842647 APY Australasian PsychiatryWibawa et al.

Australasian
Teaching and training Psychiatry
Australasian Psychiatry

Understanding MRI in clinical 2019, Vol 27(4) 396­–403

© The Royal Australian and
New Zealand College of Psychiatrists 2019

psychiatry: perspectives from Article reuse guidelines:

sagepub.com/journals-permissions
DOI: 10.1177/1039856219842647
https://doi.org/10.1177/1039856219842647

neuroimaging psychiatry registrars journals.sagepub.com/home/apy

Pierre Wibawa   Neuroimaging Fellow & Psychiatry Registrar, Neuropsychiatry Unit, Royal Melbourne Hospital, Melbourne,
VIC, and; Honorary Clinical Fellow, Department of Psychiatry, University of Melbourne, VIC, Australia
Thomas Rego  Neuroimaging Fellow & Psychiatry Registrar, Neuropsychiatry Unit, Royal Melbourne Hospital, Parkville, VIC,
and; Honorary Clinical Fellow, Department of Psychiatry, University of Melbourne, VIC, Australia
Dennis Velakoulis  Consultant Neuropsychiatrist, Neuropsychiatry Unit, Royal Melbourne Hospital, Parkville, VIC, and;
Professor, Melbourne Neuropsychiatry Centre, University of Melbourne, VIC, Australia
Frank Gaillard   Consultant Neuroradiologist, Department of Radiology, Royal Melbourne Hospital, Parkville, VIC, and;
Associate Professor, Faculty of Medicine, Dentistry & Health Sciences, University of Melbourne, VIC, Australia

Abstract
Objective: Based on the experiences of neuroimaging psychiatry registrars, we describe several reflections on improv-
ing the understanding and integration of magnetic resonance imaging in clinical psychiatry.
Conclusion: Better integration of magnetic resonance imaging into clinical psychiatry can be highly productive
when our investments are focused on understanding the gaps in clinical knowledge and the systemic barriers involv-
ing the patient and relevant clinicians.

Keywords:  magnetic resonance imaging, neuroimaging, radiology, neuropsychiatry

S
ome of us in early medical training were once opti- practical considerations in the process of an MRI investi-
mistic about the prospects of neuroimaging for gation, from the ordering to the interpretation of its
diagnosing and predicting various psychiatric dis- findings. The article focusses on the structural MRI as
orders. New findings are ample and continually reported opposed to functional MRI, the former being the com-
in the psychiatric literature, projecting an ideal soon-to- monly available clinical investigation.
be reductionist imaging utopia and attracting our ambi-
tions. However, coupled with limited health resources,
magnetic resonance imaging (MRI) has earnt the repu- Definition and physics
tation of being a low-yield and luxurious commodity.
MRI, discovered simultaneously by Bloch and Purcell in
These obstacles have led to the heterogeneous utilisation
1946, is a medical imaging technique that relies on the
of MRI and gradually degraded the role of imaging in
magnetic properties of atoms.1,2 The abundant hydrogen
mental health.
atoms and their magnetic properties (spins) in the
We then encountered the opportunity to become a neu- human body serve as the reference points as they are
roimaging fellow at the Neuropsychiatry Unit at Royal exposed to various external magnetic configurations.
Melbourne Hospital during our psychiatry training. The
Within the MRI scanner, the main static magnet, com-
role has both clinical and academic focuses on estab-
posed of a superconductor kept cool by liquid helium,
lished and evolving neuroimaging modalities as diag-
nostic, teaching and research tools. The unique
experience has refined our approach and revived our
hope for further integrating MRI into clinical psychiatry. Corresponding author:
Pierre Wibawa, Royal Melbourne Hospital, 300 Grattan Street,
To outline ways of improving the clinical practice of MRI Parkville, VIC 3050, Australia.
in psychiatry, we will highlight the key principles and Email: pierre.wibawa@gmail.com

396
Wibawa et al.

aligns the spins either in parallel or anti-parallel (polar- have even claimed that it is preferable to include as little
ised) to the axis of magnetic field. In clinical practice, information as possible on the request so as to not ‘bias’
the static magnetic field strength is either 1.5 or 3 Tesla the radiologists!
(T), although research magnets can be as high as 9 T.
On the contrary, we believe experienced neuroradiolo-
Imaging is then obtained by exposing the tissues to a gists often undergo a process of Bayesian hypothesis
rapid sequence of radiofrequency (RF) pulses of varying testing in their assessment to arrive at a most fitting for-
frequency and phase generated by RF coils. These pulses mulation. Such process employs an implicit pre-test
excite hydrogen nuclei, increasing their intrinsic energy analysis, such as identifying salient features within the
and changing their magnetic alignment. This higher presenting complaints, while, in contrast, novice clini-
energy state then dissipates (relaxation) and molecules cians often proceed with a pan-method of systems
gradually return to baseline alignment. Emitted energy review – sacrificing time, depth and value of the assess-
during relaxation is picked up by the RF receiver coils, ment.
with spatial information encoded by the phase and fre-
When sparse information is provided on MRI requests
quency components of the previously applied RF pulses.
(e.g. ‘cognitive impairment’), radiologists are faced with
This received signal is then decoded into images by the
more questions than answers. They will be pressed to
scanner’s computer system.
adopt a rigid systems review and potentially over-
Differences in tissue properties (e.g. cerebrospinal fluid, emphasise clinically irrelevant findings or, more impor-
white and grey matter or fat) exhibit different spins in tantly, not draw attention to subtle but clinically
recovery or relaxation. These can be exploited by modi- pertinent findings. Tailoring the pre-testing information
fying the sequence of RF pulses allowing a wide variety often yields more specific answers, meaningful negative
of images (sequences) to be generated, each highlighting findings and determine the significance of incidental
different tissue contrast properties as seen in T1, T2, dif- findings; for example, changing ‘persistent psychosis’ to
fusion weighted imaging (DWI), etc. (see Table 1). We ‘Schizophrenia with a new 12-month history of treat-
refer to several articles for further readings on the funda- ment-resistance and dysexecutive predominant cogni-
mental physics principles of MRI.3–5 tive impairment; query neurodegenerative disorder (e.g.
frontotemporal dementia) or relevant differential diag-
noses’. Such information also prompts the radiologist to
Safe preparation choose the suitable imaging protocol. Therefore, we sug-
The specific preparatory consideration for mental health gest considering several salient features, when formulat-
clients includes adhering to the safety guidelines and ing the question, in Table 2 and Flowchart 1.
MRI procedures. Each hospital will have a safety check- Furthermore, some questions are more readily addressed
list regarding ferrous objects or conducting, which can with other imaging modalities. For example, computed
be attracted or excited by the magnet. Certain surgical tomography (CT), which is more accessible than MRI, is
clips, tattoos and medical devices may be contraindi- excellent for assessing acute conditions such as basal
cated or require verification from the manufacturer. In skull fracture. For example, a presentation of a 70-year-
the worst-case scenario, scanning an individual with a old who has fallen on her right-side 3 hours ago with a
non-MRI-compatible device can lead to serious injury or background 2-year history of confusion, anxiety, rigidity
death, and as such care must be taken to obtain a good and visual disturbance may elicit an urgent need for a CT
medical and occupational exposure history. to exclude a traumatic bleed. While an MRI may also
Successive missed appointments, non-compliance and address both the pressing issue of excluding a traumatic
psychomotor agitation may compromise the quality of bleed and the latter diagnostic clarification of a dementia
MRI and the interdisciplinary relation. To increase the with Lewy Bodies, relevant factors such as the patient’s
likelihood of completion, we explain the indication and medical state and accessibility of the investigation (avail-
procedure to the patient and engage the family or case- ability of MRI, etc.) should be taken into account. On the
worker. Claustrophobia, noise sensitivity and the capac- other hand, certain presentation, such as a 25-year-old
ity for staying in a fixed position can be difficult for female with a 10-day history of seizure, delirium, fever
many patients. Mental state factors should be taken into and extrapyramidal symptoms, warrants for urgent MRI,
account, and, where necessary, hypnotic or generalised as the likely differential diagnoses are not expected to be
anaesthesia may be required, as motion can impact on readily visible on CT (e.g. autoimmune encephalopathy
the image quality. Where contrast is given in the elderly, or viral encephalitis). Moreover, there are several benefits
a recent renal function test is usually essential. and limitations in choosing other brain imaging modali-
ties over MRI (see Table 3).

Formulate the question

Interpreting the images
There exists an expectation, external to the field of radi-
ology, that MRI is a clear, definitive test with a concrete There are as many approaches to reviewing an MRI as
objective and unambiguous findings. Some clinicians there are radiologists; however, having a structured

397
 Australasian Psychiatry 27(4)

Table 1.  Clinical identification, uses and principles of MRI sequences

MRI Sequences Uses and relevance Basic principles
T1 Identification: white matter appears white, and grey After RF pulse, images are attained at specific time as
matter is grey. A variation with 3D acquisition may the proton spins (in hydrogen atoms) recover to align
be performed for better anatomical visualisation (e.g. with the main static magnetic field at different times,
MPRAGE). depending on the tissue properties (e.g. fat, which
Assess: brain structure and volume in all planes (e.g. realigns early, appears bright).
atrophy, congenital anomaly, neoplasm).
T2 Identification: CSF appears white and white matter is After RF pulse, rate of spins’ decay is detected at a
darker than grey matter. perpendicular axis to static magnetic field. Rate of
Assess: pathological changes or lesion (e.g. this decay (T2 relaxation time) differs with tissue
inflammatory or infectious hyperintensities in white properties (e.g. CSF and fat appear bright, white
matter or basal ganglia) matter appears dark)
FLAIR Identification: CSF appears dark and the rest appears RF pulse applies energy to specifically change the
like T2. spin 180° away (inversion) from main static magnetic
Assess: periventricular pathological changes, which alignment. As protons in different tissues recovered
may be obscured by CSF on T2 (e.g. multiple sclerosis at different rate to main alignment, from negative to
lesions) positive net magnetism, an image is captured when
CSF passes zero (shown as dark).
DWI Identification: relatively grainy image with only brain Diffusion of extracellular water/proton is restricted by
parenchyma readily seen as dark grey, and abnormal cell membranes. In acute ischaemia, oedema further
areas as bright. restricts freely diffusing water.
Assess: Acute tissue hyperintensity in conjunction with
ADC, which may be either acute ischaemia (which may
or may not be seen on T2) or old ischaemia (T2 shine-
through).
ADC maps Identification: appearance comparable to very low- Mathematical (coefficient) based maps based on
resolution T2. correlating DWI of various diffusion sensitivities. The
Assess: diffusion values of tissues without the effect of map displays an inverted DWI grayscale without T2
T2 (e.g. hypodense signal in acute ischaemia vs. high/ shine-through effect.
normal signal in T2 ‘shine-through’). Also useful in
assessing tumour cellularity.
SWI or T2*GRE Identification: dark substantia nigra and vessels. While T2 adjusts for the field inhomogeneities,
Assess: hypointensities (usually circular) representing T2*GRE or SWI detects ‘local’ magnetic
either calcium, heavy metal or blood products inhomogeneities according to the magnetic properties
‘microbleed’ (e.g. on basal ganglia in hypertension of the compound (e.g. paramagnetic compound such
or proximal to grey-white matter junction in Cerebral as iron in blood, diamagnetic compound such as
Amyloid Angiopathy). calcium in calcifications).
Flow-sensitive Identification: varies; generally flowing structure appear Static tissue is suppressed (saturated) and compared
imaging: MRA, bright with reduced stationary background tissues. with fresh unsuppressed blood, which appears bright
MRV and CSF flow Assess: structure of vessel (e.g. aneurysm, occlusion, or in signal.
studies dissection) and CSF flow.
Gadolinium Identification: T1 sequences are labelled as pre- or Contrast entering the disrupted blood-brain barrier
contrast post-contrast, showing bright blood vessels and nasal alters the proton environment and shortens T1
mucosa. recovery, which yields bright T1 signal.
Assess: enhancement indicating disruption of blood-
brain barrier (e.g. inflammation, tumour, or vasculitis).
DTI or tractography Identification: varies, including coloured images of white Voxel-analysis of directionality and magnitude of
matter tracts according to their directions. random motion of water molecules, which are being
Assess: passage and size of white matter tracts visually. contained and self-diffuse more readily within the
Values of directionality (fractional anisotropy) and white matter tract’s direction.
average displacement of water molecule from diffusion
(mean diffusivity). These may complement pre-surgical
planning.

(continued)

398
Wibawa et al.

Table 1.  (continued)

MRI Sequences Uses and relevance Basic principles

MRS Identifications: sequences with allocated spectra graph. Various types of metabolite in different tissues have
Assess: chemical composition (e.g. N-acetylaspartate different electron distribution and results in variation
and choline for neuronal integrity and cellular turnover of resonant frequencies.
respectively) of cancer or other lesions, which may
correlate with histological diagnosis and treatment
response.
Perfusion, for Identification: varies, standardised colour images may Incoming flowing arterial blood serves as an intrinsic
example arterial reflect degree of perfusion magnetic tracer. Images are compared between with
spin labelling Assess: perfusion changes may support other MR and without the tracer, which may show varying
findings in ischaemia, neoplasm and neurodegenerative distribution of perfusion according to the underlying
disease pathology.
Functional MRI, for Identification: varies, according to methodology Slight paramagnetic properties of deoxyhaemoglobin
example blood- Assess: transient regional oxygenation changes within are detected as local inhomogeneities (compared with
oxygenation-level haemoglobin following a specific task or activity (e.g. isomagnetic properties of oxygenated haemoglobins)
dependent demarcating motor cortex in pre-neurosurgical planning) during T2*GRE relaxation.

ADC: apparent diffusion coefficient; CSF: cerebrospinal fluid; DTI: diffusion tension imaging; DWI: diffusion weighted imaging; FLAIR:
fluid-attenuated inversion recovery; GRE: gradient-echo; MPRAGE: magnetisation-prepared 180° radiofrequency pulses and rapid
gradient-echo; MRA: angiography; MRI: magnetic resonance imaging; MRS: MR spectroscopy; MRV: venography; RF: radiofrequency;
SWI: susceptibility weighted imaging.

Table 2.  Salient points for MRI request

Acuity and pace of onset Single, intermittent or gradual (e.g. stroke vs. dementia)
Severity Early or advanced
Certainty Possible, probable or gene-confirmed
Predominant Frontal: dysexecutive syndrome, upper motor neuron abnormalities
neurocognitive signs Parietal: apraxia, agnosia, upper motor sensory abnormalities
Occipital: non-ocular visual problem
Temporal: language and speech deficits
Hippocampus: seizures, amnesia
Subcortical or basal ganglia: extrapyramidal symptoms, working memory and cognitive processing
deficits
Brainstem: autonomic or cranial nerve abnormalities, mixed upper and lower motor neuron findings
Spine: lower motor neuron abnormalities
Cerebellum: ataxia
Relevant risk factors Cardiovascular (e.g. elevated cholesterol, hypertension, diabetes, smoking), electrolyte
abnormalities, intravenous drug use, hypoxia, alcoholism, carbon monoxide poisoning,
unexplained weight loss, brain injury
Supportive features Treatment-resistant and atypical age/pattern/progression of onset
Differential diagnoses Alzheimer’s disease vs. frontotemporal dementia
Purpose Diagnostic, prognosis or treatment

approach can be helpful particularly to the less experi- sequence in sagittal, coronal and axial positions. Once
enced readers. Our structured neuropsychiatric assess- oriented with the structure, we screen for any signal
ment of an MRI study starts with reviewing the brain intensity changes on T2 or Fluid-Attenuated Inversion
and extracranial anatomy and volume on the T1 Recovery (FLAIR) that may classify as ‘lesions’ or patho-

399
 Australasian Psychiatry 27(4)

Figure 1.  Examples of MRI sequences in neuropsychiatric presentations.

(A) Alzheimer’s disease presenting with apraxia correlating with precuneus atrophy on T1. (B) Absence-type epilepsy with
persistent amnesia correlating with left hippocampal gliosis and temporal atrophy on T1. (C) Progressive non-fluent aphasia
showing enhancing left frontal meningioma on gadolinium-contrast T1 sequence. (D) Progressive dysexecutive behaviour showing
bifrontal leucodystrophy involving subcortical U-fibres. (E) A vasculitis presenting with stroke-like episodes of right-sided weakness
and apraxia correlating with left parietal vessel enhancing lesion on gadolinium-contrast T1 sequence. (F) Depression, lower
limb Parkinsonism and catatonia showing ‘microbleed’ on SWI in the basal ganglia and peripheral cortices likely attributed to
hypertension and cerebral amyloid angiopathy, respectively, as well as previous haemorrhagic stroke requiring a ventricular shunt on
right. A stroke presenting with aphasia and right facial droop involving left middle cerebral artery territory, showing hyperintensities
on FLAIR (G) and an acute diffusion restriction (ischaemia) on DWI (H) with corresponding dark (low values) ADC (I) over the
subcortical white matter of Broca's area, insula and Wernicke's area. Chronic small vessel ischaemic change is also seen in the
white matter, adjacent to the lateral ventricle, on FLAIR (G) without acute diffusion changes. MRI images in this figure are created
and owned by the authors.

logical changes. Any localised lesions may be then be with apparent diffusion coefficient (ADC) (see Table 1
further characterised by viewing additional sequences, and Figure 1). We sometimes utilise several online inter-
such as susceptibility weighted imaging (SWI) and DWI active step-by-step modules and normal MRI anatomical

400
Wibawa et al.

Table 3.  Benefits and limitations of MRI in comparison with CT and nuclear functional imaging

Imaging modality Benefits Limitation

CT •  D etection of bone, calcifications and •  Ionising radiation, potentially carcinogenic
acute haemorrhage •  Limited soft-tissue definition (e.g. grey-white matter
•  Vasculature (contrast) differentiation)
•  Widely available •  Bone or metal artefact obscuring brain images (e.g.
•  Quick (<5 min) posterior fossa)
•  Comparatively non-claustrophobic •  Contrast (allergy and nephrotoxicity)
MRI •  Exquisite soft-tissue differentiation •  Noise
(e.g. demyelination) •  More confined space for patient
•  No ionising radiation (e.g. pregnancy) •  Contraindications (e.g. some metallic/conductive
objects)
•  More expensive than CT
•  Long (20–45min)
Nuclear medical •  Functional changes may precede •  Limited spatial resolution
imaging: for example structural changes on CT or MRI •  Sensitivities vary with tracers and disease of interest
PET or SPECT •  Use of radioactive tracer
•  Availability and short half-life of some tracers
•  Cost (PET more than SPECT)

CT: computed tomography; MRI: magnetic resonance imaging; PET: positron-emission tomography; SPECT: single-photon
emission computed tomography.

comparisons as point of reference.6–9 Further systematic ment. For studying reference, free web-based and mobile
approaches offered in literature include anatomical,10,11 app resources in MRI neuroanatomy and pathology are
diagnostic,12,13 and case-based.14 widely available at your fingertips.16–18
In the context of pre-testing clinical features (e.g. tumour Many services hold regular interdisciplinary radiology
in CT or positive autoimmune marker in plasma), spe- meetings to address the challenging clinical conun-
cific addition of spectroscopy, flow-sensitive sequences drums. The presence of multiple experts and additional
for arterial, venous or Cerebrospinal fluid (CSF) flow, clinical history may add further clinical consensus for
perfusion (e.g. arterial spin labelling) and gadolinium- further investigation, management and differential diag-
based contrast may be considered for further assessment noses. For example, in early dementia cases when MRI
of the lesion and associated structure of interest. findings do not support the clinical findings, other suit-
Diffusion tensor imaging (DTI) may also be used for able investigations or specialist’s assessment may be sug-
evaluating the white matter tract in integrity, but this gested (e.g. nuclear medical imaging, cerebrospinal fluid
often requires statistical processing and is generally not test). Radiology meetings also serve as an alliance plat-
performed ‘routinely’.15 form for professional learning and collaboration.
Data from MRI brain studies offer a rich source of
untapped knowledge. Potential for discoveries are vast,
Making sense of it all from a numerical difference to visible ones. Various man-
While it is prudent to rely primarily on the radiologist’s ual and automated methods (e.g. FreeSurfer, Voxel-Based
report, there are several reasons for attempting to inter- Morphometry and FMRIB Software Library) are available
pret MRI as a psychiatrist. First, like all clinicians, radi- for analysing the images.19–22 However, these often require
ologists too are prone to the error of their subjectivity further training, supervision and academic collaboration.
and expertise. The appearance of a normal and patho- With the increasing accessibility of far-reaching technol-
logical brain overlaps, particularly in the elderly, and ogy such as machine learning, research-oriented centres
attracts greater variability in interpretations. Second, often integrate an anonymised MRI archive under appro-
assessing MRI improves our future pre-testing acumen priate ethical and governance approval. Sharing such data
and professional skills. Clinicians are privileged with the through international research studies and a free-content,
opportunities to contextualise their theoretical knowl- collaborative web-based encyclopaedias (e.g. https://radi-
edge, phenomenological observation and anatomical opaedia.org/) are some of the contemporary approaches
findings; to complement an in-depth bedside assess- to research and learning.

401
 Australasian Psychiatry 27(4)

Flowchart 1.  Clinical consideration of MRI in psychiatry.

Conclusion Funding
The authors received no financial support for the research, authorship, and/or publication of
Understanding MRI sits beyond the concrete interpreta-
this article.
tion of the black and white picture. Contrary to the
common sentiments in psychiatry, we found that
improving our knowledge and the interdisciplinary rela- ORCID iDs
tionship yields better clinical resolution beyond balanc- Pierre Wibawa https://orcid.org/0000-0002-7120-6984
ing the costs and availability. Frank Gaillard https://orcid.org/0000-0002-2382-872X

Disclosure
Frank Gaillard is CEO of Radiopaedia.org. The authors report no other conflict of interest. The References
authors alone are responsible for the content and writing of the paper. 1. Bloch F. Nuclear induction. Phys Rev 1946; 70: 460–474.

402
Wibawa et al.

2. Purcell EM, Torrey HC and Pound RV. Resonance absorption by nuclear magnetic 13. Hoffmann Nunes R, Abello AL and Castillo M. Critical findings in neuroradiology. Cham,
moments in a solid. Phys Rev 1946; 69: 37–38. Switzerland: Springer, 2016.

3. Grover VP, Tognarelli JM, Crossey MM, et al. Magnetic resonance imaging: principles 14. Rana AQ, Zumo LA and Sim V. Neuroradiology in clinical practice. Cham, Switzerland:
and techniques: lessons for clinicians. J Clin Exp Hepatol 2015; 5: 246–255. Springer, 2013.
4. Pooley RA. Fundamental physics of MR imaging. RadioGraphics 2005; 25: 1087–1099. 15. Jellison BJ, Field AS, Medow J, et al. Diffusion tensor imaging of cerebral white matter:
5. Farrall AJ. Magnetic resonance imaging. Practical Neurol 2006; 6: 318–325. a pictorial review of physics, fiber tract anatomy, and tumor imaging patterns. Am J
Neuroradiol 2004; 25: 356–369.
6. Renowden S. Normal anatomy of the brain on CT and MRI with a few normal variants.
Pract Neurol 2012; 12: 225–233. 16. Radiopaedia.org. The online collaborative Radiology resource, https://radiopaedia.
7. Renowden SA. Normal vascular imaging. Pract Neurol 2013; 13: 31–38. org/?lang=gb (2019, accessed 20 January 2019).

8. Luijkx T, Gaillard F, et al. Neurodegenerative MRI brain (an approach). Radiopaedia 28, 17. Brain Tutor 3D, https://www.brainvoyager.com/Mobile/BrainTutor3D_iOS.html (2018,
https://radiopaedia.org/articles/neurodegenerative-mri-brain-an-approach?lang=gb accessed 20 January 2019).
(2018, accessed 20 January 2019).
18. Stein A. ‎Brain images. Apple Store, https://itunes.apple.com/au/app/brain-images/
9. Hacking C, Gaillard F, et al. Normal brain imaging examples. Radiopaedia 47, https:// id427063170?mt=8 (2019, accessed 20 January 2019).
radiopaedia.org/articles/normal-brain-imaging-examples-1 (2018, accessed 20 January
2019). 19. Reuter M, Schmansky NJ, Rosas HD, et al. Within-subject template estimation for unbi-
ased longitudinal image analysis. Neuroimage 2012; 61: 1402–1418.
10. Small JE and Schaefer PW. Neuroradiology: key differential diagnoses and clinical ques-
tions. Philadelphia, PA: Saunders/Elsevier, 2013. 20. Ashburner J and Friston KJ. Voxel-based morphometry – the methods. Neuroimage
11. Nadgir R, Yousem DM and Yousem DM. Neuroradiology: the requisites. Philadelphia, 2000; 11: 805–821.
PA: Elsevier, 2016. 21. Jenkinson M, Beckmann CF, Behrens TE, et al. Neuroimage 2012; 62: 782–790.
12. Small JE, Noujaim DL, Ginat DT, et al. Neuroradiology: spectrum and evolution of dis- 22. Lerch JP, van der Kouwe AJW, Raznahan A, et al. Studying neuroanatomy using MRI. Nat
ease. Philadelphia, PA: Elsevier, 2019. Neurosci 2017; 20: 314.

403