**3. Three-dimensional electronic atlases based on printed two-dimensional stereotactic atlases**

Printed stereotactic atlases have an intrinsic limitation due to the fact that they consist of a two-dimensional representation of the brain, which is a three-dimensional structure (Yelnik et al., 2009).The first solution encountered in literature to overcome this problem was simply to scan the previously published 2D templates into a computer and delineate 3D volumes on them. This was the beginning of the development of deformation algorithms and volumetric visualization of anatomical structures that would change the standard on neurosurgical planning.

Orientation of sections and interval

Thalamus : 26 horizontal,24 sagittal Basal ganglia: 25 sagittal and coronal

*Definitions: PC-PO line: connects the center of posterior commissure (PC) to the ponto-medullary sulcus (posterior border of the pons) and the midsagittal plane (MSP); ICL- intercommissural line: line that passes through the superior edge of the anterior commissure (AC) and inferior edge of PC; ICP- intercommissural plane: plane obtained from the ICL, defines the horizontal plane; MSP-Midsagittal plane: obtained from midline; VACp:Verticofrontal plane is formed by VAC line, that is a vertical line traversing the posterior margin of the AC: VPCp is the verticofrontal plane perpendicular ICL crossing it at PC; MComP: Midcommissural plane is erected from the midcommissural point, that is the midpoint of the ICL;The FM-PC line is the distance between the posterior inferior margin of the foramen of Monro (FM) to the midpoint of the ventricular surface of PC;The total thalamic length (TthL) is the distance between the FM and the top of the pulvinar;The ventricular-floor plane (VFP) is the plane defined by the floor of the fourth ventricle. Perpendicular to this plane and reaching the* 

Table 1. Printed stereotactic atlases based on number of cases studied, histological protocols,

**3. Three-dimensional electronic atlases based on printed two-dimensional** 

Printed stereotactic atlases have an intrinsic limitation due to the fact that they consist of a two-dimensional representation of the brain, which is a three-dimensional structure (Yelnik et al., 2009).The first solution encountered in literature to overcome this problem was simply to scan the previously published 2D templates into a computer and delineate 3D volumes on them. This was the beginning of the development of deformation algorithms and volumetric visualization of anatomical structures that would change the standard on neurosurgical

Range of Slices Coordinate

14 mm superior to 8.1 mm inferior to ICL;4.6 to 25.8 mm from midline; 3 to 41.5 mm anterior to PC; 3 to 27mm lateral to midline

System MRI

Yes, from 2 brains, obtained 10 days and 1 year after fixation in formalin

ICP MSP VAC

Reference

Morel, 2007

**stereotactic atlases** 

planning.

Nr. of brains and sections

7 brains, frozen, cut 40-50µm thick

Staining and other investigative tissue methods

Nissl and cresyl violet, myelin;parvalb umin (PV), calbindinD-28K (CB), and calretinin (CR) ;antibodies anti tyrosine hydroxylase (TH), Acetylcholinest erase (AChE) or immunoreacted with SMI-32

*fastigium (apex of the roof of the fourth ventricle) is the fastigium-floor line(FFL).* 

planes of section, coordinate systems, and reference to neuroimaging.

#### **3.1 Creation of a three-dimensional atlas by interpolation from Schaltenbrand-Bailey's atlas (Yoshida, 1987)**

In 1987 Yoshida proposed the creation of a 3D atlas of the human brain based on the interpolation of the 2D templates of the Schaltenbrand-Bailey's atlas. Although the rendering resulted in great three-dimensional incoherency due to the lack of correction of the linear and non-linear tissue distortions present at the original sections, this atlas is important as the first of its kind and of the concept of referring to a volume instead of confining exclusively on two-dimensional representations of the neural tissue.

#### **3.2 Multiple brain atlas database and atlas-based neuroimaging system (Nowinski et al., 1997)**

Nowinski and his collaborators started in 1997 a project in which four previously published atlases were digitized, enhanced, segmented (color coded or contoured), labeled, aligned, and organized into volumes for the purpose of developing an atlas-based neuroimaging system for analysis, quantification, and real-time manipulation of cerebral structures in two and three dimensions. A software was developed that is available in a CD-ROM called "Electronic Clinical Brain Atlas-ECBA." It is impressive in terms of three-dimensional visualization of the structures; however, due to some inaccuracies inherent in the original print atlases, three dimensional structures reconstructed from Schaltenbrand & Wahren atlas are often convoluted and displayed in unrealistic shapes. As noticed and discussed by the authors in this paper, a given point in the stereotactic space may have up to three different labels on the Talairach & Tournox atlas, due to inconsistent orthogonal plates. Nevertheless, important questions about the 3D accuracy of the most-used atlases were evidenced.

#### **3.3 Automated atlas integration and interactive three-dimensional visualization tools for planning and guidance in functional neurosurgery (St-Jean et al., 1998)**

The authors created a deformable volumetric atlas of the basal ganglia and thalamus from the Schaltenbrand and Wahren atlas (SW atlas) to help neurosurgeons navigate through MRI-invisible structures. They developed also a visualization platform that permits manipulation of the merged atlas and MRI data set in two- and three-dimensional views. A really interesting and new method of correction of errors was developed by this group. After digitizing the sections and the transparent overlays from the SW atlas, they aligned and segmented the nuclear contours based on the overlays and performed an interpolation to create a 3D volume. The alignment was based on the original grid, and they noted that the grid structure present in the atlas was placed on the cryotome images subsequent to photography, and so even precise alignment with respect to the grid is no guarantee that the underlying slices were not themselves distorted during the slicing process. However, they developed a methodology for matching automatically the slices with the 3D model. The atlas is registered point-to-point (250 homologous landmarks identified by a neuroanatomist) to a model MRI or standard reference volume. Even though any MRI could be chosen, they opted by the Colin27 that is the result of an average of 27 MRI scans of the same subject. As they have a labeled MRI based on the SW atlas, an algorithm called ANIMAL (Automated Nonlinear Image Matching and Anatomical Labeling) computes a nonlinear transformation to register the patient's MRI with the pre-labeled

Review of Printed and Electronic Stereotactic Atlases of the Human Brain 159

The article published in 2003 introduces an atlas-assisted method and a tool called the Cerefy Neuroradiology Atlas (CNA), available over the internet for neuroradiology and human brain mapping. The Talairach & Tounoux atlas is presented in digital format and can be warped to the patient's MRI scan by means of a Talairach transformation. The Talairach landmarks (AC, PC, the most lateral point of the parietotemporal cortex, the most anterior point of the frontal cortex, the most posterior point of the occipital cortex, the most superior point of the parietal cortex, and the most inferior point of the temporal cortex) are set manually or semi-automatically and a linear transformation is performed. The great achievements of this atlas are the ease of use, new atlas-user interface, and availability over

**3.4 The cerefy neuroradiology atlas: A talairach–tournoux atlas-based tool for analysis of neuroimages available over the internet (Nowinski & Belov, 2003)** 

**3.5 A deformable digital brain atlas system according to talairach and tournoux** 

**4. Three-dimensional electronic atlases based on histological data** 

**4.1 The creation of a brain atlas for image guided neurosurgery using serial** 

trying to correct previous published templates.

**histological data (Chakravarty et al, 2006)** 

The authors have developed a digital version of the Talairach &Tournoux atlas. The main goal is to assist neurosurgical planning rather than brain mapping. They present a 3D representation of most of the brain structures contained in the Talairach atlas. They have also developed a tool which has a non-rigid matching capability, allowing the standard atlas structure to be warped to an individual brain MRI, even when lesions such as tumors are present. The great contribution of this work is the development of the nonlinear algorithm used to warp the atlas to the patient's MRI, despite the need for substantial nonlinear

The digitization of previous published atlas led to problems in accuracy due to errors inherent in the technique used to construct them. In order to achieve better precision and accuracy in the 3D atlases, three groups have proposed to generate their electronic threedimensional reconstructions based on own histological sections instead of aligning and

Until 2006, the group from MNI (Montreal Neurological Institute, Canada) used the atlas developed by St-Jean et al., 1998 to program neurosurgical interventions. However, some shortcomings were recognized, including limited inherent resolution in the slice direction, limited number of structures, and some small mis-registrations between the digital atlas and the Colin27 MRI average that are propagated to patient MRI data during the atlas customization procedure. In this manuscript, the authors addressed these limitations and presented a technique for the creation of a brain atlas of the basal ganglia and thalamus derived from serial histological data. The technique used was identical to the one used in St-Jean's (1998) atlas. However, in the latter instance own histological preparations were available instead of scanned figures from the Schaltenbrand & Wahren atlas. The authors digitized coronal histological sections and delineated 105 anatomical structures in them. A slice-to-slice nonlinear registration technique to correct for spatial distortions was

the internet.

**(Ganser et al., 2004)** 

transformations.

Colin27.The most important contribution from this work was the use of a special algorithm to align and segment the images and the use as reference a MRI standard volume, the Colin27.


*At variance with Table 2 data and 3D reconstructions were generated by the authors from new series of brains and details on histological procedures and data manipulation are listed.* 

Table 2. Electronic atlases derived from previously printed 2D stereotactic atlases Authors, sources, digital manipulation, and registration are listed.

Colin27.The most important contribution from this work was the use of a special algorithm to align and segment the images and the use as reference a MRI standard

plate

A 0.5-mm step atlas was interpolated from the original atlas

The print atlases listed were digitized, enhanced, segmented , labeled, aligned, and organized into volumes

Digit.sec. and transp. overlay, aligned and extracted 2D surfaces based on

Digitized and processed the original print plates

Digitized and used only the 38 coronal plates

it and interpolated sec. Correction of distortions of the original atlas and segmentation

Manual correction of rotation and overlay-

misregistrations. Some of the sources of errors cannot not be corrected

Slice-to-slice spatial inconsistencies in structure contours were considered to be small, and thus not accounted for. ANIMAL algorithm warps atlas to Colin27

Added structures to original templates to improve 3D consistency; developed algorithms to reformat the atlas to

ICP

*At variance with Table 2 data and 3D reconstructions were generated by the authors from new series of brains and* 

Table 2. Electronic atlases derived from previously printed 2D stereotactic atlases Authors,

Interpolated additional crosssections and applied algorithms to enhance the 3D coherence

Registration and atlasto-patient normalization

Registration based on max.dimensions of the brain and head of

optic tract or putamen

landmarks.Talairach's Transformation is

ANIMAL algorithm warps the pre-labeled Colin27 to patient's

Talairach landmarks are set and then Talairach's Transformation is

caudate,

used as

applied

MRI

applied

Correspondences between the atlas and the patient are established in an automatic fashion nonrigid approach;

volume, the Colin27.

Yoshida, 1987

Nowinski et al.,1997

St-Jean et al.,1998

Nowinski and Belov, 2003

Ganser et al, 2004

References Original Atlas Method

Schaltenbrand& Bailey,1959

Talairach& Tournoux,1988,

Schaltenbrand& Wahren,1977; Ono et al.,1990

Schaltenbrand& Wahren,1977

Talairach& Tournoux, 1988

Talairach& Tournoux, 1988

*details on histological procedures and data manipulation are listed.* 

sources, digital manipulation, and registration are listed.

1993;

#### **3.4 The cerefy neuroradiology atlas: A talairach–tournoux atlas-based tool for analysis of neuroimages available over the internet (Nowinski & Belov, 2003)**

The article published in 2003 introduces an atlas-assisted method and a tool called the Cerefy Neuroradiology Atlas (CNA), available over the internet for neuroradiology and human brain mapping. The Talairach & Tounoux atlas is presented in digital format and can be warped to the patient's MRI scan by means of a Talairach transformation. The Talairach landmarks (AC, PC, the most lateral point of the parietotemporal cortex, the most anterior point of the frontal cortex, the most posterior point of the occipital cortex, the most superior point of the parietal cortex, and the most inferior point of the temporal cortex) are set manually or semi-automatically and a linear transformation is performed. The great achievements of this atlas are the ease of use, new atlas-user interface, and availability over the internet.

#### **3.5 A deformable digital brain atlas system according to talairach and tournoux (Ganser et al., 2004)**

The authors have developed a digital version of the Talairach &Tournoux atlas. The main goal is to assist neurosurgical planning rather than brain mapping. They present a 3D representation of most of the brain structures contained in the Talairach atlas. They have also developed a tool which has a non-rigid matching capability, allowing the standard atlas structure to be warped to an individual brain MRI, even when lesions such as tumors are present. The great contribution of this work is the development of the nonlinear algorithm used to warp the atlas to the patient's MRI, despite the need for substantial nonlinear transformations.
