In Vivo Atlas of Deep Brain Structures With 3D

Reconstructions

Visit the link below to download the full version of this book:

https://medipdf.com/product/in-vivo-atlas-of-deep-brain-structures-with-3d-recon
structions/

Click Download Now

 S. Lucerna F. M. Salpietro
c. Alafaci F. Tomasello

In Vivo Atlas
of Deep Brain Structures
With 3D Reconstructions

With 54 Plates and 40 Figures

in 265 Separate Illustrations, Mostly in Color

Springer
 Dr. Sebastiano Lucerna
Prof. Dr. Francesco M. Salpietro
Prof. Concetta Alafaci
Prof. Francesco Tomasello

University of Messina
University Polyclinic "G.Martino", Department of Neurosurgery
Via Consolare Valeria 1,98122, Messina, Italy

ISBN 978-3-642-62710-1

Library of Congress Cataloging-in-Publication Data

In vivo atias of deep brain structures: with 3D reconstructions / S. Lucerna ... [et al.].
p. ; cm. Includes bibliographical references.
ISBN 978-3-642-62710-1 ISBN 978-3-642-56381-2 (eBook)
DOI 10.1007/978-3-642-56381-2
1. Brain - Anatomy - AtIases. 2. Brain - Imaging - Atlases. 3. Magnetic resonance imaging - Atlases.
1. Lucerna, S. (Sebastiano), 1957-.
[DNLM: 1. Brain - anatomy & histology - Atlases. 2. Magnetic Resonance Imaging - Atlases.
WL 171362002] RC473.B7 15 2002 611'.81'0222-dc21 2001049700

This work is subject to copyright. Ali rights are reserved, whether the whole or part of the mate-
rial is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microftlm or in any other way, and storage in data banks.
Duplication of this publication or parts thereof is permitted only under the provisions of the
German Copyright Law of September 9, 1965, in its current version, and permis sion for use must
always be obtained from Springer-Verlag. Violations are liable for prosecution under the German
Copyright Law.

http://www.springer.de
© Springer-Verlag Berlin Heideiberg 2002
Originally published by Springer-Verlag Berlin Heidelberg New York in 2002
Softcover reprint of the hardcover 1st edition 2002
The use of general descriptive names, registered names, trademarks, etc. in this publication does
not imply, even in the absence of a specific statement, that such names are exempt from the
relevant protective laws and regulations and therefore free for general use.
Product liability: The publishers cannot guarantee the accuracy of any information about
dosage and application contained in this book. In every individual case the user must check
such information by consulting the relevant literature.
Cover design: Erich Kirchner, Heidelberg
Typesetting and reproduction of the figures: AM-productions GmbH, Wiesloch

Printed on acid-free paper 211311110p 54321

 Preface

In the first half of the twentieth century, the study of neuroanatomy

was essentially based on the observations made by scientists on brain
cadavers fixed with standard techniques. These studies have produced
well-known tools such as the stereotactic atlas, which have proven to
be extremely useful and irreplaceable for neurosurgeons, neuroradi-
ologists, neurologists and neuroanatomists. In particular, the
Talairach and Schaltenbrandt atlases are considered the most presti-
gious and up-to-date work available today. The recent introduction of
neuroimaging, especially nuclear magnetic resonance, together with
the exciting and tremendous progress made in computer graphics,
has allowed us to approach neuroanatomy directly in living patients
with more accuracy and a high degree of detail.
This work, after a short introduction which explains the methodolo-
gy used, is divided into four types of sections: three types of sections
obtained from the same brain and orientated in the standard axial,
sagittal, and coronal spatial planes and one type of section of three-
dimensional pictures obtained from the computerized processing of
the previous pictures.
The organization and the life-size tables obtained by magnetic reso-
nance make this work similar to a classic stereotactic atlas, although
the authors do not claim to reach the high level of precision which
such atlases usually provide. The abbreviations used are based on
Latin nomenclature, in order to be understood and recognized world-
wide, and are supported by a system of color codes useful for the
identification of brain structures.
By using the above techniques, the present atlas aims to make a con-
tribution to the field of neuroanatomy. It emphasizes the importance
of deep brain structures, as they are of great interest to neurosurgeons
and neuroscientists, especially in light of the recent progress which
has been made in the surgical treatment of Parkinson disease and
other extrapyramidal disorders.

Messina, November 2001 The Authors

 Acknowledgements

The authors wish to express their gratitude to Professor M. Longo, Head

of the Section of Neuroradiology at the University Polyclinic of Messi-
na, for his invaluable help with the MRI documents. Special mention
should also be made of Dr. G. Ricciardi, assistant to Prof. Longo, for his
technical skill in performing MR imaging. Finally, we extend our most
sincere thanks to Springer and especially to Ms. D. Mennecke-BUhler,
Ms. I. Oppelt and Ms. A. Cerri (Springer Italia) for their kind coopera-
tion during the preparation of this book.

Messina, November 2001 The Authors

 Contents

The Reference System: The ca-cp Plane .... 1

MRI Technique .... 2
Magnified 2D Plate Reconstructions .... 5
Spatial Limits and Magnification Factors .... 5
Drawing Technique .... 5
3D Plate Reconstructions: NURBS Modeling .... 7
Example of How To Use the Atlas .... 9

Axial Sections: Plates 1-17 .... 11

Coronal Sections: Plates 18-40 .... 53

Sagittal Sections: Plates 41-54 .... 107

3D Reconstructions .... 141

Combination of All Structures .... 145
Putamen, Globus Pallidus Medialis and Lateralis,
Ansa Lenticularis .... 149
Caudate Nucleus, Putamen, Globus Pallidus Medialis
and Lateralis, Ansa Lenticularis .... 153
The Thalamic Region .... 157
The Amygdala, Fornix and Periventricular Regions .... 161

Bibliography .... 165

 The Reference System: The ca-cp Plane

Talairach (1955) and Schaltenbrandt (1977) sug-

gested that the anterior and posterior commissure
might be considered to have a constant relationship
with the deep cerebral structures and proposed
using a line between these two structures as the
basic reference line. Talairach found only a negligi-
ble difference between the intercommissural dis-
tance in 26 brains (from 23 mm to 28.5 mm) and
took the average of this distance as 25.5 mm in his
proportional atlases (Talairach 1957,1988,1993).
The reference system used in our atlas is taken as
the intercommissural plane passing through the
center of the anterior and posterior commissures
(Fig. 1). On such a plane, a system of Cartesian axes
has been defined which originates in the center of
the anterior commissure, with coordinates Fig. 1. The intercommissural plane
expressed in millimeters with a positive sign supe-
riorly (axial plane), anteriorly (coronal plane), and
laterally (sagittal plane) with respect to the origin.
In this way each structure can be identified in the
three-dimensional (3D) space from a triplet of
coordinates (axial, coronal, sagittal) on the corre-
sponding magnetic resonance imaging (MRI) slice
on a 1:0.87* scale and on the relevant magnified
plate. We have found it useful not to normalize such
a system with a proportional scale, because the
interindividual variability in the deep cerebral
structures can be effectively compensated for
through a simple proportion between the standard
intercommissural distance of the present atlas,
which is about 28 mm, and the corresponding
measurement on the patient's MRI. Such a system of
reference is also useful for the 3D reconstruction Fig. 2. Three-dimensional spatial limits
image tables. All plans constitute a kind of paral-
lelepiped (Fig. 2) 48 mm high, 66 mm long, and 76
mm wide. The anatomical structures of our atlas
are represented in this manner.

* Unfortunately, 1:1 scale could not be used for reasons of

technical production.
 MRI Technique

The study of basal ganglia and other deep brain However, these sequences also present significant
structures is one of the most remarkable examples drawbacks, such as a high sensitivity to motion
of the utility of morphological MR studies not only artifacts and reconstruction artifacts as a result of
for diagnostic purposes, but also for the prepara- processing algorithms and voxel anisotropy. In fact,
tion and accurate guidance of neurosurgery and the latter is a serious disadvantage because, due to
stereotactic radiotherapy. 3D-Fourier transform (3D-FFT), this sequence pro-
However, this is only true if high fIled MR imagers duces significant aliasing artifacts in the head-to-
with a homogeneous magnetic field and homoge- feet direction, when obtained on the axial plane.
nous gradients are used. Tl-weighted MR studies IR represents the last class of sequences which pro-
are the most suitable for morphological study of the vide high-quality images of the brain. These utilize
basal ganglia, both because of their high fidelity in additional RF pulses (inversion pulses) to increase
anatomical structure depiction and their high or modify the contrast between different structures.
intrinsic gray matter to white matter (GM/WM) In particular, they better visualize the cerebral gray
contrast. matter (cortex and deep nuclei) thanks to an inver-
The sequences most suitable for obtaining Tl-w sion time (TI) of 250-400 ms, with which the signal
images with these characteristics are conventional from white matter is minimized and the GM/WM
spin echo (SE), volumetric gradient echo (GE-3D), contrast optimized.
and inversion recovery (IR) sequences. All three The sequence which can be attained by combining
have complementary advantages and disadvan- an IR showing these characteristics with the advan-
tages, so that one or the other must be selected tages of fast SE is commonly known as FIRMS (fast
according to the type of information required. inversion recovery for myelin suppression).
Conventional SE is the least sensitive to motion arti- This IR was optimized at the end of the 1990s at the
facts and an intermediate number of sections can radiology department of the New Jersey Medical
be obtained in an average time (15 with a TR of 750 School (USA).
ms). On the other hand, they are the least efficient as The use of intermediary turbo factors (echo train
regards signal/noise ratio (SIN) and contrast/noise length = ETL 5-9) does not alter the features of this
ratio (C/N). sequence and enables optimal anatomical images to
The employment of turbo factors is significantly be achieved even in thin slices (2-3 mm), with high
limited since the number of sections per acquisition SIN and C/N ratios in short times (I min 15 s per
and image contrast are greatly reduced. acquisition).
Magnetization prepared rapid 3D-GE (MP-RAGE), A 250-ms IT gives the best SIN and C/N ratios, while
also known as volumetric spoiled gradient echo 16 kHz is the ideal bandwidth and 19-30 ms the
(3D-SPGR), are globally the most efficient of the optimal TE.
three groups of sequences. They demonstrate very It has been shown that images obtained by means of
good SIN and C/N ratios: a very high number of FIRMS optimized with the above parameters show
sections (> 120) can be acquired which are so thin better performance in terms of signal and GM/WM
(I -1.5 mm) that it is possible to obtain nearly contrast than other conventional SE and GE (2D or
isotropic voxels and, as a consequence, produce 3D) sequences.
high-quality multi-planar reconstructions. In addi- In order to reduce acquisition time to a minimum
tion, they are the least sensitive to distortion arti- and to enhance GM/WM contrast, FIRMS can be
facts, commonly present when imaging structures performed using a signal reconstruction technique
situated near air-bone parenchyma interfaces. based on the longitudinal magnetization magnitude
 MRI Technique 3

(magnitude reconstruction technique), rather than to maintain a perfect symmetry between facial
on its absolute value. The only drawback of this structures. Both the patients and the sections were
technique is cerebrospinal fluid (CSF) hyperintensi- always positioned by an expert radiographer (L.B.)
ty, which renders the images more like T2-weighted under the author's supervision.
than Tl-weighted ones. The first sequence to be acquired was a Tl-weight-
The greatest disadvantages of FIRMS sequence are: ed, rapid GE, lasting 16 s, with a wide FOV and one
section for each of the three fundamental planes
1. The notable deterioration of image quality when
(scout). The resulting images were used to position
the slice thickness is reduced to below 3 mm
a Tl SE (TR 300, TA 2 min) oriented on the sagittal
2. The presence of ghost artifacts due to pulsation
plane and parallel to the interhemispheric fissure.
of the main intracranial vessels and of CSF with-
To avoid distortion effects caused by excessive use
in cisterns and ventricles
of slice selection gradients, when shift corrections
A Magnetom Vision scanner (1.5 T, Siemens, Erlan- of the sections at an angle to the magnetic field's
gen, Germany) was used for this atlas. Studies were longitudinal axis of greater than 2° were needed, the
always performed within 3 days from the last quali- volunteer was repositioned.
ty control of the main magnetic field and gradient The sagittal image passing through the interhemi-
homogeneity. spheric fissure obtained with this sequence was
The FIRMS sequence which could be achieved with used to localize the anterior and posterior white
the above scanner had the following parameters: commissures. Only then was the study with FIRMS
field of view (FOV) 220· 165, matrix 168·256, slice sequences started, in which a series of nine axial
thickness 3 mm, interslice gap 3 mm, ETL 7, TI 250 sections perpendicular to the median sagittal scan
ms, TE 30, no. of acquisitions 3, TA 6 min - 7 min 20 and parallel to the intercommissural plane were
s, no. of sections 7-12. acquired. The anterior and posterior commissures
TR was the only parameter modified among the dif- were included in the central section.
ferent acquisitions (value between 3,500 ms for seven The second group of axial sections was oriented in
sections and 5,600 for 12 sections), in order to obtain exactly the same way as the first, but shifted 3 mm
a sufficient number of sections to cover all the struc- cranially in order that the two groups could subse-
tures of interest in the shortest possible time. quently be perfectly interleaved. In this way a slab of
To avoid any interference among contiguous sec- brain covering the region between 28.5 mm cranial-
tions, which is often responsible for disturbing arti- ly and 22.5 mm caudally to the intercommissural
facts mainly due to CSF flux, slices had a 3-mm plane was studied, so that the entire basal ganglia
thickness and were placed with an interslice gap of region was included.
3 mm (gap ratio =1). Subsequently, 12 coronal sections were positioned
Two images were acquired in each plane, offsetting perpendicularly both to previously acquired axial
the slices of the two studies by 3 mm in order to scans and to the interhemispheric fissure. The cen-
cover the whole region of interest and thus filling ter of the group was placed in order to correspond
previously unacquired gaps. with the anterior commissure, so that the region
We considered it useful to trigger all FIRMS between the more frontal part of the head of cau-
sequences by heart cycles; in order to do this a date nucleus anteriorly and the collicular plate pos-
rough ECG was obtained by applying three elec- teriorly would be covered.
trodes to the volunteer's back. Twelve more coronal sections were acquired with
This technique prolonged the acquisition time by the same orientation, but shifted 3 mm posteriorly
between 30% and 50% (on average from 6 min 30 s in order to cover structures not included in the first
to 9 min 05 s), but resulted in almost complete elim- group of images.
ination of ghost artifacts which, as already stated, Finally, nine sagittal scans which were also perpen-
are one of the major disadvantages of this sequence. dicular to the axial scans but parallel to the inter-
Each study began with a careful positioning of the hemispheric fissure were obtained. The most medi-
volunteer within the dedicated volumetric coil. Two al section was placed to run along the whole inter-
laser beams intersecting at right angles were used as hemispheric fissure and, therefore, included the
markers to place the head so that the basal ganglia anterior and posterior commissures, while the
region was as close as possible to the coil center and remaining sections covered the left hemisphere up
4 MRI Technique

to the internal part of the lobus insularis. In this scout obtained at the beginning of the whole proce-
case, the study was also completed by acquiring the dure. A mismatch greater than 2 mm in one of the
nine sections shifted 3 mm to the left. three axes invalidated the study.
The entire procedure, including head positioning, The gray scale of all the images obtained with
scout performing, and acquisition of FIRMS images FIRMS sequence was then inverted, so that a con-
on the three planes, took on average 1 h. For the trast between structures more similar to that of
study to be valid, the volunteer's head should not conventional Tl-weighted IR sequences could be
have moved more than 2 mm from its original posi- displayed.
tion. In order to verify this, the Tl-weighted rapid Images of the whole study were finally transferred
GE on the three axis was repeated and superim- to a PC in a DICOM III format, by means of a serv-
posed, by means of a digital image addition, to the ice class provider (SCP) program.
 Magnified 2D Plate Reconstructions

Spatial Limits and Magnification Factors These factors are the cause of the main problems of
the drawing technique, which are correct outlining
In the atlas, the spatial limits of 2D area reconstruc- of the limits of the given structures in MRI slices
tions are given, taking into account the extension of and the overlapping of different structures in the
the principal deep brain nuclei and their relation- same slice, the last being due to the 3-mm MRI slice
ships with the ventricular system: thickness.
Concerning the first problem, the currently avail-
1. Anteriorly in the more frontal part of the head of
able graphic software offers several utilities to mod-
caudate nucleus, about 27 mm from the anterior
ify the contrast of the images, gray scales, and mag-
commissure
nification, which, as a result of fine single pixel
2. Posteriorly tangential to the collicular plate,
manipulation, offer better visualization of the tran-
about 39 mm from the anterior commissure
sition zones between two different areas of the same
3. Superiorly in the more cranial part of the body of
image.
caudate nucleus, about 27 mm above the inter-
The second problem is very difficult to solve
commissural plane
because many structures, such the nuclei of the
4. Inferiorly tangential to the lower part of the pitu-
thalamus, nucleus accumbens, the hypothalamic
itary gland, about 21 mm below the intercommis-
nuclei, and many others, are less than 3 mm in their
sural plane
maximum dimension and show the same signal in
5. Laterally tangential to the internal part of the
MRI images. For these reasons we believe that the
lobus insularis, about 39 mm from the intercom-
2D and 3D plates should be considered as "high
missural plane
probabilistic areas and volumes". We have not
6. Medially on the midline
included in 2D and 3D reconstructions those struc-
The magnification factor is 3x for axial and coronal tures which are not clearly documented on MRI or
reconstructions and 2x for sagittal reconstructions. which show a high margin of reconstruction-
induced error.
Although it is impossible to achieve a perfect result
Drawing Technique
in terms of precision, reconstructions verified in
stereotactic atlases (Schaltebrand, Talairach, Mai)
The major advantages of an "in vivo brain" MRI
have shown only minor differences, which are
atlas mainly consist in the absence of "post
essentially due to interindividual anatomical varia-
mortem" structural changes and artifacts induced
tions.
by the techniques of prelevation, fixation, and histo-
The computerized post-MRI elaboration method of
logical sectioning. These changes may induce as a
imaging utilized in the present atlas is described as
principal "side effect" a linear shrinkage of the brain
follows:
which can be calculated from 5% (with a freezing
technique) to more than 20% (celloidin or paraffin 1. Interpolated enhancement of resolution from
embedding). On the other hand, MRI slices have a 256x256 pixels in DICOM 3 format images to
limited spatial resolution and tissue characteriza- 1900x1900 pixels with conversion to TIFF
tion and a relatively large section thickness and may (Tagged Image File Format) format, without
be affected by some distortion, particularly at the modifying contrast or brightness (Figs. 3-5)
image margin. 2. Correction of rotational errors greater than 0.2 0
6 Magnified 2D Plate Reconstructions

Fig. 3. MRI FIRMS original image Fig. 4. MRI FIRMS inverted image Fig. 5. MRI FIRMS image after
computerized elaboration

3. Image marking with the reference system and the

millimetric grid
4. Cropping of extracerebral structures (bone, mus-
cles, and skin)
5. Outlining the structures of interest with a Bezi-
er's curves drawing method (Fig. 6)
6. Image magnification and coloration according to
a color code system
7. Image master printing with a high-resolution
photographic printer

Fig. 6. Outlining the structures of interest using Bezier's [>

curves drawing method
 3D Plate Reconstructions: NURBS Modeling

Non-uniform rational B-splines, or NURBS, are The modeling procedure of our atlas follows essen-
some of the most commonly used geometric prim- tially six steps:
itives in 3D modeling software today. With
1. Creation of a 2D NURBS curve adapted to the
NURBS, free-form curves and surfaces as well as
boundary of the structure for each MRI slice
more traditional shapes, such as conics or quadrics
(Fig. 7)
can be clearly specified. NURBS were developed as
2. Positioning each curve in the correct 3D space
a method of precisely specifying curves and sur-
coordinates (Fig. 8)
faces for use by computer-controlled machines
3. Creation of a NURBS 3D surface by interpolating
capable of producing shapes in 3D. With this tech-
all bidimensional curves (Fig.8)
nique, 3D models have smoother borders due to
4. Rendering of objects by a raytrace method
the analytic tessellation and are a very different
(Fig.8) .
from polygonal models, which are generally
5. Positioning and regulation of the scene lights
faceted. These properties of NURBS make it possi-
(Fig.8)
ble to create organic volumes such as brain deep
6. Producing the frame of each scene view (Fig.9)
gray nuclei.

Fig. 7. Creation of a two-dimensional, adapted NURBS curve (Thalamus)

8 3D Plate Reconstructions: NURBS Modeling

I>
Fig. 8. Creation and rendering of a three-dimensional,
adapted NURBS surface (Thalamus)

Fig. 9. Defmitive three-dimensional image

(thalamic region)
 Example of How To Use the Atlas

Although it is possible to use the present atlas with scan orientation, in the author's experience, it is
conventionally oriented CT scans and MRI images, useful to tilt the gantry until it is parallel to a ideal
the authors emphasize the importance of making a line which joins the tuberculum sellae and the infe-
great effort to obtain ca-cp-oriented neuroradiolog- rior part of internal occipital protuberance (both
ical images, because, only in this way can optimal quite visible in the topogram).
precision be gained from the atlas. With MRI As mentioned above, the interindividual anatomical
images, this is very simple; in fact, the anterior and variations may induce a greater or lesser loss in pre-
posterior commissure are almost always visible on a cision; nevertheless, a proportion between the two
Tl-weighted median sagittal slice. In contrast, on intercommissural distances (atlas - patient) can
CT scan the orbitomeatal orientation (frequently help to minimize this problem.
used by neuroradiologists) is only rarely parallel to The following example (Fig.l0) shows how to use
ca-cp plane; moreover, the commissures are not the atlas for localizing the subthalamic nucleus of
clearly visible in CT slices. Therefore, for a better CT Luis (STN) in an axial MRI slice.

Fig. 10. Example of how to use the system

Step 1: Selection of patient's MRI slice corresponding to ca-cp-plane
Step 2: Measuring ca-cp distance (in this case was 24 mm)
Step 3: Calculating the magnification factor between patient and atlas
images utilizing the intercommissural distance
(24 : 28 = 0.86) t>

Step 4: Selection of corresponding axial atlas plates for STN and rela- 'i/
tives coordinates (Ax-3, Cor-IS, Sag 11)
Step 5: Converting coordinates in patient's coordinates
(-3 x 0.86 = 2.6; -16 x 0.86 = -13,8; 11 x 0.86 = 9.5)

Step 6: Application of the calculated coordinates on patient's image

Axial Sections
Plates 1-17