6 1.0 1.5 2.0 2.5
Normalized intensity Normalized intensity Normalized intensity

Image rejection Image rejection Image rejection

1 1 1
Planes

Planes

Planes
10 10 10

20 20 20
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Time (s) Time (s) Time (s)

Proceed with next step Control headpost and animal’s fixation Interpret results of data
improve habituation analysis with care

Fig. 5 | Data stability and motion artifact rejection. Detection of motion artifact and image rejection in three conditions that might be observed during
fUS of awake head-fixed mice (left to right: sparse and random, frequent and random, aligned with trial) and how to proceed further for each condition.
Details about threshold and image rejection can be found in Step 44 of the Procedure and in the manual of the Supplementary Software.

RTX-Series), DDR4-128 DIMM (Infineon), 1 TB SSD PCI-E NVMe (Samsung), 8 TB HDD SATA
(Western Digital)

Visual stimulation
● Stimulus computer (DELL, cat. no. Optiplex)

● 27″ LCD monitor (Iiyama, cat. no. ProLite XB2783HSU)

● USB-RS232 converter (StarTech, cat. no. ICUSB232V2)

fUS platform
●
Antivibration air-suspended table (TMC, cat. no. 63-500)
● Manual micromanipulator (Marzhauser Wetzlar, cat. no. MM33)

● Motorized-linear stage (Zaber Technologies, cat. no. T-LSM200A)

● Ultra-low head cap screws (Misumi, cat. no. CBSTSR2-3) and driver (iFixit, cat. no. IF145-299-4)

● CAD file of the stainless-steel headpost, see Supplementary Data, laser-cut in a 1 mm thick stainless-

steel sheet (200 × 200 mm; Salomon’s Metalen B.V., cat. no. Roestvrijstaal T-430)
● CAD file of the mouse holder, for 3D printing, see Supplementary Data

● CAD of the probe holder, for 3D printing, see Supplementary Data

● CAD experimental setup, see Supplementary Data

Software
●
Microsoft Win10 Pro 64 bits
● MATLAB (MathWorks, https://www.mathworks.com/products/matlab.html)

● CUDA toolkit (https://developer.nvidia.com/cuda-downloads)

● PsychoPy2 (http://www.psychopy.org)

● MiniScan, a custom-written MATLAB acquisition software, see Supplementary Software. See Github

repository for potential updates (https://github.com/nerf-common/whole-brain-fUS)44

● Custom-written MATLAB analysis codes, see Supplementary Software. See Github repository for

potential updates (https://github.com/nerf-common/whole-brain-fUS)44

Example dataset
The dataset analyzed in this protocol is freely available on Zenodo45 and analyzed following the
scripts in Supplementary Software (see Github/Zenodo repository for potential updates44). Follow the
instructions provided in the manual of the Supplementary Software to analyze the dataset as shown
in Fig. 6.

3556 NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot

 NATURE PROTOCOLS PROTOCOL
a Coronal view
b One session Eight sessions
c Coronal view Transversal view d
One session Eight sessions P < 0.001
1

Original map
10

Cbr-pl
20
30 VIS
40

HPF Cbr-spl
Corr. coef. 50
Transversal view
–0.5 0.5 LP/LGd 60
Filter and mask
P value < 0.05
70
80

STR
90
100
Corr. coef.
110
–0.5 0.5
120

TH
LP
Sagittal view
with μDoppler

L Corr. coef. 130

LGd
Overlay

140

HTH
A 0 0.2
Sagittal view 150
D VIS 160
Corr. coef.

MB
R 170
0 0.5 SC
180
15 20 190
VISp LGd SCs
15 200
10 SC

HB
210
z-score

10
5 220 z-score
5 10
230
0

Cbl
0
240 –10
–5 –5 D D 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
R A Time (s) Time (s)
Time (s) Time (s) Time (s)

Fig. 6 | Example of processed whole-brain fUS data. a, Superimposition of registered coronal, transverse and sagittal anatomical views with regions of
the Allen Mouse CCF, outlined in white (top to bottom; see Fig. 4a,b for registration software and landmarks for a correct alignment). b, Examples of
visually evoked activity maps for one coronal brain slice (top to bottom: original, filtered, and overlaid with the µDoppler image) for single- and
multisession average (respectively left and right columns). Regions from the Allen Mouse CCF are overlaid (black outlines). c, Coronal, transverse and
sagittal activity maps from multiple sessions from the same mouse depicting only visually evoked activated voxels (P < 0.05, eight sessions) with
segmented brain regions from the Allen Mouse CCF (black outlines). d, Left: region–time traces of 242 brain regions from single and multiple sessions
from the same mouse. Right: activated brain regions with a P value < 0.001 (GLM followed by t-test across sessions). Brain regions are ordered by
major anatomical structures and listed in Supplementary Table 1. Vertical line, stimulus start; horizontal line, stimulus end. e, Average response curves
from the VISp, the lateral geniculate nuclei (LGd) and the SCs. Data are mean z-score ± standard error of the mean. The gray vertical bar represents
the visual stimulus. A, anterior; L, left; R, right; D, dorsal. Scale bars, 1 mm. All animal experiments were approved and conducted according to the
Committee on Animal Care of the Catholic University of Leuven, Belgium, the Veterinary Department of the Canton of Basel-Stadt, Switzerland, and
European guidelines.

Reagent setup
Fentanyl/medetomidine/midazolam (FMM) solution
Mix 1 ml of fentanyl (0.05 mg/kg), 0.5 ml of medetomidine (0.5 mg/kg) and 1 ml of midazolam
(5.0 mg/kg) with 7.5 ml distilled sterile water to a final concentration of 10 ml/kg. Inject 0.2 ml/20 g
body weight. FMM solution can be stored at 4 °C for a couple of weeks. ! CAUTION Listed drugs are
controlled substances and must be used according to institutional and governmental guidelines.

Atipamezole/flumazenil/naloxone (AFN) solution

Mix 0.5 ml of atipamezole (1.25 mg/kg) 5 ml of flumazenil (0.25mg/kg) and 3 ml of naloxone
(0.6 mg/kg) with 1.5 ml distilled sterile water to a final concentration of 10 ml/kg. Inject 0.2 ml/20 g
body weight. AFN solution can be stored at 4 °C for a couple of weeks. ! CAUTION Listed drugs are
controlled substances and must be used according to institutional and governmental guidelines.

Ketamine/medetomidine solution (can be used as an alternative to FMM anesthesia)

Mix ketamine (100 mg/ml) with medetomidine (100 mg/ml) and 8 ml sterile saline, and inject
0.1 ml/20 g body weight intraperitoneally. Ketamine/medetomidine solution can be stored at 4 °C for
a couple of weeks. ! CAUTION Listed drugs are controlled substances and must be used according to
institutional and governmental guidelines.

Glucose–saline solution
Mix glucose powder into saline to yield a 5% (wt/vol) solution. Vortex to dissolve glucose crystals.
The solution can be stored at 4 °C for several weeks if sterility is preserved. For larger volumes, small
aliquots can be frozen until needed. Defrost to room temperature (~20–23 °C) before use.

NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot 3557

 PROTOCOL NATURE PROTOCOLS

Superbond C&B
Follow the instructions of the user guide provided with the product: dispense one spoon of the
polymer L-type radiopaque powder with four drops of monomer in the ceramic dish. Add one drop
of catalyst V, and mix with an application brush. Extend the window of cement pliability by cooling
down the ceramic tray with ice. Clean the ceramic dispensing dish with 70% ethanol before and
after use. ! CAUTION Resin cement is highly volatile and flammable. Avoid skin contact, and limit vapor
inhalation. Careful handling is recommended.

Silicone rubber
In a sterile dish, dispense equal amounts of parts A and B (1:1 ratio), mix to homogenize and apply
quickly with a spatula.

Equipment setup
Surgical instruments
Clean and sterilize surgical instruments and the headpost by autoclave or using a hot bead sterilizer
before starting the surgery. Gel foam, cotton tip applicators and absorption spears must be autoclaved
prior to the surgery. Clean surgical and imaging workspaces with 70% ethanol.

Ultrasound setup
Follow the instructions detailed in the miniScan manual (our demo acquisition GUI, see Fig. 3 and
Supplementary Software) to install the hardware and software schematized in Fig. 3a. In brief, the fUS
scanner is composed of four main parts: (i) the ultrasound transducer; (ii) the ultrasound electronics;
(iii) the high-performance computing workstation and (iv) the software package. Additional com-
ponents include a linear motor for moving the transducer and a second computer used here for
managing visual stimulation.
This hardware–software configuration produces images of the brain vasculature with a resolution
of 100 µm laterally (x), 110 µm in depth (z) and 300 µm in elevation (y), as previously measured6
(Fig. 3c). The ultrasound sequence generated by the software is the same as that used in Macé et al.6
and is detailed in Box 2.
The system has a default set of parameters optimized for imaging the mouse brain with the
suggested hardware, as in the example presented here. The parameters can be modified through the
file ‘parameterUser.m’ (see manual of the Supplementary Software). This file enables the settings of
the system to be changed, including changing the probe, modifying the ultrasound sequence,
changing spatial and temporal resolutions, customizing the trigger and randomizing the mechanical
scanning of the brain slices. We strongly recommend starting with the default settings before making
any modifications in the sequence.
Communication between the computers is via a serial port. When the acquisition GUI is launched,
it automatically opens a serial port for that purpose. At each brain slice acquisition, a trigger is sent
out to the external device (here the visual stimulation computer) to start the stimulus. The user can
specify when the trigger should be sent with regard to the fUS acquisition in the parameter file
(‘parameterUser.m’, see manual of the Supplementary Software). Moreover, another trigger can be
sent through the BNC OUT port of the scanner at every ultrasound emission.
All imaging equipment, including the ultrasound probe and mouse holder, should be attached
securely on an antivibration table. Equipment that may generate vibrations (fans or motors) must not
be placed on the antivibration table. It is also recommended to ground this table to reduce electronic
interference with the ultrasound probe.

Procedure

Animal preparation ● Timing ~3–4 h

! CAUTION Procedures involving live animals must be performed by trained experimenters and follow
institutional guidelines, and local and national regulations.
CRITICAL To ensure an efficient fUS imaging session under awake conditions, we recommend
c

undertaking handling habituation by the experimenter for 5 days prior to surgery, as described
in Box 1A.
CRITICAL Inject the mouse with 5% (wt/vol) glucose solution (200 µl/25 g body weight)
c

3558 NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot

 NATURE PROTOCOLS PROTOCOL
Box 2 | Ultrasound sequence
The fUS sequence has six nested levels. The timeline of the sequence is shown in Fig. 3e. Here we describe in detail each step of the default
sequence for advanced users.
Emission/reception block. A short ultrasound pulse of two wavelengths is sent out by all the elements of the probe with a specific delay to
generate a tilted plane wave that propagates through the brain. Based on the speed of sound, the backscattered echoes recorded by the probe from
3.9 µs to 16.2 µs after emission correspond to depths between 3 mm and 11 mm. A dead time is included between emissions to ensure a fixed
framerate of 500 Hz for the compound image.
Plane-wave block. A plane-wave acquisition block is obtained by repeating the emission/reception block three times and averaging the received
signal to increase the signal-to-noise ratio.
Compound image. To increase the resolution of the image, seven tilted plane-wave blocks (−6° to 6° in steps of 2°) are acquired at a rate of
3.5 kHz. The echoes of the seven plane-wave blocks are beamformed to create seven plane-wave images and coherently added to generate a
single compound image with high resolution (Fig. 3d, left)5.
µDoppler image. The compound images are repeated continuously at a framerate of 500 Hz (Fig. 3d, left). Each set of 50 (fast mode) or 250 (slow
mode) compound images is filtered to extract the blood signal (Fig. 3d, center) using a high-pass filter and an SVD filter5. Finally, the intensity of
the filtered images is averaged to obtain a vascular (or µDoppler) image in 0.1 s/10 Hz (fast mode) or 0.5 s/2 Hz (slow mode) (Fig. 3d, right).
Trial. A trial consists of acquiring N µDoppler images at a specific brain slice. In our example data, 70 images are acquired in 7 s using the fast
mode for each slice. A trigger is sent at the image Ni to the stimulation computer, here at image 35, starting the 1 s long visual stimulus (Fig. 3a).
Scan. To obtain a brain-wide functional scan, the probe is moved using the motor (Fig. 3a,b) by steps of 300 µm, and a new trial is performed at
each position. The brain slices can be acquired in a sequential or randomized way. The full scan is obtained in ~2.5 min. Note that the recording
time of a scan depends linearly on the number of slices and that a scan requires the stimulus to be repeated at each brain slice. Note also that
~300 µm between slices is optimal given the resolution of the ultrasound probe, and that oversampling will not increase the spatial resolution. Here
six scans per session were acquired, leading to a recording time of ~15 min.
The software offers a fast mode that uses 50 compound images per µDoppler image (10 Hz) and a slow mode that uses 250 compound images
(2 Hz) (Fig. 3b). Importantly, there is no change of signal-to-noise ratio between the two modes (averaging five frames from the fast mode is
equivalent to one frame of the slow mode). The fast mode is better suited to measure responses to short stimulus or timing differences between
regions, but produces a higher data volume than the slow mode.
The default settings of the sequence are optimized for the probes Vermon L15-Xtech (or Verasonics L22-14Vx) and for imaging the mouse brain.
However, advanced users can set their sequence and adapt it for different configurations by changing the parameter file (see ‘parameterUser.m’).
In particular, the parameters can be changed to cover different frequencies ranges and penetration depths, the trigger can be customized and the
acquisition of the brain slices can be randomized (see the manual in the Supplementary Software).

intraperitoneally to prevent dehydration during surgery. Give the first injection 2 h after the start of
surgery, and repeat every 2 h.
1 Fill the induction chamber with 4% (wt/vol) isoflurane. Place the animal inside the induction
chamber and wait ~3 min until the mouse becomes unconscious.
! CAUTION Isoflurane is an anesthetic gas with known adverse effects in humans and therefore must
be handled in accordance with institutional guidelines. The use of isoflurane should be conducted in a
designated area with appropriate ventilation and charcoal-based scavenging systems.
2 Inject FMM solution s.c. (0.2 ml/20 g body weight, see ‘Reagent setup’), switch off isoflurane and
move the mouse back to its cage while the drug takes full effect (~5 min). When the anesthesia is
confirmed by lack of reflexes, transfer the mouse to the stereotaxic frame.
CRITICAL STEP As an alternative to FMM anesthesia, you can instead inject ketamine/
c

medetomidine solution i.p. (0.1 ml/20 g body weight).

3 Place the animal in a prone position on a heating pad set at 37.0 ± 0.5 °C, and fix the head in the
stereotaxic frame. Adjust the animal’s body such that the head is slightly elevated with respect to the trunk.
! CAUTION Under anesthesia, animals are unable to regulate their core temperature, potentially
leading to hypothermia. The core body temperature should thus be maintained by use of a heat pad
throughout anesthesia; check body temperature throughout using a rectal thermometer.
! CAUTION Excessive elevation may cause breathing problems. To prevent other breathing issues,
pull the tongue out and to one side with dull forceps.
4 Test the depth of anesthesia by pinching the animal’s toe; the anesthesia should be deep enough for
this to elicit no response. When anesthesia is confirmed, administer i.p. injection of dexamethasone
(0.5 mg/kg body weight), and wait 5 min for it to take effect. Lubricate both eyes with Duratears
ointment to avoid dehydration.
5 Shave the top of the animal’s head, then clean and wipe it with an iso-betadine solution using a
sterile gel foam.
6 Make a midline incision with a scalpel blade from between the eyes to the back of the head.
Expose the entire width of the skull by removing both the skin and the periosteum using
fine straight scissors.

NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot 3559

 PROTOCOL NATURE PROTOCOLS

7 Under the stereomicroscope, insert closed angled forceps between the temporalis muscle and the
skull a few millimeters anterior to the bregma. Run the angled forceps gently back to lambda.
8 Apply Vetbond tissue adhesive to the retracted muscles and the edges of the incision to secure the
skin (Fig. 2a).
CRITICAL STEP Avoid getting glue near the eyes, and be careful not to glue tissues (fur, skin or

c
ears) to the stereotaxic frame.
CRITICAL STEP Retraction of lateral muscles remains essential to assess lateral parts of the skull

c
and enlarge the imaging field of view (Fig. 2i).
9 Dry the skull, and scrape it with the scalpel blade.
10 Adjust the headpost over the skull. Glue the headpost in the desired position using the tissue adhesive
Vetbond. Wait 5 min to ensure good adhesion of the headpost to the skull before applying cement.
CRITICAL STEP It is essential to prepare the skull well and glue the headpost firmly to ensure the
c

headpost is stable after the cement is applied.

CRITICAL STEP Make sure the headpost is parallel to the skull.
c c

CRITICAL STEP The design and position of the headpost determines how much of the brain can
be imaged. This step may be adapted to optimize the field of view depending on the application.
11 Mix the cement as described in ‘Reagent setup’, and cement the headpost to the skull. Wait 10 min
to ensure it is fixed firmly. Screw the headpost to the mouse holder, release and remove both the
mouth clamp and ear bars of the stereotaxic frame.
CRITICAL STEP Cool down the ceramic tray with ice to keep the cement pliable for longer.
c

! CAUTION Resin cement is highly volatile and flammable. Avoid skin contact, and limit vapor
inhalation. Careful handling is recommended.
12 With a 0.05 mm burr, mark the surface of the skull by drilling very gently, to outline the shape and
position of the window. Similarly, divide and mark the window into small 3 × 3 mm square plaques.
13 Drill skull plaques with a 0.05 mm burr by applying slight pressure to the skull with gentle
movements.
! CAUTION To prevent drill-induced overheating of the brain tissue and clear bone debris, apply
sterile saline frequently throughout the surgical procedure. However, always drill when the bone is
dry as it is easier and more precise.
CRITICAL STEP Due to skull stratification, the bone might bleed during the thinning process.
c

Avoid long periods of bleeding.

? TROUBLESHOOTING
14 Keep drilling until a very thin and homogeneous preparation has been obtained (Fig. 2b). Apply
sterile saline to make the bone more transparent: if no blood vessels can be seen through the
thinned skull, repeat Steps 13–14 as necessary.
CRITICAL STEP Note that well-performed thinning surgery substantially reduces the risk of tissue
c

inflammation.
! CAUTION At this step, any mistake could break the thinned skull and damage the dura mater and
the brain surface (Fig. 2i).
! CAUTION The sagittal suture should be treated with caution; any mistake could rupture the
superior sagittal sinus and kill the animal by exsanguination.
? TROUBLESHOOTING
15 Crack the thinned skull at a corner of the window. Insert the angled forceps between the square of
skull and the dura mater to dissociate them.
! CAUTION Do not puncture the dura mater. Do not let the dura mater dry.
16 Carefully remove one square of skull after the other in one hemisphere. Continuously clean and
rinse the brain with sterile saline (Fig. 2c).
17 Continue to carefully remove the skull plaques from the second hemisphere (Fig. 2d). Delicately
insert a 45°-angled forceps right under the skull, between the bone and the dura mater. Scratch
carefully the bottom part of the skull to disconnect the dura mater from the bone. Proceed mm after
mm, until the two tissues along the superior sagittal sinus have been completely disconnected.
CRITICAL STEP The sagittal suture should be treated with caution as the superior sagittal sinus
c

lies under it. Special care must be taken while crossing the bregma suture; the dura mater strongly
connects with the bone and might be damaged if not disconnected carefully. If this procedure seems
too risky, a thin band of bone can be left over the sinus. However, bear in mind that the remaining
strip will result in attenuation of the signal in the medial part of the brain during the imaging
sessions. Before continuing, we recommend waiting 10 min to ensure that no bleeding arises.
? TROUBLESHOOTING

3560 NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot

 NATURE PROTOCOLS PROTOCOL
18 Fill the window with 5 µl enrofloxacin solution (5 mg/kg), wait 10 min for it to take effect and then
rinse with sterile saline.
19 Cut the plastic sheet with scissors to a size that fits the cranial window.
20 Carefully position the plastic sheet on the cranial window, and glue the edges to the bone.
! CAUTION Because of the curvature of the brain, it is possible for air to get trapped between the
plastic sheet and the brain. If the plastic sheet is well sealed, these air pockets will fill with
cerebrospinal fluid over the following days. The window should be fully transparent (no air gaps) at
the time of imaging; thus it is important the sheet is well sealed.
21 Mix the cement as described in ‘Reagent setup’, and cement the plastic sheet to the bone. Wait
10 min to ensure the cement grips well (Fig. 2e).
22 Prepare the silicone rubber as described in ‘Reagent setup’, and quickly fill the window up with it.
Wait 10 min until it hardens (Fig. 2f).
23 Inject AFN solution s.c. (0.2 ml/20 g body weight, see ‘Reagent setup’) to reverse the FMM anesthesia.
If the alternative ketamine/medetomidine anesthesia has been used, inject atipamezole (1.0–2.5 mg/kg)
to reverse the anesthesia. Inject the painkiller buprenorphine i.p. (0.05–0.1 mg/kg) and the antibiotic
cefazolin (6.25 mg/kg). Add 1 ml of antibiotic emdotrim to 300 ml of drinking water.
24 Remove the rectal probe, and free the animal from head fixation.
25 Allow the mouse to recover from the anesthesia in its cage, placed on a heating pad, and observe
regularly (Fig. 2g). Ease the mouse’s recovery by providing dry food soaked in water and recovery
gel to facilitate chewing and hydration.
PAUSE POINT We recommend at least 5 d of recovery before starting head-fixation habituation
j

and imaging. During the first 3 d, we provide buprenorphine, dexamethasone and cefazolin and
inject dexamethasone for an additional 2 d.

Imaging session ● Timing ~45 min

CRITICAL To ensure an efficient fUS imaging session under awake conditions, we recommend
c

undertaking head-fixation habituation prior to the imaging session for a minimum of 3 d, as described
in Box 1B.
CRITICAL Throughout the procedure, buttons of the acquisition GUI are indicated as [GUI panel/
c

Button]. An image of the GUI is shown in Fig. 3f and Supplementary Software.

26 On the day of imaging, allow the animal to walk into the tube. Hold the headpost, and screw it to
the mouse holder (Fig. 2h).
? TROUBLESHOOTING
27 Clean both the protective silicone rubber and the headpost with 70% ethanol using cotton tips. Rinse
with sterile saline. Insert 1 mm of the 45°-angled forceps between the silicone rubber and the cement.
Disconnect the silicone rubber from the cement all around the inner window of the headpost. Add a
drop of sterile saline to smooth the disconnection of the silicone with the plastic window, and use fine
forceps to catch and slowly remove the silicone plug from the headpost (Fig. 2h)
! CAUTION Take care while removing the silicone rubber not to remove the plastic sheet or damage
the brain tissue.
28 Cover the imaging window with ultrasound gel (Fig. 2h).
! CAUTION Air bubbles trapped in the gel are echogenic and thus attenuate the signal (Fig. 2i).
Load the ultrasound gel in a vial, and centrifuge (~500g for 2 min) before applying.
29 Place the animal under the ultrasound probe. Roughly adjust the ultrasound probe to the imaging
window using the manual micromanipulator (Fig. 2h)
! CAUTION To avoid scratches on the probe and compressing the animal’s brain, do not press the
probe down against the headpost. The probe does not need to be in contact with the brain; ideally,
place it ~3 mm above the surface.
30 In MATLAB, open the acquisition GUI (‘start_miniScan.m’, Fig. 3f) that enables recording of
different types of images (Bmode, µDoppler and Functional in single plane or full brain volume).
Set the file name and the communication ports with the motor [GUI panel: motor settings/Button:
serial port] and with the stimulus computer [GUI panel: stimulus/Button: serial port]. Select the fast
mode to acquire at 10 Hz; otherwise, images are acquired in slow mode at 2 Hz. A description of
how to use this software can be found in the manual of the Supplementary Software. The default
settings are optimized for this experimental example, but users can modify most of the parameters
(probe, ultrasound sequence, trigger, random or sequential scanning, etc.) by modifying the
‘parameterUser.m’ file (see manual of the Supplementary Software).

NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot 3561

 PROTOCOL NATURE PROTOCOLS

! CAUTION To avoid damage to the probe, free the space around it while initiating the motor.
! CAUTION High emission power can damage the probe. Be aware that increasing the voltage or the
duty cycle might damage the probe.
31 Switch on the button [GUI panel: single-plane/Button: Bmode] for real-time display of
morphological images, and adjust the ultrasound probe with the manual micromanipulator to
center the brain in the middle of the image.
32 Check the imaging quality using single plane µDoppler images with the button [GUI panel: single-
plane/Button: Doppler] while moving along the cranial window using the button [GUI panel:
motor settings/Button: move].
CRITICAL STEP This step checks whether there is any brain damage caused by the cranial
c
window implantation and whether any air bubbles or other artifacts are present that will degrade
the image quality. Examples of such cases are shown in Fig. 2i.
? TROUBLESHOOTING
33 Move the ultrasound probe to the anterior edge of the cranial window to define the first plane of the scan.
! CAUTION For the sake of reproducibility over sessions, users are encouraged (i) to identify a brain
slice that can easily and reproducibly be localized by the user using the single-plane µDoppler
mode, (ii) to navigate to this slice at the beginning of the imaging session and (iii) to use this slice as
a reliable origin. Then, the user should identify the coordinates of the anterior and posterior limits
of the cranial window with respect to the origin using the single-plane µDoppler mode. These limits
can be identified by a typical signal attenuation due to the presence of the skull (Fig. 2i). The
registration step of the analysis will ensure a fine alignment of the scans across sessions, but
following this procedure will avoid excessive shifts that could affect the extreme parts of the scan
upon averaging, particularly when the number of slices is optimized to exactly span the cranial
window size. Finally, note that the motor works in both directions. Be sure to identify the direction
of the motor with regard to the mouse position. Here we assume that increasing values correspond
to a movement of the probe from the anterior to the posterior part of the brain.
34 Run the anatomical scan [GUI panel: scan-volume/Button: Doppler] with a 100 µm step size
between planes. The anatomical scan records a single µDoppler image at each position of the probe
and saves the data as a single 3D matrix file. Example settings of the anatomical scan can be found
in the manual of the Supplementary Software.
? TROUBLESHOOTING
35 Run the functional scan [GUI panel: scan-volume/Button: functional] with a 300 µm step size
between planes (typically 20 planes). The functional scan acquires 70 µDoppler images at 10 Hz in
fast mode (or 2 Hz when the fast mode button is unchecked) in each imaging plane and sends a
trigger to start the stimulus (trigger sent at image 35). Data is saved as a single 4D matrix file (3D +
time). Settings of the functional scan can be found in the manual of the Supplementary Software.
! CAUTION When using a new stimulation paradigm, for the first functional scan acquisition, check
that the stimulation is correctly triggered before acquiring multiple scans.
? TROUBLESHOOTING
36 Acquire additional functional scans by repeating Step 35, or move to the next step to close the
imaging session.
37 Move the ultrasound probe away from the mouse using the manual micromanipulator, wipe the gel
away and clean the plastic window.
38 Prepare the silicone rubber as described in ‘Reagent setup, and quickly fill the cranial window with
it. Wait 10 min until hardened. Unscrew the headpost from the mouse holder, and put the animal
back in its homecage. Cleaning and grooming behaviors are signs of good welfare.
PAUSE POINT Allow at least a 24-h rest period before the next imaging session.
j

Analysis of a single session ● Timing ~1 h/session

CRITICAL In this section we provide a complete analysis package for data processing running with
c

MATLAB including: (i) the registration of the fUS dataset with the Allen Mouse CCF reference atlas
(Fig. 4a–c), (ii) the visualization of activity using correlation maps, (iii) the segmentation of the data into
anatomical regions and (iv) the statistical procedure for coherent analysis of a single session or multiple
sessions (Fig. 4d,e). Detailed documentation about files and functions used hereafter can be found in the
manual of the Supplementary Software. The output of a single fUS session consists of one anatomical
scan A (x,y,z) and a set of functional scans Fi (x,y,z,t), all used in the following analysis steps.

3562 NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot

 NATURE PROTOCOLS PROTOCOL
Registration with the reference brain atlas
39 Open the registration GUI (‘registration_ccf.m’) to register the anatomical scan A(x,y,z) with the
Allen Mouse CCF (see ‘example01_registering.m’).
40 Use the coronal view of the scan to center the volume on the atlas. Move the scan with the
computer mouse (left-click for translation, right-click for rotation), and adjust the surface of the
brain to the outline of the atlas. If needed, scale the (x,y,z) dimensions to fit the outlines of the atlas.
The registration GUI is illustrated in Fig. 4a.
41 Rotate along the anteroposterior axis to align the superior sagittal sinus to the sagittal atlas outlines.
If needed, rotate along the mediolateral axis to align the brain midline.
? TROUBLESHOOTING
42 Adjust the position of the brain scan to the atlas using brain landmarks, such as lateral ventricles,
hippocampus and large vessels.
43 Repeat Steps 40–42 to ensure correct registration of the anatomical scan. An example of correct
registration with landmarks is provided in Fig. 4b.
The output of the ‘registration_ccf.m’ function is an affine coordinate transformation:
~ r þ~
r0 ¼ M~ a
where~r ¼ ðx; y; z Þ are the original coordinates of the anatomical scan, M is the rotation and scaling
matrix and ~
a the translation vector.
This affine transformation can be applied to any matrix having both the same voxel size as the
Allen Mouse CCF (i.e., 50 µm3 isotropic) and the same axis origin as the anatomical scan.
? TROUBLESHOOTING
Data stability and frame rejection
CRITICAL These steps are important as they check the stability of the data and identify images
c

affected by the movement artifacts (see ‘example02_filter_average.m’).

44 Run the function ‘image_stability.m’. This function computes the average value of each frame,
displays a histogram of the average values and fits it with a Gaussian curve (Fig. 5). Values higher
than three times the standard deviation are considered as outliers due to movement artifacts. The
position (plane and time) of the outliers is also displayed in a figure.
PAUSE POINT Three scenarios are possible. If there are few outliers (<10%) randomly
j

distributed, the acquisition can be considered as ‘stable’ (Fig. 5, left). A high number of outliers
(>10%) indicates a mechanical instability (Fig. 5, middle). In this case, check both the headpost and
the animal fixation. If the outliers systematically happen around the stimulus time, it suggests that
the artifact is generated by a behavioral movement related to the task, and this time period must be
considered as a blind time as the data cannot be recovered through averaging (Fig. 5, right).
? TROUBLESHOOTING
45 Run the function ‘image_rejection.m’. This function eliminates the outlier frames identified as in
Step 44 and replaces them by an interpolation of the neighboring accepted frames.
Session average
46 Average all the motion-filtered scans from the same session (see ‘example02_filter_average.m’).
1 XN
Faver ðx; y; z; t Þ ¼ F ðx; y; z; t Þ
i¼1 i
N

These scans all share the same transformation matrix for registering.

Extract activity maps

47 This process enables visualization of the activated voxels. Follow option A to perform Pearson’s
correlation or option B to run a generalized linear model (GLM) analysis. Choose option A if you
have a single type of stimulus (for example, here a flickering checkerboard presented for 1 s).
Choose option B if you have multiple types of stimuli (for example, drifting gratings of different
directions). Option B (GLM) is a more flexible method of generating activity maps as it can be
extended to multiple stimuli or events by using multiple regressors. Note that option B also works
in the case of one stimulus type, in which case it gives the same result as option A.
(A) Pearson’s correlation.
(i) Run the function ‘map_correlation.m’ (see ‘example03_correlation.m’) to compute the
correlation of the session dataset with the stimulus. This function defines the stimulus

NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot 3563

 PROTOCOL NATURE PROTOCOLS

pattern s(t) as the correlation of the stimulation window w(t) by the hemodynamic
response hrf(t),
sðtÞ ¼ wðtÞ  hrf ðtÞ;
normalizes the data and the stimulus with zero-means
1 Xn
F aver ðx; y; z; tÞ ¼ Faver ðx; y; z; tÞ  F ðx; y; z; tÞ;
i¼1 aver
n
1 Xn
sðtÞ ¼ sðtÞ  sðtÞ;
n i¼1

computes the Pearson’s correlation in each voxel of the image. The output of the function is
a volumetric correlation map:
Pn
i¼1 F aver ðx; y; z; tÞsðtÞ
cðx; y; zÞ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pn 2 Pn 2 :
ffi
F
i¼1 aver ðx; y; z; tÞ i¼1 s ðtÞ

A default model of the hrf is provided in the script (based on the hrf from the SPM software46).
Voxels with c > 0.23 are considered as significantly activated for this example. This
threshold is defined by selecting a P-value threshold of 0.05, converting the P-value
threshold into the corresponding z-score using a two-tailed t-test (zp= 2.8), and computing
ffiffiffiffiffiffiffiffiffiffiffi
a z-score value from the correlation using Fisher’s transform z ¼ n  3ArctanhðcÞ, with
n being the number of samples. For n = 70 and z = 2.8, the threshold is c = 0.23.
(B) Run a GLM.
(i) Run the function ‘map_glm.m’ (see ‘example04_glm’). This function defines a model with
two regressors
yðtÞ ¼ β1 X1 ðtÞ þ β2 X2 ðtÞ;
where the first regressor is the stimulus convolved with the hemodynamic response
X1 ðtÞ ¼ wðtÞ  hrf ðtÞ;
and the second regressor is a constant X2(t) = 1.
The function fits each voxel of the averaged scan Faver(x,y,z,t) with this model using the
Matlab function glmfit (no regularization step used). The output is a set of estimators for
each voxel β1(x,y,z), β2(x,y,z) and the t-score map corresponding to each estimator. The
standard outputs of the GLM can be used for visualization of activity and contrasts between
stimuli, and for statistical analysis at the subject or group level.
48 Register the activity map with the Allen Mouse CCF by running the function ‘register_data.m’. This
function interpolates anisotropic voxels of the scan (100 × 110 × 300 µm) to isotropic voxels of the
Allen Mouse CCF (50 × 50 × 50 µm) to generate a 3D matrix cint(x,y,z) and transforms the
interpolated data into the register coordinates Cð~ r0Þ ¼ cint ð~ r ¼ M 1 ð~
rÞ, where ~ r0  ~
aÞ. The
registered data can be displayed with the anatomical outlines (see ‘draw_borders.m’).

Analysis of region-time traces

CRITICAL This analysis enables temporal activity to be analyzed in each anatomical region defined by
c

the Allen Mouse CCF (see ‘example05_segmentation.m’).

49 Run ‘segmentation_ccf.m’ to segment the averaged scan. This function interpolates the functional data
from a voxel size of 48 × 100 × 300 µm to an isotropic voxel size of 50 × 50 × 50 µm. The data are
resampled to the same voxel size as the reference atlas to make them compatible with the registration
procedure of Steps 39–43. Then, the function registers the interpolated data using the affine
transformation matrix obtained after Step 43. The functional data is interpolated and registered as follows:
Fint ðx; y; z; t Þ ¼ interpolateðFAver Þ;
 
r; t Þ ¼ Fint M 1 ð~
Freg ð~ r ~
aÞ; t :

All voxels from the same region are then integrated as follows:
X
Sðk; t Þ ¼ F ð~
r 2 region k reg
~
r; t Þ

3564 NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot

 NATURE PROTOCOLS PROTOCOL
50 Run the function ‘select_brain_regions.m’ to extract a selection of brain regions. Generate a
region–time matrix z(k,t), normalized as a z-score with zero average and standard deviation 1 in the
baseline,
Sðk; t Þ  SðkÞ
z ðk; t Þ ¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ;
Pn 2
Sðk; t Þ  SðkÞ
t¼1
P
n
with SðkÞ ¼ n 1
Sðk; tÞ and n being the number of samples in the baseline.
t¼1

Analysis of multiple sessions ● Timing ~1 h

CRITICAL This section focuses on the analysis of multiple sessions from a single mouse. This section
c

should be followed after each session i has already been through Steps 39–50, and both the activity map
Ci(x,y,z) and a region–time matrix zi(k,t) have been extracted. This analysis aims to obtain a statistical
view of what is significantly activated over multiple imaging sessions.

Activity maps
CRITICAL The activity maps Ci(x,y,z) from each session i are extracted after running Steps 39–47 of
c

the Procedure. We used the activity maps based on correlation in this procedure, but it can be
conceptually applied in the same way to activity maps (t-score maps) extracted using the GLM.
51 Run the function ‘map_statistics’ (see ‘example06_multisession_map.m’) to compute the mean
value Cmean, the standard deviation σ and the t-score tscore of Ci in each voxel
1 XN
Cmean ðx; y; z Þ ¼ C ðx; y; zÞ;
i¼1 i
N
1 XN
σ 2 ðx; y; z Þ ¼ ðCi ðx; y; z Þ  Cmean ðx; y; zÞÞ2 ;
N i¼1

Cmean ðx; y; z Þ
tscore ðx; y; z Þ ¼ :
σ ðx; y; z Þ

Then, compute the P value with a Student’s distribution

pðx; y; z Þ ¼ 1  tcdf ðtscore ðx; y; z Þ; N Þ;
where tcdf is the cumulative distribution function of the Student’s t distribution. For a low number
N, a pixel is considered significantly activated if P < 0.05. For larger numbers, one should apply a
post hoc multiple comparison of the P value (e.g., Bonferroni, false discovery rate) to avoid false
positives.
52 Display the significantly activated voxels of the averaged activity map with the anatomical outlines
(see ‘draw_borders.m’).

Analysis of region–time traces

CRITICAL The region–time matrices zi(k,t) from each session i are extracted after running Steps
c

48–50 of the Procedure.

53 Run the function ‘segmentation_statistics.m’ to perform a statistical study over all the sessions (see
‘example07_multisession_segmentation.m’). This function computes the mean value
1 XN
zmean ðk; t Þ ¼ pffiffiffiffi z ðk; tÞ:
i¼1 i
N

pffiffiffiffi
The division by N keeps the zmean normalized with standard deviation equal to 1 in the baseline as
in Step 47(B).
54 Define a significantly activated region by studying the correlation with the stimulus as in Step 47
(A) by:
generating the stimulus pattern stim(t) as the correlation of the stimulation window w(t) by the
hemodynamic response hrf(t)
sðt Þ ¼ wðt Þ  hrf ðtÞ;

NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot 3565

 PROTOCOL NATURE PROTOCOLS

subtracting the mean

1 Xn
sðt Þ ¼ sðt Þ  sðtÞ;
n t¼1

computing the correlation in each region k and each session i,

Pn
zi ðk; tÞsðtÞ
ri ðkÞ ¼ Pn t¼12 ðk; tÞ
Pn 2 ;
z
t¼1 i t¼1 s ðtÞ

computing the mean value, the standard deviation and the t-score of ri(k),
1 XN
rmean ðkÞ ¼ r ðkÞ;
i¼1 i
N
1 XN
σ 2 ðkÞ ¼ ðr ðkÞ  rmean ðkÞÞ2 ;
i¼1 i
N

rmean ðkÞ
tscore ðkÞ ¼ ;
σ ðkÞ

and computing the P value with a Student’s distribution

pðkÞ ¼ 1  tcdf ðtscore ðkÞ; N Þ;
where tcdf is the cumulative distribution function of the Student’s t distribution.
55 Display the averaged region–time matrix zmean(k,t) with the significantly activated regions with p(k)
< 0.001.

Troubleshooting
Troubleshooting advice can be found in Table 1.

Table 1 | Troubleshooting table

Step Problem Possible reason Solution

13 Skull bleeding A blood vessel was hit in the skull Rinse with sterile saline and soak up blood with foam; in most cases,
bleeding will stop. Apply bone wax if bleeding persists (see
‘Reagents’)
14 Drill burr breaches skull Pressure applied was too strong If brain tissue is not damaged and no bleeding arises from under the
skull, keep drilling while reducing the strength. If brain tissue is
damaged, terminate the surgery. Refine your motor skills for future
surgeries. See Fig. 2i for an example of the impact of cortical
damages on the µDoppler image quality
17 Sinus bleeding Tweezer or skull bone breached the If the sinus breach is small, intensively pour saline and absorb the
sinus. The dura covering the sinus was blood arising from the sinus with sterile gel foam. Maintain a light
not correctly dissociated from the skull pressure on the slit for 5 min to contain the bleeding. If the bleeding
and was disrupted while removing the stops, continue the procedure carefully, bearing in mind there is a
skull plaque high risk of reopening. Inject the appropriate volume of glucose
solution (see ‘Reagent setup’) to balance the loss of blood. If the
breach is too large, the bleeding cannot be stopped or the sinus is too
damaged to be saved, terminate the surgery. Refine your motor skills
for future surgeries
26 Removal of the headpost Inadequate headpost fixation to the The headpost has been removed while the mouse was in its home
mouse skull bone cage: no solution; ensure the headpost is more securely fixed in
future surgeries. Terminate the experiment
The headpost has been removed during fixation, and the brain has
been damaged: no solution; ensure the headpost is more securely
fixed in future surgeries. Terminate the experiment
Table continued

3566 NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot

 NATURE PROTOCOLS PROTOCOL
Table 1 (continued)
Step Problem Possible reason Solution

The headpost has been removed during fixation, but no brain

damage/bleeding occurred: anesthetize (Steps 1–4), and implant a
new headpost more securely (Steps 9–11 and 18–25)
32 Signal artifacts Narrow cranial window No solution. Enlarge the cranial window in future surgeries
(examples of results Cortical bulge No solution. Refine your motor skills, and be more delicate at
seen as a consequence of dissociating skull plaques from the dura. Terminate the experiment
the following reasons are
shown in Fig. 2i) Cortical damage No solution. Refine your motor skills, and be more delicate at
dissociating skull plaques from the dura. Terminate the experiment
Air bubble trapped in the ultrasound gel Remove the ultrasound gel carefully, and reapply, avoiding formation
of air bubbles. Air bubbles can be removed from the gel by prior gel
centrifugation
Signal saturation from skull Adjust gain, and/or smoothly move the probe away from this plane
34 Motor moves away User forgot to hit ‘Set Origin’ in the Immediately hit the ‘STOP’ button in the Motor settings of the GUI.
from limits ‘Motor settings’ section of the GUI Manually reposition the motor over the window, and hit ‘Set Origin’
35 Motion artifact during Loose fixation of the headpost and/or Control and reinforce the headpost and/or mouse holder fixation
imaging session the mouse holder
Light headpost-to-skull fixation Control the headpost sealing, and reinforce the fixation to the skull
following Steps 9–11 of the Procedure
Excessive movement of the mouse in Reduce mouse stress and head-fixation habituation by reinforcing
the tube training stages. Follow recommendations provided in Box 1
41 Anterior planes are Scan has been performed from posterior For the acquired dataset, flip the antero-posterior order in the both
located at posterior to anterior position ScanAnat and ScanFus files. For future experiments, scan the brain
position from anterior to posterior position
43 Mistransformation of the Human error; the volume scan was not Repeat the transformation process, following tips and tricks shown in
dataset transformed appropriately Fig. 4a,b
44 Frame rejection >10% Loose fixation of the headpost and/or Control and reinforce the headpost and/or mouse holder fixation
the mouse holder
Light headpost-to-skull fixation Control the headpost sealing, and reinforce the fixation to the skull
following Steps 9–11 of the Procedure
Excessive movement of the mouse in Reduce mouse stress and head-fixation habituation by reinforcing
the tube training stages. Follow recommendations provided in Box 1

Timing
New fUS users should allow extra time to set up and troubleshoot their system installation, and thus
ensure a fully working system before beginning experiments. The surgical time will decrease sub-
stantially with practice.
Steps 1–25, animal preparation: ~3–4 h
Steps 26–38, imaging session: ~45 min
Steps 39–50, analysis of a single session: ~1 h
Steps 51–55, analysis of multiple sessions: ~1 h
Box 1A, presurgery mouse handling: ~20 min
Box 1B, postsurgery head-fixation habituation: ~1 h

Anticipated results
This protocol describes how to insert cranial windows for fUS that are stable for several months
(Steps 1–25). A brain-wide view of stimulus-evoked activity in an awake head-fixed mouse can
subsequently be obtained in a short ~15 min imaging session. Outputs are standardized as brain-wide
activity maps and activity traces in ~250 brain regions: these are suitable for statistical comparison
over multiple sessions, and thus significance at the subject level can be inferred using the analysis
software.
As an example, we imaged the brain network activated by a short visual stimulus of 1 s. The visual
stimulus consisted of square checkerboards (10° spatial frequency) flickering at 5 Hz presented on a
large screen directly in front of the mouse. We repeated six functional scans per imaging session
(~2 min per scan), and we accumulated eight imaging sessions from the same mouse. By processing a

NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot 3567

 PROTOCOL NATURE PROTOCOLS

single session (Steps 39–50), the six functional scans acquired in that session (in ~15 min) were
averaged, and the resulting activity map (Step 47; Fig. 6b, left) showed some significantly activated
voxels. Single-session images are variable owing to the animal’s ongoing behavior, and results from
individual sessions must therefore be interpreted with caution. As an example, in the results averaged
for Fig. 6b, we observed an asymmetry in the visual response although the stimulus was presented to
both eyes.
An important advantage of this protocol is that data from different sessions can be combined, as
they are registered and analyzed in a standardized way (Steps 51–55). Two types of output are
extracted for each session: one activity map and one region–time matrix (Figs. 4 and 6). The
Supplementary Software can be used to combine the activity maps and region–time matrices of
multiple sessions to obtain one average map/time-region matrix per animal (Steps 51–55, Fig. 6).
On the activity map, by averaging the eight sessions from the same mouse (Fig. 6b, right) and
choosing a threshold of P < 0.05 uncorrected, we observed a significant activation in the superior
colliculus (SC), visual cortex (VIS), dorsal lateral geniculate complex (LGd) and other thalamic
regions that are known regions of the mouse visual system. We also observed that the response
became symmetric in the VIS and that fewer false-positive voxels are visible outside the brain. The
software enables these whole-brain maps to be displayed in 3D by selecting different views, super-
imposed on the atlas to facilitate localizing the active voxels (Fig. 6c).
The region–time analysis gives a global view of the activity with a dynamic aspect (Fig. 6d). In the
single session (Steps 49–50, Fig. 6d left), we visualized responses (i.e., high z-scores) after the pre-
sentation of the stimulus, for example in the SC and VIS. When the eight sessions were averaged
(Steps 53–55, Fig. 6d right), we confirmed that these regions are significantly activated (P < 0.001
uncorrected), and we observed that more regions show significant activity than is visible in an
individual session. The region–time analysis brings information about the dynamics of the response
in different regions. In particular, we can see that activity peaks later in the primary visual cortex
(VISp) compared with the SC and LGd, which is consistent with the fact that the SC and LGd receive
direct retinal inputs whereas the VISp gets visual inputs indirectly. This qualitative observation is
confirmed when inspecting the activity curves in these specific regions (Fig. 6e, VISp response peaks
~1 s later than in the SC and LGd).
The activity maps and the region–time representations provide complementary interpretation of
the fUS data. While the activity map displays full spatial information, it is based on a priori
knowledge of the hemodynamic response function, and ignores temporal differences between dif-
ferent regions. On the other hand, the region–time view keeps the temporal dynamics at the cost of
losing some spatial information. For example, a small part of a region can be significantly activated
and spatially detected but diluted and thus underestimated when averaged with inactivated voxels
from the same region. The fact that regions have different sizes, unlike the voxel-based analysis,
should also raise caution when interpreting the results.
As for other imaging modalities, individual sessions may give variable results. This variability can
be largely explained by the spontaneous ongoing behavior and/or brain state of the mice. Indeed, it
was observed using cortex-wide imaging methods that spontaneous behavior explains most of the
variance in neuronal activity even in the presence of a stimulus or task50,51. We propose here a simple
analysis method, but more advanced regression methods, which take into account the behavior
variables, could be implemented by the users to refine the activity maps.
We believe that this comprehensive protocol will support implementation of the fUS technology to
the neuroscience community. Getting access to brain-wide activity during behavior will hopefully
contribute to better understanding brain functions at the network level.

Data availability
The example dataset is available on Zenodo45 and can be downloaded at https://doi.org/10.5281/
zenodo.4382638.

Code availability
All the codes used for data acquisition and for data analysis referenced in this protocol are provided
as Supplementary Software. A copy is also available on a GitHub/Zenodo repository, along with
potential updates44 and can be downloaded at https://github.com/nerf-common/whole-brain-fUS.

3568 NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot

 NATURE PROTOCOLS PROTOCOL
References
1. Luo, L., Callaway, E. M. & Svoboda, K. Genetic dissection of neural circuits. Neuron 57, 634–660 (2008).
2. Lerner, T. N., Ye, L. & Deisseroth, K. Communication in neural circuits: tools, opportunities, and challenges.
Cell 164, 1136–1150 (2016).
3. Hong, G. & Lieber, C. M. Novel electrode technologies for neural recordings. Nat. Rev. Neurosci. 20,
330–345 (2019).
4. Mace, E. et al. Functional ultrasound imaging of the brain. Nat. Methods 8, 662–664 (2011).
5. Mace, E. et al. Functional ultrasound imaging of the brain: theory and basic principles. IEEE Trans. Ultrason.
Ferroelectr. Freq. Control 60, 492–506 (2013).
6. Macé, É. et al. Whole-brain functional ultrasound imaging reveals brain modules for visuomotor integration.
Neuron 100, 1241–1251.e7 (2018).
7. Urban, A. et al. Chronic assessment of cerebral hemodynamics during rat forepaw electrical stimulation
using functional ultrasound imaging. Neuroimage 101, 138–149 (2014).
8. Urban, A. et al. Real-time imaging of brain activity in freely moving rats using functional ultrasound. Nat.
Methods 12, 873–878 (2015).
9. Brunner, C. et al. Evidence from functional ultrasound imaging of enhanced contralesional
microvascular response to somatosensory stimulation in acute middle cerebral artery occlusion/
reperfusion in rats: a marker of ultra-early network reorganization? J. Cereb. Blood Flow Metab. 38,
1690–1700 (2018).
10. Brunner, C. et al. Mapping the dynamics of brain perfusion using functional ultrasound in a rat model of
transient middle cerebral artery occlusion. J. Cereb. Blood Flow Metab. 37, 263–276 (2017).
11. Rau, R. et al. 3D functional ultrasound imaging of pigeons. Neuroimage 183, 469–477 (2018).
12. Demené, C. et al. Multi-parametric functional ultrasound imaging of cerebral hemodynamics in a cardio-
pulmonary resuscitation model. Sci. Rep. https://doi.org/10.1038/s41598-018-34307-9 (2018).
13. Bimbard, C. et al. Multi-scale mapping along the auditory hierarchy using high-resolution functional
ultrasound in the awake ferret. eLife https://doi.org/10.7554/eLife.35028 (2018).
14. Dizeux, A. et al. Functional ultrasound imaging of the brain reveals propagation of task-related brain activity
in behaving primates. Nat. Commun. https://doi.org/10.1038/s41467-019-09349-w (2019).
15. Blaize, K. et al. Functional ultrasound imaging of deep visual cortex in awake nonhuman primates. Proc. Natl
Acad. Sci. USA https://doi.org/10.1073/pnas.1916787117 (2020).
16. Imbault, M., Chauvet, D., Gennisson, J.-L., Capelle, L. & Tanter, M. Intraoperative functional ultrasound
imaging of human brain activity. Sci. Rep. 7, 7304 (2017).
17. Soloukey, S. et al. Functional ultrasound (fUS) during awake brain surgery: the clinical potential
of intra-operative functional and vascular brain mapping. Front. Neurosci. https://doi.org/10.3389/fnins.2019.
01384 (2020).
18. Demene, C. et al. Functional ultrasound imaging of brain activity in human newborns. Sci. Transl. Med. 9,
eaah6756 (2017).
19. Sieu, L.-A. et al. EEG and functional ultrasound imaging in mobile rats. Nat. Methods 12, 831–834 (2015).
20. Rabut, C. et al. 4D functional ultrasound imaging of whole-brain activity in rodents. Nat. Methods 16,
994–997 (2019).
21. Brunner, C. et al. A platform for brain-wide volumetric functional ultrasound imaging and analysis of circuit
dynamics in awake mice. Neuron https://doi.org/10.1016/j.neuron.2020.09.020 (2020).
22. Rahal, L. et al. Ultrafast ultrasound imaging pattern analysis reveals distinctive dynamic brain states and
potent sub-network alterations in arthritic animals. Sci. Rep. 10, 10485 (2020).
23. Sans-Dublanc, A. et al. Optogenetic fUSI for brain-wide mapping of neural activity mediating collicular-
dependent behaviors. Neuron https://doi.org/10.1016/j.neuron.2021.04.008 (2021).
24. Heo, C. et al. A soft, transparent, freely accessible cranial window for chronic imaging and electrophysiology.
Sci. Rep. https://doi.org/10.1038/srep27818 (2016).
25. Ghanbari, L. et al. Cortex-wide neural interfacing via transparent polymer skulls. Nat. Commun. https://doi.
org/10.1038/s41467-019-09488-0 (2019).
26. Rynes, M. L. et al. Assembly and operation of an open-source, computer numerical controlled (CNC) robot
for performing cranial microsurgical procedures. Nat. Protoc. 15, 1992–2023 (2020).
27. Kılıç, K. et al. Chronic imaging of mouse brain: from optical systems to functional ultrasound. Curr. Protoc.
Neurosci. https://doi.org/10.1002/cpns.98 (2020).
28. Boido, D. et al. Mesoscopic and microscopic imaging of sensory responses in the same animal. Nat. Commun.
https://doi.org/10.1038/s41467-019-09082-4 (2019).
29. Aydin, A. K. et al. Transfer functions linking neural calcium to single voxel functional ultrasound signal. Nat.
Commun. https://doi.org/10.1038/s41467-020-16774-9 (2020).
30. Fonseca, M. S., Bergomi, M. G., Mainen, Z. F. & Shemesh, N. Functional MRI of large scale activity in
behaving mice. Preprint at bioRxiv https://doi.org/10.1101/2020.04.16.044941 (2020).
31. Dinh, T. N. A., Jung, W. B., Shim, H.-J. & Kim, S.-G. Characteristics of fMRI responses to visual stimulation
in anesthetized vs. awake mice. Neuroimage https://doi.org/10.1016/j.neuroimage.2020.117542 (2021).
32. Weisenburger, S. & Vaziri, A. A guide to emerging technologies for large-scale and whole-brain optical
imaging of neuronal activity. Annu. Rev. Neurosci. 41, 431–452 (2018).

NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot 3569

 PROTOCOL NATURE PROTOCOLS

33. Jun, J. J. et al. Fully integrated silicon probes for high-density recording of neural activity. Nature 551,
232–236 (2017).
34. Steinmetz, N. A., Zatka-Haas, P., Carandini, M. & Harris, K. D. Distributed coding of choice, action and
engagement across the mouse brain. Nature 576, 266–273 (2019).
35. Sych, Y., Chernysheva, M., Sumanovski, L. T. & Helmchen, F. High-density multi-fiber photometry for
studying large-scale brain circuit dynamics. Nat. Methods 16, 553–560 (2019).
36. Lake, E. M. R. et al. Simultaneous cortex-wide fluorescence Ca2+ imaging and whole-brain fMRI.
Nat. Methods 17, 1262–1271 (2020).
37. Gottschalk, S. et al. Rapid volumetric optoacoustic imaging of neural dynamics across the mouse brain. Nat.
Biomed. Eng. 3, 392–401 (2019).
38. Demené, C. et al. Spatiotemporal clutter filtering of ultrafast ultrasound data highly increases Doppler and
fUltrasound sensitivity. IEEE Trans. Med. Imaging 34, 2271–2285 (2015).
39. Goldey, G. J. et al. Removable cranial windows for long-term imaging in awake mice. Nat. Protoc. 9,
2515–2538 (2014).
40. Hillman, E. M. C. Coupling mechanism and significance of the BOLD signal: a status report. Annu. Rev.
Neurosci. https://doi.org/10.1146/annurev-neuro-071013-014111 (2014).
41. Avants, B. B. et al. A reproducible evaluation of ANTs similarity metric performance in brain image
registration. Neuroimage https://doi.org/10.1016/j.neuroimage.2010.09.025 (2011).
42. Pallast, N. et al. Processing pipeline for atlas-based imaging data analysis of structural and functional mouse
brain MRI (AIDAmri). Front. Neuroinform. https://doi.org/10.3389/fninf.2019.00042 (2019).
43. Wang, Q. et al. The Allen Mouse Brain Common Coordinate Framework: a 3D reference atlas. Cell
https://doi.org/10.1016/j.cell.2020.04.007 (2020).
44. ClemBrunner & nerf-common. nerf-common/whole-brain-fUS: whole-brain-fUS v1.0. Zenodo https://doi.
org/10.5281/zenodo.4585348 (2021).
45. Mace, E. & Montaldo, G. Whole-brain visually evoked activity maps in head-fixed awake mice. Zenodo
https://doi.org/10.5281/ZENODO.4382638 (2020).
46. Friston, K. J. et al. Statistical parametric maps in functional imaging: a general linear approach. Hum. Brain
Mapp. 2, 189–210 (1994).
47. Gouveia, K. & Hurst, J. L. Optimising reliability of mouse performance in behavioural testing: the major role
of non-aversive handling. Sci. Rep. https://doi.org/10.1038/srep44999 (2017).
48. Hurst, J. L. & West, R. S. Taming anxiety in laboratory mice. Nat. Methods 7, 825–826 (2010).
49. Guo, Z. V. et al. Procedures for behavioral experiments in head-fixed mice. PLoS ONE 9, e88678 (2014).
50. Musall, S., Kaufman, M. T., Juavinett, A. L., Gluf, S. & Churchland, A. K. Single-trial neural dynamics are
dominated by richly varied movements. Nat. Neurosci. 22, 1677–1686 (2019).
51. Stringer, C. et al. Spontaneous behaviors drive multidimensional, brainwide activity. Science 364,
eaav7893 (2019).

Acknowledgements
We acknowledge the following grants to G.M. and A.U.: the Leducq Foundation (15CVD02), FWO (MEDI- RESCU2-AKUL/17/049,
G091719N, and 1197818N), VIB TechWatch (fUSI-MICE), internal NERF TechDev fund (3D-fUSI project). We acknowledge
the following grant to E.M.: Human Frontier Science Program Postdoctoral Fellowship (LT000769/ 2015) and the support of this work by
the Max Planck Society. We acknowledge the following grant to B.R.: Swiss National Science Foundation grants (3100330B_163457), the
National Center of Competence in Research Molecular Systems Engineering grant, European Research Council (669157, RETMUS), and
DARPA (HR0011-17- C-0038, Cortical Sight). We thank M. Krumin for assistance with the design of the mouse holder, D. Kil for
assistance with the design of the headpost and mouse holder, T. Lambert for help with the PsychoPi software, and A. Savoye and
J. Tantivit for technical help at the beginning of the project. We also thank NERF animal caretakers I. Eyckmans, F. Ooms and S. Luijten
for their help with the management of the mice.

Author contributions
E.M., B.R., G.M. and A.U. designed the project. E.M., G.M. and A.U. designed, assembled and tested the fUS system. E.M., C.B., M.G. and
G.M. performed experiments. E.M. and G.M. designed and performed the data analysis. G.M. developed the software included with the
manuscript. C.B., E.M., M.G., G.M, B.R. and A.U. wrote the manuscript.

Competing interests
A.U. is the founder and a shareholder of AUTC company commercializing neuroimaging solutions for preclinical and clinical research.

Additional information
Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41596-021-00548-8.
Correspondence and requests for materials should be addressed to G.M. or E.M.

Peer review information Nature Protocols thanks Anne Churchland, Pablo Blinder, Yves Boubenec and the other, anonymous reviewer(s)
for their contribution to the peer review of this work.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

3570 NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot

 NATURE PROTOCOLS PROTOCOL
Received: 23 September 2020; Accepted: 30 March 2021;
Published online: 4 June 2021

Related links
Key references that demonstrate the development and use of the protocol
Macé, É. et al. Neuron 100, 1241–1251.e7 (2018): https://doi.org/10.1016/j.neuron.2018.11.031
Sans-Dublanc, A. et al. Neuron (2021): https://doi.org/10.1016/j.neuron.2021.04.008

NATURE PROTOCOLS | VOL 16 | JULY 2021 | 3547–3571 | www.nature.com/nprot 3571