A new rapid landmark-based regional MRI segmentation method of the brain

Similar documents
Visual Rating Scale Reference Material. Lorna Harper Dementia Research Centre University College London

Regional and Lobe Parcellation Rhesus Monkey Brain Atlas. Manual Tracing for Parcellation Template

Neuroanatomy lecture (1)

Differences in brain structure and function between the sexes has been a topic of

Medical Neuroscience Tutorial Notes

Piano playing skills in a patient with frontotemporal dementia: A longitudinal case study

ORIGINAL CONTRIBUTION. Application of Automated Medial Temporal Lobe Atrophy Scale to Alzheimer Disease

Automated detection of abnormal changes in cortical thickness: A tool to help diagnosis in neocortical focal epilepsy

Procedia - Social and Behavioral Sciences 159 ( 2014 ) WCPCG 2014

MITELMAN, SHIHABUDDIN, BRICKMAN, ET AL. basic necessities of life, including food, clothing, and shelter. Compared to patients with good-outcome schiz

Cover Page. The handle holds various files of this Leiden University dissertation

A Dozen Neuroanatomical Landmarks Every Radiologist Should Know

Visualization strategies for major white matter tracts identified by diffusion tensor imaging for intraoperative use

Brain tissue and white matter lesion volume analysis in diabetes mellitus type 2

Assessing Brain Volumes Using MorphoBox Prototype

Supplementary Online Content

Automated Whole Brain Segmentation Using FreeSurfer

Pediatric MS MRI Study Methodology

Cerebrum-Cerebral Hemispheres. Cuneyt Mirzanli Istanbul Gelisim University

Supplementary Online Content

Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative

SWI including phase and magnitude images

fmri and Voxel-based Morphometry in Detection of Early Stages of Alzheimer's Disease

Gross Organization I The Brain. Reading: BCP Chapter 7

Diffusion-Weighted and Conventional MR Imaging Findings of Neuroaxonal Dystrophy

Supplementary Information Methods Subjects The study was comprised of 84 chronic pain patients with either chronic back pain (CBP) or osteoarthritis

Announcement. Danny to schedule a time if you are interested.

CEREBRUM. Dr. Jamila EL Medany

Review of Longitudinal MRI Analysis for Brain Tumors. Elsa Angelini 17 Nov. 2006

Clinically focused workflow with unique ability to integrate fmri, DTI, fiber tracks and perfusion in a single, multi-layered 3D rendering

Neuro-Imaging in dementia: using MRI in routine work-up Prof. Philip Scheltens

Principles Arteries & Veins of the CNS LO14

PRESERVE: How intensively should we treat blood pressure in established cerebral small vessel disease? Guide to assessing MRI scans

NIH Public Access Author Manuscript Proc SPIE. Author manuscript; available in PMC 2014 February 07.

Human Brain Myelination from Birth to 4.5 Years

Four Tissue Segmentation in ADNI II

A few notions of brain anatomy

Figure S1: P maps of differences between groups in thickness of gray matter for. the subgroup of 16 Pure Tourette syndrome subjects and their age and

Fully-automated volumetric MRI with normative ranges: Translation to clinical practice

brain MRI for neuropsychiatrists: what do you need to know

mr brain volume analysis using brain assist

It is well known that the topography of the brain correlates

Voxel-based morphometry in clinical neurosciences

Online appendices are unedited and posted as supplied by the authors. SUPPLEMENTARY MATERIAL

Twelve right-handed subjects between the ages of 22 and 30 were recruited from the

APPLICATION OF PHOTOGRAMMETRY TO BRAIN ANATOMY

Overview of the Nervous System (some basic concepts) Steven McLoon Department of Neuroscience University of Minnesota

Yin-Hui Siow MD, FRCPC Director of Nuclear Medicine Southlake Regional Health Centre

Psychology, 3 Department of Anatomy, Histology and Embryology,

MR imaging is emerging as a central tool for in vivo quantification

Brain anatomy tutorial. Dr. Michal Ben-Shachar 459 Neurolinguistics

Cortico-Striatal Connections Predict Control over Speed and Accuracy in Perceptual Decision Making

SUPPLEMENTARY INFORMATION In format provided by Frank et al. (JULY 2010)

Diffusion Tensor Imaging in Psychiatry

CISC 3250 Systems Neuroscience

W hite matter lesions (WML) in elderly people result

Dissection of the Sheep Brain

Department of Cognitive Science UCSD

Morphometric Analysis of Cortical Sulci Using Parametric Ribbons: A Study of the Central Sulcus

Different regional patterns of cortical thinning in. Alzheimer s disease and frontotemporal dementia (FTD) are

Magnetic Resonance Imaging. Basics of MRI in practice. Generation of MR signal. Generation of MR signal. Spin echo imaging. Generation of MR signal

Author's response to reviews

BIOL Dissection of the Sheep and Human Brain

Automated morphometry in adolescents with OCD and controls, using MR images with incomplete brain coverage

A Dorsolateral Prefrontal Cortex Semi-Automatic Segmenter

Medical Neuroscience Tutorial Notes

Neuroimaging for Diagnosis of Psychiatric Disorders

Supplementary online data

NEURO IMAGING 2. Dr. Said Huwaijah Chairman of radiology Dep, Damascus Univercity

The Nervous system is divided into 2 major divisions: 1) Central Nervous System (CNS): found within bones & consists of:

High spatial resolution reveals excellent detail in pediatric neuro imaging

Copyright 2002 American Academy of Neurology. Volume 58(8) 23 April 2002 pp

1. The basic anatomy of the Central Nervous System (CNS)

Student Lab #: Date. Lab: Gross Anatomy of Brain Sheep Brain Dissection Organ System: Nervous Subdivision: CNS (Central Nervous System)

NIH Public Access Author Manuscript Med Image Comput Comput Assist Interv. Author manuscript; available in PMC 2013 December 06.

DISSECTION OF THE SHEEP'S BRAIN

University, Montreal, Quebec, Canada

CNS Imaging. Dr Amir Monir, MD. Lecturer of radiodiagnosis.

Differential diagnosis of Frontotemporal Dementia FTLD using visual rating scales

SOP: Cerebral Ultrasound

ANATOMY & PHYSIOLOGY DISSECTION OF THE SHEEP BRAIN LAB GROUP:

FRONTAL LOBE. Central Sulcus. Ascending ramus of the Cingulate Sulcus. Cingulate Sulcus. Lateral Sulcus

THE ESSENTIAL BRAIN INJURY GUIDE

Structural MRI in Frontotemporal Dementia: Comparisons between Hippocampal Volumetry, Tensor- Based Morphometry and Voxel-Based Morphometry

Sex Differences in Cortical Thickness Mapped in 176 Healthy Individuals between 7 and 87 Years of Age

Cerebro-vascular stroke

THE cerebral surfaces are molded into several irregular. The occipital lobe convexity sulci and gyri. Laboratory investigation

Shape Modeling of the Corpus Callosum for Neuroimaging Studies of the Brain (Part I) Dongqing Chen, Ph.D.

Visual rating and volumetry of the medial temporal lobe on magnetic resonance imaging in dementia: a comparative study

Hallucinations and conscious access to visual inputs in Parkinson s disease

SPAMALIZE s Cerebellum Segmentation routine.

Cerebral Cortex 1. Sarah Heilbronner

We are IntechOpen, the world s leading publisher of Open Access books Built by scientists, for scientists. International authors and editors

The most common cause of dementia is Alzheimer disease.

Chapter 3. Structure and Function of the Nervous System. Copyright (c) Allyn and Bacon 2004

Automated Volumetric Cardiac Ultrasound Analysis

M555 Medical Neuroscience Lab 1: Gross Anatomy of Brain, Crainal Nerves and Cerebral Blood Vessels

Assessment of Adipose Tissue from Whole Body 3T MRI Scans

Biological Bases of Behavior. 3: Structure of the Nervous System

O Connor 1. Appendix e-1

Transcription:

Journal of the Neurological Sciences 194 (2002) 35 40 www.elsevier.com/locate/jns A new rapid landmark-based regional MRI segmentation method of the brain A.L.W. Bokde a, S.J. Teipel a, *, Y. Zebuhr a, G. Leinsinger b, L. Gootjes a,c, R. Schwarz a, K. Buerger a, P. Scheltens d, H.-J. Moeller a, H. Hampel a, ** a Dementia and Neuroimaging Section, Department of Psychiatry, Ludwig-Maximilian University, Nussbaumstr. 7, 80336 Munich, Germany b Department of Radiology, Ludwig-Maximilian University, Munich, Germany c Department of Clinical Neuropsychology, Free University, Amsterdam, Netherlands d Department of Neurology, Free University Hospital, Amsterdam, Netherlands Received 2 July 2001; received in revised form 2 November 2001; accepted 6 November 2001 Abstract Background: Neurodegenerative and cerebrovascular diseases show a distinct distribution of regional atrophy and subcortical lesions. Objective: To develop an easily applicable landmark-based method for segmentation of the brain into the four cerebral lobes from MRI images. Method: The segmentation method relies on a combination of anatomical landmarks and geometrical definitions. It is applied on the surface reconstruction of the MRI volume. The internal borders between the lobes are defined on the axial slices of the brain. The reliability of this method was determined from MRI scans of 10 subjects. To illustrate the use of the method, it was applied to MRI scans of an independent group of 10 healthy elderly subjects and 10 patients with vascular dementia to determine the regional distribution of white matter hyperintensities (WMH). Results: The intra-rater relative error (and intra-class correlation coefficient) of the lobe segmentation ranged from 1.6% to 6.9% (from 0.91 to 0.99). The inter-rater relative error (and intra-class correlation coefficient) ranged from 1.4% to 5.2% (from 0.96 to 0.99). Density of WMH was significantly higher in all four lobes in VD patients compared to controls ( p < 0.05). Within each group, WMH density was significantly higher in frontal and parietal than in temporal and occipital lobes ( p < 0.05). Conclusion: This landmark based method can accommodate age and disease-related changes in brain morphology. It may be particularly useful for the study of neurodegenerative and cerebrovascular disease and for the validation of template-based automated techniques. D 2002 Elsevier Science B.V. All rights reserved. Keywords: Brain lobes; Regional segmentation; Anatomical landmark; White matter hyperintensities 1. Introduction Neuropathological studies suggest a differential susceptibility of cerebral cortical regions for different neurodegenerative processes [1,2] and cerebro-vascular disease [3,4]. A major challenge in clinical research on neurodegenerative or cerebrovascular disorders of the brain is the reliable identification of regional pattern of disease-related cerebral atrophy and the regional distribution of vascular cerebral lesions. * Corresponding authors. Tel.: +49-89-5160-5860; fax: +49-89-5160-5856. ** Co-corresponding author. E-mail addresses: stt@psy.med.uni-muenchen.de (S.J. Teipel), hampel@psy.med.uni-muenchen.de (H. Hampel). To show the differential involvement in-vivo, two basic methodological approaches exist to segment cerebral regions (or calculate lobar volumes) on MRI scans: (a) an automated method relying on template and (b) a manual method based upon landmarks. There have been several reports on automated methods [5 8] for quantification of the lobe volumes by first normalizing the magnetic resonance images to the Talairach and Tournoux template [9] and segmenting the brain regions based on the template. The template defines every pixel as a member of a specific region and that segmentation is then transferred to the MRI scan of interest. The advantage of automated methods is speed, reliability, and ease of use by the operator. The disadvantage is that it does not take into account differences in shape and variability of the cortex, which are more pronounced in patient populations. In addition, the normalization process will inevitably cause 0022-510X/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved. PII: S0022-510X(01)00667-0

36 A.L.W. Bokde et al. / Journal of the Neurological Sciences 194 (2002) 35 40 an averaging process to take place and individual differences may be lost, which can hide differences in structure between normal populations and patient populations. There are several reports on manual methods for tracing a single or two lobes of the brain, typically the temporal and/or frontal lobe, on a slice by slice basis (see for example Refs. [10 13]) or on a reconstructed 3-D model of the cortex [14]. There is a report on the regional segmentation of the five lobes in the cortex [15]. These methods trace each individual region and segmentation methods are used to delineate brain from CSF. The main advantage of these methods is that differences in shape and variability can be accounted for in application to normal subjects and patient populations. The disadvantages are that it can be time consuming, landmarks can be difficult to detect, and the reliability between and within raters over time has to be confirmed. The objective of this work is the development of a rapid and easy to use landmark-based technique for regional segmentation of the brain into different lobes. The segmentation is manual, with the traces done with reference to prominent sulcal points and drawing well-defined lines when no sulci and/or gyri reference landmarks are available. We will demonstrate the usefulness of the technique with its application to regional measures of white matter hyperintensities (WMH) in a group of patients with vascular dementia and an age-matched healthy control group. 2. Materials and methods 2.1. Subjects The technique was evaluated in a total of 10 subjects that comprised five normal healthy subjects, three patients with clinical probable Alzheimer s disease, and one patient each with clinical probable vascular dementia and fronto-temporal degeneration. The technique was applied to a vascular dementia patient group of 10 patients with average age of 67.7 yearsf7.8 (averagefs.d.) and an age matched control group (age = 69.0 yearsf5.6) of 10 subjects. Alzheimer s disease was diagnosed according to NINCDS-ADRDA criteria [16], vascular dementia was diagnosed according to NINDS-AIREN criteria [17] and fronto-temporal degeneration was diagnosed according to criteria of Neary et al. [18]. All subjects (and/or their legal care-givers) gave informed written consent to participate in this project, which was approved by the Ethics Committee of Ludwig Maximilian University Medical Faculty. 2.2. Magnetic resonance imaging The lobe segmentation was done on images obtained with an MPRAGE sequence (TR = 11.6 ms, TE = 4.9 ms, TI = 300 ms, TD = 0, flip angle = 12, slab thickness = 180 mm), sagittally oriented volumes with 128 slices, a slice thickness of 1.2 mm, 256256 pixels per slice, and in-plane pixel size of 1.0 mm. The scanner used was a 1.5 T Siemens Magnetom Vision MRI scanner (Siemens, Erlangen, Germany). Some volume sequences had 512512 in-plane pixels that were reduced to 256256 pixels per slice before processing. To obtain the measures of white mater hyperintensities (WMH), a fast FLAIR (TR = 9000 ms, TE = 110 ms, TI = 2500 ms, echo train length = 7) sequence was used, with each volume having 20 slices, a slice thickness of 6 mm, a pixel size of 0.93750.9375 mm, and each slice had 256256 pixels. 2.3. Description of segmentation technique The initial step of this technique was to remove all nonbrain tissue and cerebro-spinal fluid (CSF) regions using a semi-automated threshold technique. This segmentation technique consisted of a histogram of the volume of data and picking two thresholds between which the values corresponded mainly to tissue outside of the brain. Then a seed was placed in the skull area of the image and region growing technique was applied, with the growth limited to pixel values within the range defined by the two thresholds. As the final step, the mask produced in the previous step was edited so that no brain tissue was deleted and all nonbrain tissue was removed. Next, the brain images were re-oriented, if necessary, such that in the coronal and axial views the inter-hemispheric fissure was vertical on the screen. The vertical alignment was important because the regional segmentation depends on a complete lateral view of each hemisphere when constructing the surface projection. We used the atlas by Michio et al. [19] for sulcal anatomical information. The technique was implemented with Analyze AVW (Mayo Foundation, Rochester, MN, USA). The steps taken to implement this technique will now be described. First, orient brain images in the sagittal orientation, and go to the slice which shows the biggest part of the falx in the left hemisphere. Then go to the first slice in which the parieto-occipital sulcus is clearly visible. Mark the location (point B in Fig. 1) of the sulcus on the posterior side of the brain (we used lines outside the brain to denote the location of sulcus). Brain tissue below the parietooccipital sulcus is the medial surface of the occipital lobe (with the exception of the cerebellum). Trace the occipital lobe on this slice. Based on our experience, the first slice on which the parieto-occipital sulcus is clearly visible is between 5 to 8 mm from the inter-hemispheric fissure. Redo the process for the right hemisphere. Reconstruct a 3D-model of the brain (surface projection), which will show the lines indicating the location of the parieto-occipital sulcus. We always start with the left hemisphere. Identify the central sulcus by the following procedure: look up the three frontal gyri, which run in the anterior posterior direction. At the posterior end of these gyri, the first gyrus which runs perpendicular to them is the

A.L.W. Bokde et al. / Journal of the Neurological Sciences 194 (2002) 35 40 37 Fig. 1. Medial surface showing line marking of the parieto-occipital sulcus in the posterior region of the brain. Point B is the posterior end of the parieto-occipital sulcus. The occipital lobe has been segmented from the rest of the brain. precentral gyrus. The central sulcus is the posterior border of the precentral gyrus. Draw a trace along the central sulcus. This trace demarcates the border between the frontal and parietal lobes. The trace should extend from the interhemispheric fissure to the sylvian fissure. Because in most cases the central sulcus and sylvian fissure do not intersect, the trace extends with a straight line from the lower end of the central sulcus to the sylvian fissure. Next draw a trace along the sylvian fissure anteriorly and include the entire frontal cortex (see Fig. 2). Draw another trace around the rest of the cortex that includes the parietal, occipital and temporal lobes, but exclude the cerebellum and brain stem. The cerebellum and brain stem can be edited from the images at a later stage. Draw a trace along the sylvian fissure (from anterior to posterior direction) and terminate the trace on the beginning of the ascending terminal ramus of the sylvian fissure. The angular gyrus is located around the ascending terminal gyrus. The end of the trace is point A in Fig. 2. Draw a straight line from the parietoccipital sulcus (indicated in the first step, point B) to point A. At the center of this line, point C in Fig. 1, draw a line perpendicular to the lower side of the brain-point H in Fig. 1. Point D along this line is where it intersects the tissue non-tissue border of the brain. The mid-point along line CD is denoted point X. Draw a straight line from X to A. Next draw a straight line from point B through point X and continue the line until outside of the brain tissue image. The preoccipital notch, a landmark for the border of the parietal and temporal lobes, was not visible in any of the subjects that we analyzed even with the high resolution images that we obtained. Thus the artificial lines that are described above were developed to delineate the borders between the temporal, occipital and parietal lobes. At this point, we have defined all four lobes on the surface of the brain. The frontal lobe is the region indicated with F in Fig. 2, the parietal lobe is indicated with P, the occipital lobe is an O and the temporal lobe is indicated with a T. One can see that the frontal lobe is one area in our tracing, the parietal lobe three areas, the occipital lobe two areas and the temporal lobe one area. The surface of the lobes are now painted with different colors so that when we view individual slices the lobe segmentation will be visible along the cortex surface. The next step is to define the lines separating the lobes in the interior volume of the brain. The brain images are now re-oriented to the axial configuration with the surface of the brain in different colors depending upon region (see Fig. 3 for an example). To separate the frontal and parietal lobes, on every slice draw a perpendicular line from the bottom of the central sulcus to the inter-hemispheric fissure. In the brains in which the inferior end of the central sulcus does not reach the sylvian fissure, the perpendicular line will go from the border between the two lobes to the interhemispheric fissure. Between occipital and parietal lobes draw the shortest straight line from the defined border on the brain surface to the ventricle. The internal border between the occipital and temporal lobes, and between the parietal and temporal lobes, is a straight perpendicular line from the border on the brain surface to the interhemispheric fissure. The cerebellum is well defined and no processing was done on this structure. Fig. 2. Tracings of the regional segmentation of the left hemisphere. The regions are denoted by F frontal cortex, T temporal cortex, P parietal cortex and O occipital cortex. The other letters (A, B, C, D, E, H, X) denote the intersections of lines or reference points that are described in text. The brain image has been filtered to darken it so that the white lines and letters can be better seen.

38 A.L.W. Bokde et al. / Journal of the Neurological Sciences 194 (2002) 35 40 divided by the average of both measures) to assess the extent of difference between measures. Additionally, we calculated the intra-class correlation coefficient to assess reliability between and within raters. The ICC does not only take into account correlations between measures, but is also sensitive to systematic differences in absolute values of measures between ratings [21]. A linear model was used to test for the effect of diagnosis on the distribution of WMH. The variables in the model were the WMH density on the temporal, occipital, parietal and frontal lobes and diagnosis. To test for differences in WMH density in the four regions between groups, we used the Mann Whitney test. 4. Results Fig. 3. Transfer of the regional segmentation to the FLAIR image, with the different colors (shown in grey only) on the surface of the brain indicating the different lobes of the brain. External lines to the brain image have been added so that different regions can be clearly distinguished. The regions outlined in the white matter region are the white matter hyperintensities. All segmentation measures were done by two independent investigators blinded to clinical information from the 10 subjects. Additionally, segmentation from the 10 subjects were repeated by one investigator. For the second measurement the 10 subjects were blinded a second time and randomly mixed with broader set of 42 scans which were measured for the first time. With the above definition of the regions, the structural volume was registered to the FLAIR volume with AIR 3.08 using rigid body registration [20]. The FLAIR image was overlaid over the co-registered segmented MRI. In the FLAIR volumes, the WMH were measured using a semiautomated thresholding technique (see Fig. 3). Areas of WMH were divided by lobe volume to obtain a value for the density of WMH for each lobe. 3. WMH measures were done by one investigator blinded to clinical information 3.1. Statistics To assess inter-rater reliability, scans from the 10 subjects were segmented by two independent investigators; for assessment of intra-rater-reliability, scans were measured twice by one investigator blinded to clinical diagnosis. We used the relative error (positive difference between measures In the first group of data, comprised of healthy controls and various dementia groups, the measured surface areas between independent raters and within one rater of the four different lobes is displayed in Table 1. The relative error and the intra-class correlation coefficients are displayed in Tables 2 and 3, respectively. In the pilot data of the vascular dementia group and the age-matched healthy controls (the quantitative data shown in Table 4), using a linear model with the lobe region and diagnosis as independent variables, we found that both variables contributed significantly to the explanatory power of the model ( p < 0.05). The difference of WMH density between groups, using the Mann Whitney test, was significant ( p < 0.05) in all four regions with higher densities in the vascular dementia patients. Within each group, the WMH density was significantly greater in the frontal and parietal lobes compared to the Table 1 Surface areas of the different lobes measured by the two independent raters and the two measurements made by the single rater Measured surface areas of the two raters Rater 1(a) Rater 1(b) Rater 2 Frontal Right 4842F615 4911F626 4859F652 Left 4642F673 4705F717 4620F673 Temporal Right 3173F441 3135F396 3153F492 Left 2958F320 2914F276 3003F319 Parietal Right 3218F378 3183F360 3202F336 Left 3150F350 3105F351 3185F333 Occipital Right 1398F249 1441F207 1419F219 Left 1566F261 1655F273 1575F268 Values are meanf1 SD (mm 2 ).

A.L.W. Bokde et al. / Journal of the Neurological Sciences 194 (2002) 35 40 39 Table 2 Relative error of surface areas of the different lobes between two independent raters and within a single rater Inter-rater Intra-rater temporal and occipital lobes (using the Wilcoxon test, p < 0.05). 5. Discussion Relative errorfsd (%) Relative errorfsd (%) Frontal Right 1.4F0.9 1.6F1.3 Left 1.5F1.6 1.9F1.3 Temporal Right 3.5F3.1 4.3F2.4 Left 3.1F3.3 3.1F1.7 Parietal Right 1.8F1.1 2.0F1.3 Left 2.2F1.4 2.4F2.6 Occipital Right 5.2F4.2 6.9F5.1 Left 2.2F1.9 5.6F4.1 In the present study we evaluated a newly developed method for regional segmentation of brain lobar volumes from structural MRI based on anatomical landmarks and well-defined lines. To investigate the potential use of this technique on regional changes in different patient populations, we applied it to a small group of vascular dementia patients and age-matched healthy controls. The intra-class correlation coefficient is highest for the frontal lobe. The correlation coefficients are above 0.96 for Table 3 Intra-class correlation coefficient of surface areas of the different lobes between two independent raters and within a single rater Inter-rater Intra-rater Intra-class correlation coefficient Frontal Right 0.9952 0.9960 Left 0.9963 0.9953 Temporal Right 0.9657 0.9699 Left 0.9610 0.9712 Parietal Right 0.9963 0.9917 Left 0.9877 0.9802 Occipital Right 0.9678 0.9153 Left 0.9916 0.9863 Intra-class correlation coefficient Table 4 Regional white matter hyperintensity densities of the vascular dementia patient group and the healthy control group Healthy controls (meanfsd) (10 4 mm 1 ) Vascular dementia (meanfsd) (10 4 mm 1 ) Frontal 3.53F3.60 44.83F61.26 Temporal 1.02F1.19 3.69F2.73 Parietal 4.23F4.90 50.91F55.58 Occipital 0.82F1.43 11.62F12.46 all regions except the right occipital region that has a value of 0.91. The relative error follows the same pattern as the intra-class correlation coefficient with the lowest values for the frontal lobes and highest values for the occipital lobe. The intra-class correlation coefficient in the other regions are not as high as the frontal lobe (and the relative error is not as low) because the segmentation of these regions is through a combination of anatomical landmarks and lines drawn with respect to anatomical landmarks and to each other. The frontal lobe segmentation is done with reference to anatomical landmarks only. The results are consistent within raters and between raters. Comparing our reliability results to those obtained by Aylward et al. [14] on the frontal lobe, both studies had an intra-class correlation coefficient of 0.99 both within and between raters for each of the frontal lobes. Aylward et al. [14] obtained a mean relative error between two raters of 1.2% whereas our mean inter-rater relative error was 1.6% and 1.9% for the right and left frontal lobes, respectively. Both studies had 10 subjects for validation but Aylward et al. s [14] study only had healthy subjects, whereas we had a group composed of healthy subjects and neuro-degenerative disease patients. The reliability of our technique, for the same region, is similar to Aylward et al. s [14] but with a more heterogeneous group. In Fukui and Kertesz [15], the intra- and inter-rater correlation coefficient calculated over all regions, in five subjects, are 0.98 and 0.94, respectively. In our method, when we calculated the intra- and inter-rater correlation coefficients analogue to this study [15], we obtained 0.99 for both coefficients. The high reliability values obtained indicate that the area measures are consistent across the various different groups that we employed. In addition, it can be applied when there are large changes in shape and asymmetries between the lobes without altering the robustness of the method. Changes in sulci and gyri variations, such as may exist in patient populations, will not negatively impact the robustness of the method. It can also serve as a technique for establishing baseline measures on which automatic techniques may be compared to. The regional segmentation can be accomplished in 15 to 20 min per stripped brain, which is obtained after two to five brains of training. This performance is based on the experience of the two raters, one a psychiatrist with 5 years experience in MRI research and the other a dental school student (the intra-rater reliability

40 A.L.W. Bokde et al. / Journal of the Neurological Sciences 194 (2002) 35 40 values are from this rater). We believe that the time needed to segment the brain, as well as obtain the high correlations that we do, is possible with raters who are not experts in neuro-anatomy but have been trained to recognize the specific landmarks specified here. Using the technique, we found greater density of WMH in the frontal and parietal lobes compared to the occipital and temporal lobes both in VD patients and in healthy controls. This is consistent with neuropathological evidence for predominant frontal lobe involvement of cerebral white matter in vascular disease [3] and the architecture of the cerebral vasculature with long penetrating arteries crossing the frontal lobe white matter. The difference between VD patients and healthy controls agrees with previous findings based on semi-quantitative rating scales that showed significantly greater WMH load in VD than in healthy aging [22]. These findings illustrate the potential use of the method for questions relating to the regional distribution of subcortical lesions such as in multiple sclerosis and vascular disease. In conclusion, this method is a powerful tool for the regional segmentation of structural MRI volumes. A method to determine the regional distribution of lobar atrophy and subcortical lesions in-vivo is of high clinical value, because it allows to increase diagnostic accuracy in neurodegenerative and cerebrovascular disorders and to follow the longitudinal course of regional cerebral changes over time. Acknowledgements Part of the presented material originates from the doctoral thesis of Y. Zebuhr (Ludwig-Maximilian University, Munich, Germany; in preparation). Part of this work was supported by a grant of Eisai (Frankfurt) and Pfizer (Karlsruhe), Germany to H.H. and S.J.T. and by a grant of the Medical Faculty of the Ludwig-Maximilian University, Munich, Germany to S.J.T. References [1] Braak H, Griffing K, Braak E. Neuroanatomy of Alzheimer s disease. Alzheimer s Res 1997;3:235 47. [2] Bergmann M, Kuchelmeister K, Schmid KW, Kretzschmar HA, Schroder R. Different variants of frontotemporal dementia: a neuropathological and immunohistochemical study. Acta Neuropathol 1996;92:170 9. [3] Yamanouchi H, Sugiura S, Tomonaga M. Decrease in nerve fibres in cerebral white matter in progressive subcortical vascular encephalopythy of Binswanger type. J Neurol 1989;236:382 7. [4] Erkinjuntti T, Benavente O, Eliasziw M, Munoz DG, Sulkava R, Haltia M, et al. Diffuse vacuolization (spongiosis) and arteriolosclerosis in the frontal white matter occurs in vascular dementia. Arch Neurol 1996;53:325 32. [5] Andreasen NC, Rajarethinam R, Cizadlo T, et al. Automatic atlasbased volume estimation of human brain regions from the MR images. J Comput Assist Tomogr 1996;20:98 106. [6] Collins DL, Holmes CJ, Peters TM, Evans AC. Automatic 3-D modelbased neuroanatomical segmentation. Hum Brain Mapp 1995;3: 190 208. [7] Thompson PM, MacDonald D, Mega MS, Holmes CJ, Evans AC, Toga AW. Detection and mapping of abnormal brain structure with a probabilistic atlas of cortical surfaces. J Comput Assist Tomogr 1997;2:567 81. [8] Goldszal AF, Davatzikos C, Pham DL, Yan MXH, Bryan RN, Resnick SM. An image processing system for qualitative and quantitative volumetric analysis of brain images. J Comput Assist Tomogr 1998; 22:827 37. [9] Talairach J, Tournoux P. Co-Planar Stereotaxic Atlas of the Human Brain. New York: Thieme Medical; 1988. [10] Jack JCR, Twomey CK, Zinsmeister A, et al. Anterior temporal lobes and hippocampal formations: normative volumetric measurements from MR images in young adults. Radiology 1989;172:549 54. [11] Kertesz A, Polk M, Black SE, Howell J. Sex, handedness, and the morphometry of cerebral asymmetries on magnetic resonance imaging. Brain Res 1990;530:40 8. [12] Buchsbaum MS. The frontal lobes, basal ganglia, and temporal lobes as sites for schizophrenia. Schizophr Bull 1990;16:379 89. [13] Turetsky B, Cowell PE, Gur RC, Grossman RI, Shtasel DL, Gur RE. Frontal and temporal lobe brain volumes in schizophrenia: relationship to symptoms and clinical subtype. Arch Gen Psychiatry 1995; 52:1061 70. [14] Aylward EH, Augustine A, Li Q, Barta PE, Pearlson GD. Measurement of frontal lobe volume on magnetic resonance imaging scans. Psychiatry Res 1997;75:23 30. [15] Fukui T, Kertesz A. Volumetric study of lobar atrophy in Pick complex and Alzheimer s disease. J Neurol Sci 2000;174:111 21. [16] McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM. Clinical diagnosis of Alzheimer s disease: report of the NINCDS- ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer s Disease. Neurology 1984; 34:939 44. [17] Roman GC, Tatemichi TK, Erkinjuntti T, et al. Vascular dementia: diagnostic criteria for research studies. Report of the NINDS-AIREN International Workshop. Neurology 1993;43:250 60. [18] Neary D, Snowden JS, Gustafson L, et al. Frontotemporal lobar degeneration: a consensus on clinical diagnostic criteria. Neurology 1998;51: 1546 54. [19] Michio O, Kubik S, Abernathey CD. Atlas of the Cerebral Sulci 51. Stuttgart: Thieme Verlag; 1990. [20] Woods RP, Grafton ST, Watson JD, Sicotte NL, Mazziotta JC. Automated image registration: I. General methods and intrasubject, intramodality validation. J Comput Assist Tomogr 1998;22:139 52. [21] Bartko JJ, Carpenter WT. On the methods and theory of reliability. J Nerv Ment Dis 1976;163:307 17. [22] Schmidt R. Comparison of magnetic resonance imaging in Alzheimer s disease, vascular dementia and normal aging. Eur Neurol 1992; 32:164 9.