Neuroradiology
Original Research
Hippocampal Volume and the Mini-Mental State Examination in the Diagnosis of Amnestic Mild Cognitive Impairment
Audio Available | 
Share
OBJECTIVE. The purpose of this study was to examine the diagnostic efficacy of hippocampal volumetry and the Mini-Mental State Examination (MMSE) in differentiating amnestic mild cognitive impairment from the normal changes of aging.
SUBJECTS AND METHODS. The conditions of healthy older persons (n = 17) and persons with amnestic mild cognitive impairment (n = 18) were classified on the basis of results of comprehensive neuropsychological assessment. All subjects underwent MRI at 4 T, including high-resolution coronal T1-weighted images. Hippocampal volume was calculated by tracing right and left hippocampal formations on coronal slices and normalizing by single-slice intracranial area. Receiver operating characteristic analysis and logistic regression analysis were performed to evaluate the diagnostic efficacy of hippocampal volume and the MMSE in differentiating subjects with amnestic mild cognitive impairment from subjects with normal cognitive function.
RESULTS. The mean ± SE area under the receiver operating characteristics curve (AUC) for left hippocampal volume (0.886 ± 0.056) was greater than that for right hippocampal volume (0.614 ± 0.097). The AUC for MMSE score (0.745 ± 0.085) was intermediate and not statistically different from that of hippocampal volume measurements alone. The AUC for the combination of left hippocampal volume and MMSE score (0.92) was significantly greater than that of MMSE score alone (p < 0.05).
CONCLUSION. In the diagnosis of amnestic mild cognitive impairment, use of left hippocampal volume may be more efficacious than use of right hippocampal volume and may add to the value of routine screening MMSE. Automated hippocampal volumetry may become a useful diagnostic adjunct in settings in which sophisticated neuropsychological testing is not readily available.
Keywords: Alzheimer's disease, brain, high-resolution, hippocampal volume, Mini-Mental State Examination, MRI, neuroimaging
| Introduction |   |
|---|
Amnestic mild cognitive impairment is a condition that represents increased risk of development of Alzheimer's disease. Many experts [1] consider it a prodromal stage of Alzheimer's disease. The early pathologic changes of Alzheimer's disease, especially in the medial temporal lobe (MTL), may begin years before clinical symptoms, and this region retains the greatest density of neurofibrillary tangles as the disease progresses [2]. The role of the MTL, particularly the hippocampus, in episodic memory [3] has led a number of researchers to target MTL structures in the search for in vivo imaging markers of amnestic mild cognitive impairment. Numerous studies [4] have shown greater atrophy of the hippocampus in amnestic mild cognitive impairment and Alzheimer's disease, and lower baseline hippocampal volumes are known to be predictive of decline to Alzheimer's disease. Hippocampal asymmetry has been reported to be present in older adults with subjective memory symptoms [5], mild cognitive impairment [6], and Alzheimer's disease [7]. This finding suggests that age and degenerative processes do not necessarily affect the brain equally across hemispheres.
In primary care, screening for memory deficits and dementia relies largely on results of brief cognitive tests, such as the Mini-Mental State Examination (MMSE). The MMSE [8] is used to test the major domains of cognitive function, including language, visuospatial ability, and memory. Out of a maximum score of 30, persons with multimodal functional deficits, such as those of Alzheimer's disease, typically score less than 24 [9]. However, among persons in the preclinical stages of Alzheimer's disease, such as those with amnestic mild cognitive impairment, MMSE scores are less sensitive, tending to fall within the normal range (3 24) [10]. As a nonspecific screening instrument, the MMSE is perhaps more appropriate for directing clinicians toward more comprehensive diagnostic tests. Such tests, although currently necessary to assess for amnestic mild cognitive impairment and probable Alzheimer's disease, often are not performed in routine practice. The purpose of this study was to assess the diagnostic efficacy of hippocampal volumetry alone in combination with the MMSE in differentiating amnestic mild cognitive impairment from the normal changes of aging.
| Subjects and Methods | |
|---|
The study received approval from the Duke University Medical Center institutional review board and was conducted in compliance with the Health Insurance Portability and Accountability Act. All subjects provided written informed consent before testing and neuropsychological evaluation.
Subjects were recruited as part of a large functional MRI study. They responded to advertisements asking for persons 55-85 years old with memory problems. Our study group consisted of the first 37 subjects (20 persons with amnestic mild cognitive impairment and 17 control subjects) who underwent imaging. One subject with amnestic mild cognitive impairment was excluded because of imaging acquisition error, and a second was excluded because of marked base-of-skull artifact. The other 35 subjects were 18 persons with amnestic mild cognitive impairment and 17 controls.
Subjects were limited to those between the ages of 55 and 85 years and those free of contraindications to MRI, including claustrophobia and presence of metallic objects, such as cardiac pacemakers, aneurysm clips, and foreign bodies in the eyes. The 18 subjects with amnestic mild cognitive impairment (mean age, 74.4 ± 7.7 years; age range, 55.5-83.0 years) tended to be slightly older than the 17 controls (mean age, 70.2 ± 3.6 years; age range, 63.3-78.0 years), but the difference was not statistically significant (p > 0.05).
All subjects underwent baseline psychometric evaluation, which included the MMSE; logical memory and visual reproduction from the Wechsler Memory Scale, third edition [11]; the California Verbal Learning Test, second edition [12]; and the Beck Depression Inventory [13].
Subjects were classified as having amnestic mild cognitive impairment if they did not meet the criteria for dementia defined by National Institute of Neurological and Communicative Diseases and the Stroke/Alzheimer's Disease and Related Disorders Association [14] or by the Diagnostic and Statistical Manual of Mental Disorders, fourth edition [15]; if they had impairment (≥ 1 SD below age-adjusted norms) in delayed verbal or visual recall; if they had an MMSE score of 24 or higher; if they had normal independent functioning as defined by a clinical dementia rating of 0.5 (questionable dementia) [16]; and if they did not have other factors that might better explain memory loss (e.g., current major depression). It was required that the study clinicians reach consensus on each subject's diagnosis. Subjects were classified as controls if they had an MMSE score of 28 higher, did not meet the aforementioned criteria for dementia, had normal or near normal independent function, and had a clinical dementia rating global score of zero and normal memory.
Imaging was performed with a 4-T MRI system (Signa LX 8.0 NVi, GE Healthcare) equipped with 41-mT/m gradients, a birdcage radiofrequency transmitter, and a receiver head coil. Axial T2-weighted spin-echo images (TR/TE 3,000/80; matrix size, 256 × 256; slice thickness, 3.75 mm; interslice gap, 0.0 mm; field of view, 240 mm) were obtained for anatomic screening purposes and were reviewed by a neuroradiologist for significant intracranial abnormalities. For directed hippocampal volumetric analysis, 128 contiguous T1-weighted images (3D inversion recovery prepare fast spoiled gradient-recalled acquisition in the steady state; 12.2/5.4; inversion time, 300 milliseconds; flip angle, 20°; field of view, 240 mm; matrix size, 256 × 256; slice thickness, 1.5 mm) through the entire brain were obtained in the oblique coronal plane perpendicular to the long axis of the hippocampus.
Mouse-driven software (ITK-SnAP 1.0, Cognitica Corporation) was used to make manual hippocampal tracings on the contiguous oblique coronal T1-weighted images. This program is included in the open National Library of Medicine Insight Segmentation and Registration Toolkit. It was designed as a user-friendly interface for displaying medical images in three planes simultaneously, for segmenting level threshold anatomic structures, and for manual editing of the segmentations. Hippocampal tracings were performed by one investigator, who was blinded to clinical classification.
Each hippocampus was traced independently with reference to previous hippocampal MR volumetric studies [4, 17] and the hippocampal atlas by Duvernoy [18]. The hippocampus was first identified at the most posterior portion where the hippocampal tail became visible under the fornix and proceeded on contiguous slices to the most anterior slice on which the hippocampus could be delineated from the amygdala. The selected area was reviewed for consistency on the axial and sagittal representations. The traced region encompassed the subiculum, regions CA1-4, and the internal and external digitations. The fimbria lining the dorsal aspect of the hippocampus and the vertical digitations and semilunar gyrus extending superiorly toward the amygdala were omitted. Each hippocampus spanned approximately 20-25 slices in each subject (Fig. 1A, 1B, 1C, 1D, 1E, 1F, 1G, 1H, 1I, 1J, 1K, 1L).
After manual tracing was completed, the number of voxels included was calculated. For simplicity, we did not convert voxel counts to volume (1 voxel = 1.318 mm3), and volume was reported as voxel count. Intrarater reliability in six cases (three subjects with amnestic mild cognitive impairment and three controls) was calculated for the single experimenter. The kappa values were 0.91 and 0.88 for the right and left raw hippocampal voxel counts. For calculation of interrater reliability, images of another six randomly selected subjects (three with amnestic mild cognitive impairment and three controls) were traced by a second observer. The kappa values were 0.90 and 0.85 for the right and left raw hippocampal voxel counts.
Normalization of right and left hippocampal voxel counts was accomplished by division of raw counts by intracranial area, also measured in voxels. This value was calculated for each subject from a manually traced single-slice intracranial area as measured on the oblique coronal T1-weighted image at the level of the anterior commissure. This method has been described by Laakso et al. [19] as an appropriate index for intracranial volume correction. Intrarater reliability for this measurement of intracranial area voxel count was greater than 0.99. Interrater reliability for intracranial area voxel count was considerably lower at 0.79, possibly because the template used was a single slice, making it more challenging to reproduce across different brains. Dividing the right and left hippocampal voxel counts by intracranial area resulted in the normalized values.
Hippocampal volume and cognitive test scores decrease with age regardless of associated cognitive decline [20]. Results of linear regression indicated that left and right hippocampal volumes varied significantly with age (p < 0.05). Age correction, therefore, was conducted for the right and left normalized hippocampal voxel counts and for the raw MMSE scores according to the following equation: 
where valuen is the right or left normalized hippocampal voxel count or raw MMSE score; agesubj is the age of the subject; agemean is the mean age of all subjects in the study; and Bage is the slope calculated by linear regression for each volume or test score with age as the independent variable. The result was three measures corrected for age for each subject: right hippocampal volume, left hippocampal volume, and MMSE score.
 View larger version (217K) | Fig. 1A —71-year-old man in normal health. Oblique coronal 1.5-mm MR images show bilateral hippocampal formations (crosshatching). For conciseness, only alternating slices are shown. |
 View larger version (215K) | Fig. 1B —71-year-old man in normal health. Oblique coronal 1.5-mm MR images show bilateral hippocampal formations (crosshatching). For conciseness, only alternating slices are shown. |
 View larger version (216K) | Fig. 1C —71-year-old man in normal health. Oblique coronal 1.5-mm MR images show bilateral hippocampal formations (crosshatching). For conciseness, only alternating slices are shown. |
 View larger version (216K) | Fig. 1D —71-year-old man in normal health. Oblique coronal 1.5-mm MR images show bilateral hippocampal formations (crosshatching). For conciseness, only alternating slices are shown. |
 View larger version (218K) | Fig. 1E —71-year-old man in normal health. Oblique coronal 1.5-mm MR images show bilateral hippocampal formations (crosshatching). For conciseness, only alternating slices are shown. |
 View larger version (217K) | Fig. 1F —71-year-old man in normal health. Oblique coronal 1.5-mm MR images show bilateral hippocampal formations (crosshatching). For conciseness, only alternating slices are shown. |
 View larger version (218K) | Fig. 1G —71-year-old man in normal health. Oblique coronal 1.5-mm MR images show bilateral hippocampal formations (crosshatching). For conciseness, only alternating slices are shown. |
 View larger version (218K) | Fig. 1H —71-year-old man in normal health. Oblique coronal 1.5-mm MR images show bilateral hippocampal formations (crosshatching). For conciseness, only alternating slices are shown. |
 View larger version (218K) | Fig. 1I —71-year-old man in normal health. Oblique coronal 1.5-mm MR images show bilateral hippocampal formations (crosshatching). For conciseness, only alternating slices are shown. |
 View larger version (216K) | Fig. 1J —71-year-old man in normal health. Oblique coronal 1.5-mm MR images show bilateral hippocampal formations (crosshatching). For conciseness, only alternating slices are shown. |
 View larger version (216K) | Fig. 1K —71-year-old man in normal health. Oblique coronal 1.5-mm MR images show bilateral hippocampal formations (crosshatching). For conciseness, only alternating slices are shown. |
 View larger version (213K) | Fig. 1L —71-year-old man in normal health. Oblique coronal 1.5-mm MR images show bilateral hippocampal formations (crosshatching). For conciseness, only alternating slices are shown. |
Repeated-measures analysis of covariance was calculated with side (right and left normalized hippocampal voxel counts) as the within-subject factor and group (amnestic mild cognitive impairment or control) as the between-subject factor. Age was the covariate. A post hoc paired-samples Student's t test of right and left normalized hippocampal voxel counts was performed separately for subjects with amnestic mild cognitive impairment and control subjects.
Receiver operating characteristic (ROC) curves were generated for comparison of age-adjusted normalized hippocampal volumes with MMSE score. Area under the ROC curve (AUC) and SE were calculated for each ROC curve. ROC curves depict the sensitivity (ability to indicate presence of disease when disease is truly present) and 1 - specificity (ability to exclude disease when truly absent) of a test at each possible cutoff point [21]. AUC reflects overall test accuracy, but unlike measures of accuracy (sum of true-positive results and true-negative results divided by study population), which depend on disease prevalence, AUC is independent of prevalence.
Binary logistic regression analysis was performed to investigate the independence of contributions of left hippocampal volume and MMSE score to the diagnosis of amnestic mild cognitive impairment. The regression equation was reconstructed and applied to determine the probability that a given subject would have amnestic mild cognitive impairment on the basis of that subject's MMSE and left hippocampal volume with optimal weighting of the two test results. The ROC curve (left hippocampal volume plus MMSE score) was based on these predicted probabilities.
The AUC for MMSE score was compared with those of right hippocampal volume, left hippocampal volume, and left hippocampal volume plus MMSE score. The method described by Hanley and McNeil [22] was used to examine the significance of differences in AUC. This calculation was performed for each of the following comparisons: right hippocampal volume versus MMSE score, left hippocampal volume versus MMSE score, combination of left hippocampal volume and MMSE versus MMSE score.
| Results | |
|---|
The amnestic mild cognitive impairment and control groups did not differ in sex, age, or education (Table 1). As expected, the score on the delayed-recall measure of the California Verbal Learning Test among subjects with amnestic mild cognitive impairment (5.2 ± 2.4) differed significantly from that of controls (10.8 ± 2.5) (p < 0.001). Group differences in MMSE score (amnestic mild cognitive impairment group, 26.8 ± 1.3; controls, 28.2 ± 1.3) also reached significance (p <0.05).
Analysis of covariance showed a significant interaction between brain side and group. Results of post hoc Student's t tests revealed that mean left hippocampal volume was significantly smaller in the subjects with amnestic mild cognitive impairment than in controls (p < 0.01) (Table 2). In subjects with amnestic mild cognitive impairment, the left hippocampus (mean, 0.131) was significantly more atrophied than the right (mean, 0.142) (p < 0.01). By contrast, controls had comparable right (mean, 0.151) and left (mean, 0.154) volumes (p > 0.05).
 View larger version (14K) | Fig. 2 —Graph shows receiver operating characteristic (ROC) curves of predicted probability of amnestic mild cognitive impairment based on hippocampal volumes compared with Mini-Mental State Examination (MMSE) score. Solid line indicates area under ROC curve (AUC) of 0.886 for left hippocampal volume; dashed line, AUC of 0.745 for MMSE; dotted line, AUC of 0.614 for right hippocampal volume. |
 View larger version (12K) | Fig. 3 —Graph shows receiver operating characteristic (ROC) curve of predicted probability of amnestic mild cognitive impairment calculated from combined left hippocampal volume and Mini-Mental State Examination (MMSE) score compared with predicted probability based on MMSE score alone. Solid line indicates area under ROC curve (AUC) of 0.918 for combination of left hippocampal volume and MMSE; dotted line, AUC of 0.745 for MMSE. |
Hippocampal volume versus MMSE score—ROC curves for comparison of sensitivity and 1 - specificity of hippocampal measurements and psychometric scores are shown in Figures 2 and 3. Corresponding AUC data are shown in Table 3. Of the hippocampal measurements, left hippocampal volume (AUC, 0.886) was the most accurate for classification of amnestic mild cognitive impairment. This value, however, was not statistically more accurate than MMSE score (AUC, 0.745) (p = 0.147). Right hippocampal volume (AUC, 0.614) was a poor indicator of clinical classification, lower even than MMSE score, although the difference was not statistically significant (Fig. 2).
Combinination of left hippocampal volume with MMSE score—The ROC curve generated from the combination of left hippocampal volume and MMSE score in logistic regression (AUC, 0.918) (Fig. 3) showed a statistically significant difference in AUC from the ROC curve of MMSE alone (p < 0.05). The optimal value for differentiating subjects with amnestic mild cognitive impairment from controls was 0.53.
The logistic regression equation was used to calculate probability of classification for each subject. The threshold for a group was set at 0.50. Subjects with a calculated probability of 0.50 or greater were classified as controls and those with a calculated probability less than 0.50 as having amnestic mild cognitive impairment. Use of the MMSE led to correct classification of 11 of 18 subjects with amnestic mild cognitive impairment and 11 of 17 controls (sensitivity, 0.61; specificity, 0.65) (Table 4). After the addition of left hippocampal volume to MMSE score in the logistic regression equation, classification improved to 15 of 18 subjects with amnestic mild cognitive impairment and 14 of 17 controls (sensitivity, 0.83; specificity, 0.82) (Table 4).
| Discussion | |
|---|
We tested the diagnostic utility of hippocampal volumetry in the assessment of amnestic mild cognitive impairment. In accordance with results of previous studies [4-7, 17], our finding was that subjects with amnestic mild cognitive impairment had more hippocampal atrophy than healthy subjects and that hippocampal volume decreased significantly with increasing age. We also found that left hippocampal volume had more discriminatory value than right hippocampal volume. The addition of left hippocampal volume increased both the sensitivity and specificity of the MMSE to more than 0.80, a level that would be clinically acceptable if replicated in a broader sample.
Although measures of left hippocampal atrophy in this study helped differentiate subjects with amnestic mild cognitive impairment from healthy older adults, group distributions of left hippocampal volume nevertheless overlap. As a result, six subjects would still be misclassified after MMSE score and left hippocampal volume were combined. There may be several explanations for these normal variations, including the presence of silent preclinical disease and measurement error. This finding suggests the continued need for additional measures such as neuropsychological testing and other imaging markers. The MMSE is a simple screening tool and is not intended for use as a diagnostic test, although it often is used as such in primary care. This practice reinforces the need to find simple markers that can supplement the MMSE in the detection of disease in its early stages.
We found more left than right hippocampal atrophy in subjects with amnestic mild cognitive impairment. Hippocampal asymmetry has been reported frequently, not only among persons with pathologic conditions but also in the healthy population [7, 23] and among older adults with subjective memory symptoms [5]. Studies have shown a change in persons with asymmetry in amnestic mild cognitive impairment and Alzheimer's disease. This change suggests a preferentially unilateral change responsible for cognitive impairment. The laterality of this asymmetry, however, has varied between greater atrophy on the right side [7] and greater atrophy on the left side [6]. Our subjects with amnestic mild cognitive impairment were identified largely because of verbal delayed recall deficits, which in previous studies [24] have been found to correlate with left hippocampal volume. Petersen et al. [25] reported that the volume of the left hippocampus in persons with Alzheimer's disease was predictive of performance on a verbal learning task. Right hippocampal volume, however, was more closely associated with performance on a nonverbal task of visual representation.
There were limitations to this study, such as small sample size, cross-sectional design, lower interrater reliability regarding intracranial area, and lack of disease controls (e.g., depression, other forms of dementia). Furthermore, although 4-T magnets may not be common in clinical practice, we used an ultra-high-field imaging unit to acquire high-resolution images so that we could examine which measure was most sensitive to disease state in a group without having to acquire onerously large numbers of subjects. Our findings can now be confirmed with lower-resolution images. Although hippocampal volumetric studies have been conducted with variable sample sizes, in few studies has the relation between hippocampal volume and MMSE score been examined from a practical diagnostic point of view. Our findings must be interpreted in this context.
Manual tracing of the hippocampi is labor intensive and operator dependent. Thus an automated protocol for hippocampal volumetry would be more practical in a clinical setting. Warping and segmentation algorithms for automated volumetry of the hippocampus have been evaluated in subjects with Alzheimer's disease and amnestic mild cognitive impairment [26]. Such algorithms, however, usually require manual input, thus observer bias is not completely removed. Moreover, the accuracy of automated methods may suffer when a substantial amount of image noise or poor contrast enhancement is present.
The findings of our study suggest that the combination of left hippocampal volume and MMSE score may have reasonably high diagnostic accuracy for separating amnestic mild cognitive impairment from the normal changes of aging. Larger studies with automated volumetry in representative samples are needed to replicate the findings. Future studies also should examine the value of adding hippocampal volumetry to more sensitive neuropsychological measures.
Address correspondence to J. R. Petrella.
Supported by National Institute on Aging grant R01AG019728.
FOR YOUR INFORMATION
The comprehensive book based on the ARRS 2007 annual meeting categorical course on Neuroradiology is now available. For more information or to purchase a copy, see www.arrs.org.
We thank the subjects who participated in this study.
| References | |
|---|
