Radiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online before print May 5, 2008, 10.1148/radiol.2481070876
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
2481070876v1
248/1/194    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Colliot, O.
Right arrow Articles by Lehéricy, S.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Colliot, O.
Right arrow Articles by Lehéricy, S.
(Radiology 2008;248:194-201.)
© RSNA, 2008

Neuroradiology

Discrimination between Alzheimer Disease, Mild Cognitive Impairment, and Normal Aging by Using Automated Segmentation of the Hippocampus1

Olivier Colliot, PhD, Gaël Chételat, PhD, Marie Chupin, PhD, Béatrice Desgranges, PhD, Benoît Magnin, MSc, Habib Benali, PhD, Bruno Dubois, MD, PhD, Line Garnero, PhD, Francis Eustache, PhD, and Stéphane Lehéricy, MD, PhD

1 From the Cognitive Neuroscience and Brain Imaging Laboratory, Centre National de la Recherche Scientifique, UPR640-LENA (O.C., M.C., L.G.), Institut National de la Santé et de la Recherche Médical (INSERM) U678 (B.M., H.B.), INSERM U610 (B.M., B.Dubois, S.L.), and Department of Neuroradiology, Center for Neuroimaging Research (S.L.), Université Pierre et Marie Curie-Paris 6, Hôpital de la Pitié-Salpêtrière, 47, boulevard de l'Hôpital, 75651 Paris Cedex 13, France; and INSERM–Ecole Pratique des Hautes Etudes, Université de Caen Basse-Normandie, U923, E0218, Cyceron, Centre Hospitalo-Universitaire de Caen, Caen, France (G.C., B.Desgranges, F.E.). Received May 18, 2007; revision requested July 26; revision received October 10; accepted December 28; final version accepted January 29, 2008. Address correspondence to O.C. (e-mail: olivier.colliot{at}chups.jussieu.fr).


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCE IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
 References
 
Purpose: To prospectively evaluate the accuracy of automated hippocampal volumetry to help distinguish between patients with Alzheimer disease (AD), patients with mild cognitive impairment (MCI), and elderly controls, by using established criteria for patients with AD and MCI as the reference standard.

Materials and Methods: The regional ethics committee approved the study and written informed consent was obtained from all participants. The study included 25 patients with AD (11 men, 14 women; mean age ± standard deviation [SD], 73 years ± 6; Mini-Mental State Examination (MMSE) score, 24.4 ± 2.7), 24 patients with amnestic MCI (10 men, 14 women; mean age ± SD, 74 years ± 8; MMSE score, 27.2 ± 1.4) and 25 elderly healthy controls (13 men, 12 women; mean age ± SD, 64 years ± 8). For each participant, the hippocampi were automatically segmented on three-dimensional T1-weighted magnetic resonance (MR) images with high spatial resolution. Segmentation was performed by using recently developed software that allows fast segmentation with minimal user input. Group differences in hippocampal volume were assessed by using Student t tests. To obtain robust estimates of P values, the correct classification rate, sensitivity, and specificity, bootstrap methods were used.

Results: Significant hippocampal volume reductions were detected in all groups of patients (–32% in AD patients vs controls, P < .001; –19% in MCI patients vs controls, P < .001; and –15% in AD patients vs MCI patients, P < .01). Individual classification on the basis of hippocampal volume resulted in 84% correct classification (sensitivity, 84%; specificity, 84%) between AD patients and controls and 73% correct classification (sensitivity, 75%; specificity, 70%) between MCI patients and controls.

Conclusion: This automated method can serve as an alternative to manual tracing and may thus prove useful in assisting with the diagnosis of AD.

© RSNA, 2008


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCE IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
 References
 
Alzheimer disease (AD) is the most common cause of dementia in the elderly (1). Early and accurate diagnosis of AD can be challenging. In recent years, the early clinical signs of AD have been extensively investigated, leading to the concept of amnestic mild cognitive impairment (MCI) (2–4). A challenge for modern neuroimaging is to help in the diagnosis of early AD, particularly with amnestic MCI patients. Early diagnosis of AD patients allows early treatment with cholinesterase inhibitors, which have been shown to delay institutionalization and improve or stabilize cognition and behavioral symptoms (5,6).

Three-dimensional magnetic resonance (MR) imaging with high spatial resolution allows visualization of subtle anatomic changes and thus can help detect brain atrophy in the initial stages of the disease. The hippocampus is known to be affected in the earliest stages of AD (7,8). Many reporters in studies have thus assessed hippocampal atrophy in AD by using manual segmentation at MR (9–19). These studies have demonstrated that manual MR volumetry of the hippocampus can help distinguish patients with AD from elderly controls with a high degree of accuracy (80%–90%). However, manual segmentation of the hippocampus requires a high degree of anatomic training and is observer dependent and time consuming (≥1 hour). Although more suited to clinical practice, visual evaluation of atrophy at multiplanar MR is difficult and prone to subjectivity (20).

We have developed an automated method that is able to segment the hippocampus on MR images (21). This method has been compared with manual segmentation in young healthy participants and patients with AD and has proved reliable, fast, and accurate (about 8% relative volume error when compared with manual segmentation) (21). Thus, the purpose of our study was to prospectively evaluate the accuracy of automated hippocampal volumetry to distinguish patients with AD, MCI, and elderly controls, by using established criteria for patients with AD and MCI as the reference standard.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 ADVANCE IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
 References
 
Participants
The regional ethics committee approved this study and written informed consent was obtained from all participants. In our study, 25 patients with AD (11 men, 14 women; mean age ± standard-deviation (SD), 73 years ± 6; age range, 62–81 years; Mini-Mental State Examination (MMSE) score, 24.4 ± 2.7; score range, 19–29) and 24 patients with amnestic MCI (10 men, 14 women; mean age ± SD, 74 years ± 8; age range, 55–87 years; MMSE score, 27.2 ± 1.4; score range, 24–29) were selected from the database of patients prospectively recruited at the Centre Hospitalo-Universitaire of Caen, France. From this database, we included participants whose scanning parameters followed the protocol described below and whose MR images were free of substantial visible motion artifacts. These samples partly overlap with those of previous publications (22–25). The diagnosis for probable AD was made according to the National Institute of Neurological and Communicative Disorders and Stroke–Alzheimer Disease and Related Disorders Association criteria (26). The diagnosis of MCI was made on the basis of the criteria from Petersen et al (27). All MCI patients were evaluated every 6 months over an 18-month period to assess conversion (ie, whether they met National Institute of Neurological and Communicative Disorders and Stroke–Alzheimer Disease and Related Disorders Association criteria for probable AD). Patients were declared as converters if they had impaired performances (>1.5 SD below the normal mean, according to age and education, when available) in at least one of the general intellectual function scales, as well as in at least two areas of cognition, including memory, that lead to impaired daily activities as judged by the clinicians from the consultation interviews. Post hoc exclusion criteria included presence of substantial neurologic, psychiatric, or any other medical disease that could affect brain function or structure and normal episodic memory performance at follow-up. At completion of the 18-month follow-up period, eight (of 23, 35%) MCI patients were declared as converters (three men, five women; mean age ± SD, 77 years ± 4; age range, 71–82 years; MMSE score, 26.5 ± 1.4; score range, 24–28), 15 patients still had isolated memory deficits (nonconverters: seven men, eight women; mean age ± SD, 72 years ± 9; age range, 55–87 years; MMSE score, 27.7 ± 1.2; score range, 26–29), and one female MCI patient refused follow-up and was thus excluded from the analysis of converters and nonconverters. The annual conversion rate was thus 23%.

AD and MCI patients were compared with 25 elderly healthy controls (13 men, 12 women; mean age ± SD, 64 years ± 8; range, 51–84 years) with normal memory performance, as assessed by using tests of episodic, semantic, and working memory and without vascular lesions visible at MR. To exclude vascular lesions, all controls were found to have normal signal intensity on one or more T1-, T2- or fluid-attenuated inversion recovery–weighted MR images; notably, no substantial white matter T2–fluid-attenuated inversion recovery–weighted areas of hyperintensity (fewer than five pinpoint areas of hyperintensity <4 mm in size) (28). The controls were screened for the absence of cerebrovascular risk factors, mental disorders, substance abuse, head trauma, substantially abnormal MR findings or biologic abnormality, and incipient dementia by using a memory test battery. Control participants were recruited by means of advertisements in local newspapers. Control participants were required to be over 50 years old. There was no specific sex criterion. The recruitment of participants (controls and patients) began in 1999 and ended in 2004.

MR Acquisition
Within 2 months from inclusion in this study for controls and a few days for MCI and AD patients, each participant underwent T1-weighted volumetric MR imaging, which consisted of a set of 128 adjacent transverse sections parallel to the anterior commissure–posterior commissure line with a section thickness of 1.5 mm and a pixel size of 0.9375 x 0.9375 mm, by using the spoiled gradient-echo sequence (repetition time msec/echo time msec, 10.3/2.1; field of view, 24 x 18 cm; and matrix, 256 x 192). All MR data sets were acquired with the same 1.5-T imager (Signa Advantage Echospeed; GE Healthcare, Milwaukee, Wis).

Automated Segmentation of the Hippocampus
The segmentation of the hippocampus was performed by using an automated method we previously developed. This approach segments the hippocampus and the amygdala simultaneously on the basis of competitive region-growing between these structures. It includes prior knowledge of the relative positions of these structures with respect to anatomic landmarks, which are automatically identified. During the iterative segmentation process, 11 sets of landmarks were automatically retrieved at the border of the deforming structures. Two of these landmarks were located at the interface between the hippocampus and the amygdala, based on the location of the alveus and the temporal horn of the lateral ventricle. Three landmarks were defined for the hippocampus only, two based on the location of the alveus and one based on the location of the hippocampal sulcus. One was defined for the amygdala only, based on the location of the isthmus of the temporal lobe. Finally, five were defined for both the hippocampus and the amygdala, based on the location of the parahippocampal gyrus and the temporal horn of the lateral ventricle. More details can be found in a previously published report (21). It should be noted that all of these landmarks were found automatically by using the algorithm and that no intervention from the operator was required. It is thus not necessary to be able to locate these landmarks on the MR image to use the automated segmentation software.

Segmentations (Figs 1, 2) were processed by a trained operator (O.C., with 6 years experience in structural MR analysis), who was blinded to all clinical data and diagnostic categories and was trained to use the segmentation software by processing a set of 20 images.


Figure 1
View larger version (38K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1: MR images of hippocampal segmentation process. (a) Sagittal view shows four bounding box limits around amygdalohippocampal complex. (b) Sagittal view shows two seed voxels placed in hippocampal head (red) and amygdala (green); these seeds do not necessarily belong to same section. (c) Sagittal view shows algorithm automatically segmenting structure. (d) Three-dimensional surface renderings corresponding to automated segmentations of hippocampus and amygdala. Here, T1-weighted images were acquired by using inversion-recovery fast spoiled gradient-echo sequence (124 adjacent transverse sections; section thickness, 1.3 mm; pixel size, 0.9375 x 0.9375 mm; 14.3/6.3; inversion time msec, 600; field of view, 24 x 18 cm; and matrix, 256 x 192). Image presented to illustrate segmentation method and is not part of this study.

 

Figure 2A
View larger version (46K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2a: MR images show automated hippocampal segmentation. (a) Coronal and sagittal reconstructions of patient with AD. (b) Coronal and sagittal reconstructions of healthy elderly control. T1-weighted MR images acquired by using spoiled gradient-echo sequence (128 adjacent transverse sections parallel to anterior commissure–posterior commissure line; section thickness, 1.5 mm; pixel size, 0.9375 x 0.9375 mm; 10.3/2.1; field of view, 24 x 18 cm; and matrix, 256 x 192).

 

Figure 2B
View larger version (43K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2b: MR images show automated hippocampal segmentation. (a) Coronal and sagittal reconstructions of patient with AD. (b) Coronal and sagittal reconstructions of healthy elderly control. T1-weighted MR images acquired by using spoiled gradient-echo sequence (128 adjacent transverse sections parallel to anterior commissure–posterior commissure line; section thickness, 1.5 mm; pixel size, 0.9375 x 0.9375 mm; 10.3/2.1; field of view, 24 x 18 cm; and matrix, 256 x 192).

 
The method requires initialization from the operator. First, a bounding box is manually defined around the amygdalohippocampal complex (Fig 1a) by selecting six sections, which correspond to the boundaries of the hippocampus and the amygdala in each direction. The dimensions of the bounding box are typically around 30 x 50 x 20 voxels. Then, two seeds are placed in the hippocampus and the amygdala (Fig 1b). These seeds constitute the starting points of the deformation process. They are positioned close to the center of the amygdala and the center of the head of the hippocampus. Last, starting from these two seeds, the algorithm automatically aggregates voxels depending on the signal intensities, regularity of the region, and the detection of anatomic landmarks, and converges to the segmentation of the two structures (Fig 1c, 1d). Additionally, three parameters of the algorithm can be adjusted depending on the participant, one radiometric and two geometric parameters. The radiometric parameter controls the ratio between the signal intensity characteristics (mean and SD) of the hippocampus and the amygdala and those of the gray matter. Even though the amydgala and the hippocampus are mainly gray matter structures, their signal intensity on T1-weighted MR images is slightly different from that of the other gray matter. Specifically, we observed that their mean signal intensity is slightly lower than the mean signal intensity of the gray matter of the bounding box, whereas their SD is higher. The ratio between the mean and SD of the signal intensity of the amygdala and the hippocampus and those of the other gray matter may depend on the image contrast. For this reason, three preset values controlling these signal intensity ratios can be chosen depending on the visual contrast of the image. One of the geometric parameters controls the degree of shape anisotropy in the head of the hippocampus. The other geometric parameter controls the anisotropy in the tail of the hippocampus. These geometric parameters are adjusted when the hippocampus is atrophied. The average total processing time for each amygdalohippocampal complex (measured on a subset of 10 participants: three controls, two MCI patients, and five AD patients) was 11 minutes for the complete procedure, including bounding box definition, seed positioning, parameter adjustment, and automatic deformation by using the algorithm. Intrarater reproducibility was assessed by performing the segmentation twice, after a 1-week interval, on a randomly selected subsample of 10 participants (three AD patients, three MCI patients, and four controls). The rater was blinded to all previous parameters or visualization adjustments. The mean intrarater relative volume difference ± SD was 7% ± 7. The difference between the two measurements was not significant (P = .4, paired Student t test).

Normalization with Total Intracranial Volume
For subsequent classification of AD patients and healthy controls, hippocampal measurements were normalized to the total intracranial volume. The total intracranial volume was computed by using software (SPM5; Wellcome Department of Imaging Neuroscience, London, England), according to the following procedure:

1. Correction for intensity nonuniformity, spatial normalization to a common stereotaxic space, and tissue classification into gray matter, white matter, and cerebrospinal fluid by using a unified procedure (29).

2. Creation of an approximate intracranial mask by summing the three unmodulated tissue probability maps and creating a threshhold level of the result at 0.5 (this removes some non–cerebrospinal fluid voxels that were included in the cerebrospinal fluid map by using the segmentation process). Modulation is a procedure that allows preserving the amount of tissue during the normalization process by multiplying the tissues with the Jacobian determinants of the deformation (30).

3. Masking modulated tissue maps with the previous result.

4. Total intracranial volume is the sum of the masked modulated gray matter, white matter, and cerebrospinal fluid maps.

The normalized hippocampal volume is defined as NHV = TIVm x HV/TIV, where TIVm is the average total intracranial volume computed across all participants, which is constant, and HV is the hippocampal volume. The role of the constant multiplicative factor TIVm is simply to preserve the order of magnitude of NHV similar to that of HV. Last, normalized hippocampal volumes were averaged over both hemispheres. In the following, all reported volumes thus represent the average of the left and right hippocampal volumes.

Statistical Analysis
The statistical analysis was conducted (O.C.) by using in-house software (developed by B.M. and H.B).

Group analysis.—Group differences in normalized hippocampal volume between AD patients, MCI patients, and healthy controls were assessed by using Student t tests. To obtain a more robust estimate of the P value, we used a bootstrap method (31). We denoted S as the group of all subjects, S1 as the group of controls, and S2 as the group of patients. We worked with the null hypothesis that there were no differences between the mean values of the groups. We resampled the set S by using the null hypothesis, creating resampled sets S1 and S2 by drawing replacement participants from both groups. For the nth resampling S1,n, S2,n we computed the corresponding value T,n of the t test. We performed 5000 resamplings and calculated the percentile corresponding to the initial value of the t test in the set of values (T,n, n = 1–5000). According to the bootstrap theory, this percentile is a good estimate of the P value of our test. A P value of less than .05 indicated a significant difference.

Individual analysis.—For the automatic classification of AD patients versus controls, MCI patients versus controls, and AD patients versus MCI patients, each participant was assigned to the closest group. Specifically, if S1 and S2 are two groups of participants with respective means, defined as m1 and m2, a new individual with volume x is assigned to S1 if (x – m1) is less than (x – m2) and to S2 if otherwise. To obtain a robust estimate of the correct classification rate, sensitivity, and specificity, we used a bootstrap procedure for training set selection. To this purpose, we drew, without replacement, approximately 75% of each group to obtain a training set (S1, S2) and to estimate the means (m1 and m2). The remaining 25% were used as a test set. The procedure was repeated 5000 times. We thus obtained the correct classification rates, sensitivity, and specificity for the 5000 drawings.

The classification was also analyzed by using receiver operating characteristic curves, which indicate the relationship between sensitivity and 1–specificity for each intergroup discrimination. We computed the area under the curve, which is an index of overall discriminative ability.

Subgroup analysis.—To ensure that our findings were not biased by means of age or sex confounding effects, the same group and individual analyses were also performed on smaller groups of age-matched participants. To this purpose, we selected a group of 17 AD patients (six men, 11 women; mean age ± SD, 70 years ± 4; age range, 62–76 years; MMSE score, 24.1 ± 2.8; score range, 19–28), a group of 17 MCI patients (six men, 11 women; mean age ± SD, 70 years ± 6; age range, 55–79 years; MMSE score, 27.1 ± 1.3, range, 25–29), and of a group of 17 healthy elderly controls (10 men, seven women; mean age ± SD, 68 years ± 7; range, 60–84 years). We also performed the same analysis on smaller groups of sex-matched participants. To this purpose, we selected a group of 22 AD patients (10 men, 12 women; mean age ± SD, 73 years ± 5; age range, 62–81 years; MMSE score, 24.4 ± 2.5; score range, 19–28), a group of 22 MCI patients (10 men, 12 women; mean age ± SD, 73 years ± 7; age range, 55–85 years; MMSE score, 27.4 ± 1.4; score range, 24–29), and a group of 22 healthy elderly controls (10 men, 12 women; mean age ± SD, 64 years ± 8; range, 51–84 years).


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 ADVANCE IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
 References
 
Group Analysis
Normalized hippocampal volumes ± SD were 1.95 cm3 ± 0.46 (range, 0.98–3.10 cm3) for AD patients, 2.30 cm3 ± 0.46 (range, 1.28–3.10 cm3) for MCI patients, and 2.86 cm3 ± 0.46 (range, 1.74–4.05 cm3) for control participants. Significant hippocampal volume reductions were found in both AD (–32% [0.91 of 2.86 cm3] volume reduction, P < .001) and MCI patients (–19% [0.56 of 2.86 cm3] volume reduction, P < .001), compared with elderly controls (Table 1). AD patients also had significantly smaller hippocampi compared with MCI patients (–15% [0.35 of 2.30 cm3] volume reduction, P < .01). Among MCI patients, converters were found to have smaller hippocampi at baseline than did nonconverters (1.97 vs 2.47 cm3, 20% [0.50 of 2.47 cm3] volume reduction) (Fig 3). The converter subgroup was too small to compute a P value.


View this table:
[in this window]
[in a new window]

  		Table 1. Pairwise Intergroup Comparisons of Hippocampal Volumes

 

Figure 3
View larger version (9K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3: Normalized hippocampal volumes of MCI converters and nonconverters. Circles indicate volume of each participant, horizontal bars indicate mean of each group.

 
Individual Analysis
Correct classification rates were 84% for AD patients and 73% for MCI patients with respect to elderly controls and 69% for AD patients with respect to MCI patients (Table 2).


View this table:
[in this window]
[in a new window]

  		Table 2. Classification between Groups, Given the Normalized Hippocampal Volume

 
Regarding the receiver operating characteristic curves for intergroup discrimination (Fig 4), the area under the curve was 0.913 for AD patients versus controls, 0.808 for MCI patients versus controls and 0.721 for AD versus MCI patients.


Figure 4A
View larger version (22K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 4a: Receiver operating characteristic curves for intergroup classification. (a) AD patients versus controls. (b) MCI patients versus controls. (c) MCI patients versus AD patients.

 

Figure 4B
View larger version (23K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 4b: Receiver operating characteristic curves for intergroup classification. (a) AD patients versus controls. (b) MCI patients versus controls. (c) MCI patients versus AD patients.

 

Figure 4C
View larger version (23K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 4c: Receiver operating characteristic curves for intergroup classification. (a) AD patients versus controls. (b) MCI patients versus controls. (c) MCI patients versus AD patients.

 
Subgroup Analysis
For the age-matched subgroups of 17 participants each, normalized hippocampal volumes ± SD were 2.06 cm3 ± 0.48 (range, 0.98–3.10 cm3) for AD patients, 2.30 cm3 ± 0.50 (range, 1.28–3.10 cm3) for MCI patients, and 2.87 cm3 ± 0.50 (range, 1.74–4.05 cm3) for control participants. Significant hippocampal volume reductions were found in AD (–28% [0.81 of 2.87 cm3] volume reduction, P < .001) and MCI (–20% [0.57 of 2.87 cm3] volume reduction, P < .001) groups, compared with elderly controls. Correct classification rates were 81% for AD patients and 70% for MCI patients with respect to elderly controls.

For the sex-matched subgroups of 22 participants each, normalized hippocampal volumes ± SD were 1.93 cm3 ± 0.49 (range, 0.98–3.10 cm3) for AD patients, 2.27 cm3 ± 0.46 (range, 1.28–3.10 cm3) for MCI patients, and 2.84 cm3 ± 0.49 (range, 1.74–4.05 cm3) for control participants. Significant hippocampal volume reductions were found in both AD (–32% [0.91 of 2.84 cm3] volume reduction, P < .001) and MCI (–20% volume [0.57 of 2.84 cm3] reduction, P < .001) patients, compared with elderly controls. Correct classification rates were 82% for AD patients and 71% for MCI patients with respect to elderly controls.


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
ADVANCE IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
 References
 
Our study showed that automated segmentation was able to detect significant volume differences in both AD patients and patients with amnestic MCI. The results are in concordance with a number of studies regarding manual hippocampal segmentation, which have shown hippocampal atrophy in patients with AD (10–15,18,19) and in patients with MCI (32–37). Compared with elderly controls, we found an average of 32% hippocampal volume loss in patients with AD. This value is in the range of those reported in studies that used manual volumetry, with volume loss in AD between 23% and 34% (10,12,17,18,33). In patients with MCI, values ranged from 8% to 15% (32–35,37). We found a slightly higher relative volume loss in these patients (19%). However, it is important to note that MCI criteria are highly variable from one study (or laboratory) to another, as are the annual rates of conversion to AD. As a result, the degree of hippocampal atrophy, which appropriately depends on the underlying etiology, the severity of symptoms, and the proportion of converters, similarly shows inconsistencies. In our opinion, the slightly higher degree of atrophy found in our study should be related to the relatively high conversion rate of our MCI sample (23% each year), which is higher than that usually reported (about 15%). This rate should, in turn, reflect the use of strict criteria, including objective memory deficits, normal global cognition (as attested by using objective tests), and excluding impairment in other areas of cognition (also objectively assessed) or depression.

By using normalized hippocampal volumes, 84% of AD patients were correctly classified with respect to elderly controls. This value falls in the range of classification results given the manual segmentation of the hippocampus, which was between 82% and 90% for AD (10,12,17,18,33). For MCI patients, classification rates ranged from 60% to 74% (32–35,37). We found a discrimination rate of 73%. As for the degree of atrophy, this relatively high value should be related to the high conversion rate and thus to the use of strict criteria.

We found that MCI patients who later converted to AD had a 20% smaller hipppocampal volume at baseline than did nonconverters. This result should be interpreted with caution owing to the small number of converters. Nevertheless, this is in agreement with several studies regarding manual segmentation, which have reported that baseline hippocampal volume is an indicator of future progression to AD (38–42). This is also in concordance with studies based on visual rating, which demonstrated medial temporal atrophy in patients who subsequently converted to AD (43–45).

While several automated hippocampal segmentation methods have been proposed (46–51), few of them have been applied in patients with AD and/or MCI and rarely did researchers in studies report the accuracy of their technique to classify MCI or AD and controls. Carmichael et al (52) have assessed the performance of automated atlas-based segmentation by using several freely available registration methods (Automated Image Registration [University of California at Los Angeles, Los Angeles, Calif], Statistical Parametric Mapping [Wellcome Department of Imaging Neuroscience], Functional MR imaging Linear Image Registration Tool [University of Oxford, Oxford, England], and a fully deformable approach) in AD and MCI patients. They concluded that these approaches are less precise when applied to AD patients than controls but this should be tempered by the fact that these techniques were not specifically designed for this task. Fischl et al (53) proposed a general method, derived from a probabilistic atlas, to automatically label different noncortical structures, including the hippocampus, and applied this technique to patients with mild and questionable AD. The method helped identify significant group differences in terms of hippocampal volume but the authors did not investigate the classification of individual participants. Csernansky et al (54) used the high-dimensional brain mapping approach, on the basis of fluid registration with a template, to obtain hippocampal volumes and hippocampal shape differences between patients with very mild AD and controls. By using a classification on the basis of volume and shape features, they achieved a sensitivity of 83% and a specificity of 78%. However, they did not assess the classification performance on the basis of volume alone. By using a similar high-dimensional brain mapping approach, Hsu et al (55) compared automated and manual segmentations in AD and cognitively impaired patients. They reported good correlations between manual and automated measurements. However, they did not investigate the accuracy of this technique for the classification of individual patients.

The results of our study require confirmation with larger groups of participants. However, the use of bootstrap resampling techniques allows computing robust estimates for relatively small groups of patients. To keep the control group as large as possible, we decided to not exclude control participants on the basis of age or sex. As a consequence, the mean age of the healthy controls was lower than those of the AD and MCI patients. However, to ensure that our findings were not biased by age or sex confounding effects, we also performed the same analysis on smaller groups of age-matched and sex-matched participants and obtained similar results. Nevertheless, future studies on larger age- and sex-matched groups of participants are required to confirm our results. Finally, in our study, the automated segmentations were performed by only one rater, excluding the ability to evaluate interoperator repeatability. In a previous paper (21), we assessed the interrater reproducibility of the automated segmentation in healthy controls and patients with AD and reported a high reproducibility (the average volume difference between operators was 4% for healthy controls and 7% for AD patients).

By using automated segmentation of the hippocampus, we were able to individually classify Alzheimer disease, mild cognitive impairment, and control participants with a high degree of accuracy. This method can serve as an alternative to manual tracing and may become a useful tool to assist in the diagnosis of Alzheimer disease.


   ADVANCE IN KNOWLEDGE
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCE IN KNOWLEDGE
IMPLICATION FOR PATIENT CARE
 References
 


   IMPLICATION FOR PATIENT CARE
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCE IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
References
 


   FOOTNOTES
 

Abbreviations: AD = Alzheimer disease • MCI = mild cognitive impairment • MMSE = Mini-Mental State Examination • SD = standard deviation

Author contributions: Guarantor of integrity of entire study, O.C.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; literature research, O.C., G.C., M.C., S.L.; clinical studies, G.C., B.Desgranges, B.Dubois, F.E.; statistical analysis, O.C., G.C., B.M., H.B., L.G., S.L.; and manuscript editing, all authors

Authors stated no financial relationship to disclose.


   References
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCE IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
 References
 

  1. Ferri CP, Prince M, Brayne C, et al. Global prevalence of dementia: a Delphi consensus study. Lancet 2005;366:2112–2117.[CrossRef][Medline]
  2. Dubois B, Albert ML. Amnestic MCI or prodromal Alzheimer's disease? Lancet Neurol 2004;3:246–248.[CrossRef][Medline]
  3. Petersen RC. Mild cognitive impairment as a diagnostic entity. J Intern Med 2004;256:183–194.[CrossRef][Medline]
  4. Winblad B, Palmer K, Kivipelto M, et al. Mild cognitive impairment: beyond controversies, towards a consensus—report of the International Working Group on Mild Cognitive Impairment. J Intern Med 2004;256:240–246.[CrossRef][Medline]
  5. Whitehead A, Perdomo C, Pratt RD, Birks J, Wilcock GK, Evans JG. Donepezil for the symptomatic treatment of patients with mild to moderate Alzheimer's disease: a meta-analysis of individual patient data from randomised controlled trials. Int J Geriatr Psychiatry 2004;19:624–633.[CrossRef][Medline]
  6. Ritchie CW, Ames D, Clayton T, Lai R. Metaanalysis of randomized trials of the efficacy and safety of donepezil, galantamine, and rivastigmine for the treatment of Alzheimer disease. Am J Geriatr Psychiatry 2004;12:358–369.[CrossRef][Medline]
  7. Braak H, Braak E. Staging of Alzheimer's disease-related neurofibrillary changes. Neurobiol Aging 1995;16:271–278.[CrossRef][Medline]
  8. Delacourte A, David JP, Sergeant N, et al. The biochemical pathway of neurofibrillary degeneration in aging and Alzheimer's disease. Neurology 1999;52:1158–1165.[Abstract/Free Full Text]
  9. Fox NC, Warrington EK, Freeborough PA, et al. Presymptomatic hippocampal atrophy in Alzheimer's disease: a longitudinal MRI study. Brain 1996;119(pt 6):2001–2007.[Abstract/Free Full Text]
  10. Lehéricy S, Baulac M, Chiras J, et al. Amygdalohippocampal MR volume measurements in the early stages of Alzheimer disease. AJNR Am J Neuroradiol 1994;15:929–937.[Abstract]
  11. Seab JP, Jagust WJ, Wong ST, Roos MS, Reed BR, Budinger TF. Quantitative NMR measurements of hippocampal atrophy in Alzheimer's disease. Magn Reson Med 1988;8:200–208.[Medline]
  12. Jack CR Jr, Petersen RC, O'Brien PC, Tangalos EG. MR-based hippocampal volumetry in the diagnosis of Alzheimer's disease. Neurology 1992;42:183–188.[Abstract/Free Full Text]
  13. Killiany RJ, Moss MB, Albert MS, Sandor T, Tieman J, Jolesz F. Temporal lobe regions on magnetic resonance imaging identify patients with early Alzheimer's disease. Arch Neurol 1993;50:949–954.[Abstract/Free Full Text]
  14. Laakso MP, Partanen K, Riekkinen P, et al. Hippocampal volumes in Alzheimer's disease, Parkinson's disease with and without dementia, and in vascular dementia: an MRI study. Neurology 1996;46:678–681.[Free Full Text]
  15. Jack CR Jr, Petersen RC, Xu YC, et al. Medial temporal atrophy on MRI in normal aging and very mild Alzheimer's disease. Neurology 1997;49:786–794.[Abstract/Free Full Text]
  16. Jack CR Jr, Petersen RC, Xu Y, et al. Rate of medial temporal lobe atrophy in typical aging and Alzheimer's disease. Neurology 1998;51:993–999.[Abstract/Free Full Text]
  17. Laakso MP, Soininen H, Partanen K, et al. MRI of the hippocampus in Alzheimer's disease: sensitivity, specificity, and analysis of the incorrectly classified subjects. Neurobiol Aging 1998;19:23–31.[CrossRef][Medline]
  18. Frisoni GB, Laakso MP, Beltramello A, et al. Hippocampal and entorhinal cortex atrophy in frontotemporal dementia and Alzheimer's disease. Neurology 1999;52:91–100.[Abstract/Free Full Text]
  19. Juottonen K, Laakso MP, Partanen K, Soininen H. Comparative MR analysis of the entorhinal cortex and hippocampus in diagnosing Alzheimer disease. AJNR Am J Neuroradiol 1999;20:139–144.[Abstract/Free Full Text]
  20. Scheltens P, Launer LJ, Barkhof F, Weinstein HC, van Gool WA. Visual assessment of medial temporal lobe atrophy on magnetic resonance imaging: interobserver reliability. J Neurol 1995;242:557–560.[CrossRef][Medline]
  21. Chupin M, Mukuna-Bantumbakulu AR, Hasboun D, et al. Anatomically constrained region deformation for the automated segmentation of the hippocampus and the amygdala: method and validation on controls and patients with Alzheimer's disease. Neuroimage 2007;34:996–1019.[CrossRef][Medline]
  22. Baron JC, Chételat G, Desgranges B, et al. In vivo mapping of gray matter loss with voxel-based morphometry in mild Alzheimer's disease. Neuroimage 2001;14:298–309.[CrossRef][Medline]
  23. Chételat G, Desgranges B, De La Sayette V, Viader F, Eustache F, Baron JC. Mapping gray matter loss with voxel-based morphometry in mild cognitive impairment. Neuroreport 2002;13:1939–1943.[CrossRef][Medline]
  24. Chételat G, Landeau B, Eustache F, et al. Using voxel-based morphometry to map the structural changes associated with rapid conversion in MCI: a longitudinal MRI study. Neuroimage 2005;27:934–946.[CrossRef][Medline]
  25. Desgranges B, Baron JC, Lalevée C, et al. The neural substrates of episodic memory impairment in Alzheimer's disease as revealed by FDG-PET: relationship to degree of deterioration. Brain 2002;125:1116–1124.[Abstract/Free Full Text]
  26. 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;34:939–944.[Abstract/Free Full Text]
  27. Petersen RC, Doody R, Kurz A, et al. Current concepts in mild cognitive impairment. Arch Neurol 2001;58:1985–1992.[Abstract/Free Full Text]
  28. Meguro K, Constans JM, Courtheoux P, Theron J, Viader F, Yamadori A. Atrophy of the corpus callosum correlates with white matter lesions in patients with cerebral ischaemia. Neuroradiology 2000;42:413–419.[CrossRef][Medline]
  29. Ashburner J, Friston KJ. Unified segmentation. Neuroimage 2005;26:839–851.[CrossRef][Medline]
  30. Ashburner J, Friston KJ. Voxel-based morphometry: the methods. Neuroimage 2000;11:805–821.[CrossRef][Medline]
  31. Efron B, Tibshirani RJ. An introduction to the bootstrap. New York, NY: Chapman & Hall, 1993.
  32. Convit A, De Leon MJ, Tarshish C, et al. Specific hippocampal volume reductions in individuals at risk for Alzheimer's disease. Neurobiol Aging 1997;18:131–138.[CrossRef][Medline]
  33. Xu Y, Jack CR Jr, O'Brien PC, et al. Usefulness of MRI measures of entorhinal cortex versus hippocampus in AD. Neurology 2000;54:1760–1767.[Abstract/Free Full Text]
  34. De Santi S, de Leon MJ, Rusinek H, et al. Hippocampal formation glucose metabolism and volume losses in MCI and AD. Neurobiol Aging 2001;22:529–539.[CrossRef][Medline]
  35. Du AT, Schuff N, Amend D, et al. Magnetic resonance imaging of the entorhinal cortex and hippocampus in mild cognitive impairment and Alzheimer's disease. J Neurol Neurosurg Psychiatry 2001;71:441–447.[Abstract/Free Full Text]
  36. Killiany RJ, Hyman BT, Gomez-Isla T, et al. MRI measures of entorhinal cortex vs hippocampus in preclinical AD. Neurology 2002;58:1188–1196.[Abstract/Free Full Text]
  37. Pennanen C, Kivipelto M, Tuomainen S, et al. Hippocampus and entorhinal cortex in mild cognitive impairment and early AD. Neurobiol Aging 2004;25:303–310.[CrossRef][Medline]
  38. Tapiola T, Pennanen C, Tapiola M, et al. MRI of hippocampus and entorhinal cortex in mild cognitive impairment: a follow-up study. Neurobiol Aging 2008;29:31–38.[CrossRef][Medline]
  39. Jack CR Jr, Petersen RC, Xu YC, et al. Prediction of AD with MRI-based hippocampal volume in mild cognitive impairment. Neurology 1999;52:1397–1403.[Abstract/Free Full Text]
  40. Devanand DP, Pradhaban G, Liu X, et al. Hippocampal and entorhinal atrophy in mild cognitive impairment: prediction of Alzheimer disease. Neurology 2007;68:828–836.[Abstract/Free Full Text]
  41. Convit A, de Asis J, de Leon MJ, Tarshish CY, De Santi S, Rusinek H. Atrophy of the medial occipitotemporal, inferior, and middle temporal gyri in non-demented elderly predict decline to Alzheimer's disease. Neurobiol Aging 2000;21:19–26.[Medline]
  42. Jack CR Jr, Shiung MM, Weigand SD, et al. Brain atrophy rates predict subsequent clinical conversion in normal elderly and amnestic MCI. Neurology 2005;65:1227–1231.[Abstract/Free Full Text]
  43. DeCarli C, Frisoni GB, Clark CM, et al. Qualitative estimates of medial temporal atrophy as a predictor of progression from mild cognitive impairment to dementia. Arch Neurol 2007;64:108–115.[Abstract/Free Full Text]
  44. Geroldi C, Rossi R, Calvagna C, et al. Medial temporal atrophy but not memory deficit predicts progression to dementia in patients with mild cognitive impairment. J Neurol Neurosurg Psychiatry 2006;77:1219–1222.[Abstract/Free Full Text]
  45. de Leon MJ, Golomb J, George AE, et al. The radiologic prediction of Alzheimer disease: the atrophic hippocampal formation. AJNR Am J Neuroradiol 1993;14:897–906.[Abstract]
  46. Kelemen A, Szekely G, Gerig G. Elastic model-based segmentation of 3-D neuroradiological data sets. IEEE Trans Med Imaging 1999;18:828–839.[CrossRef][Medline]
  47. Yang J, Duncan JS. 3D image segmentation of deformable objects with joint shape-intensity prior models using level sets. Med Image Anal 2004;8:285–294.[CrossRef][Medline]
  48. Pitiot A, Delingette H, Thompson PM, Ayache N. Expert knowledge-guided segmentation system for brain MRI. Neuroimage 2004;23(suppl 1):S85–S96.[CrossRef][Medline]
  49. Shen D, Moffat S, Resnick SM, Davatzikos C. Measuring size and shape of the hippocampus in MR images using a deformable shape model. Neuroimage 2002;15:422–434.[CrossRef][Medline]
  50. Duchesne S, Pruessner J, Collins DL. Appearance-based segmentation of medial temporal lobe structures. Neuroimage 2002;17:515–531.[CrossRef][Medline]
  51. Hogan RE, Mark KE, Wang L, Joshi S, Miller MI, Bucholz RD. Mesial temporal sclerosis and temporal lobe epilepsy: MR imaging deformation-based segmentation of the hippocampus in five patients. Radiology 2000;216:291–297.[Abstract/Free Full Text]
  52. Carmichael OT, Aizenstein HA, Davis SW, et al. Atlas-based hippocampus segmentation in Alzheimer's disease and mild cognitive impairment. Neuroimage 2005;27:979–990.[CrossRef][Medline]
  53. Fischl B, Salat DH, Busa E, et al. Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron 2002;33:341–355.[CrossRef][Medline]
  54. Csernansky JG, Wang L, Joshi S, et al. Early DAT is distinguished from aging by high-dimensional mapping of the hippocampus: dementia of the Alzheimer type. Neurology 2000;55:1636–1643.[Abstract/Free Full Text]
  55. Hsu YY, Schuff N, Du AT, et al. Comparison of automated and manual MRI volumetry of hippocampus in normal aging and dementia. J Magn Reson Imaging 2002;16:305–310.[CrossRef][Medline]



This article has been cited by other articles:


Home page
 BrainHome page
R. S. Desikan, H. J. Cabral, C. P. Hess, W. P. Dillon, C. M. Glastonbury, M. W. Weiner, N. J. Schmansky, D. N. Greve, D. H. Salat, R. L. Buckner, et al.
Automated MRI measures identify individuals with mild cognitive impairment and Alzheimer's disease
Brain, May 21, 2009; (2009) awp123v1.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
2481070876v1
248/1/194    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Colliot, O.
Right arrow Articles by Lehéricy, S.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Colliot, O.
Right arrow Articles by Lehéricy, S.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
RADIOLOGY RADIOGRAPHICS RSNA JOURNALS ONLINE