Combination of MRI hippocampal volumetry and arterial spin labeling MR perfusion at 3-Tesla improves the efficacy in discriminating Alzheimer's disease from cognitively normal elderly
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Unc
orre
cted
Aut
hor P
roof
Journal of Alzheimer’s Disease xx (20xx) x–xxDOI 10.3233/JAD-131868IOS Press
1
Combination of MRI HippocampalVolumetry and Arterial Spin Labeling MRPerfusion at 3-Tesla Improves the Efficacyin Discriminating Alzheimer’s Diseasefrom Cognitively Normal Elderly Adults
1
2
3
4
5
Henry Ka-Fung Maka,f,∗, Wenshu Qiana, Kwok Sing Ngb, Queenie Chanc, You-Qiang Songd,f ,Leung Wing Chue,f and Kelvin Kai-Wing Yaug
6
7
aDepartment of Diagnostic Radiology, The University of Hong Kong, Hong Kong SAR, China8
bLi Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong SAR, China9
cPhilips Healthcare, Hong Kong SAR, China10
dDepartment of Biochemistry, The University of Hong Kong, Hong Kong SAR, China11
eDepartment of Medicine, The University of Hong Kong, Hong Kong SAR, China12
f Alzheimer’s Disease Research Network, The University of Hong Kong, Hong Kong SAR, China13
gDepartment of Management Sciences, City University of Hong Kong, Hong Kong SAR, China14
Handling Associate Editor: J. Wesson Ashford
15
Accepted 18 February 2014
Abstract.16
Background: Structural magnetic resonance imaging has been employed for evaluation of medial temporal atrophy in patientswith Alzheimer’s disease (AD). Arterial spin labeling (ASL) technique could detect cerebral perfusion abnormalities in AD.
17
18
Objective: We hypothesized that combination of hippocampal volumetry and cerebral blood flow yield higher accuracy thaneither method alone in discriminating AD patients from cognitively normal older adults.
19
20
Materials and Methods: 13 AD patients and 15 healthy controls were studied using a 3-tesla scanner. Standardized T1W 3Dvolumetric Fast Field Echo and QUASAR ASL sequences were employed for cerebral volumetry and perfusion respectively.Manual right and left hippocampal volumetry was performed by ANALYZE software, with total intracranial volume normaliza-tion. ASL data were analyzed by institutional specially-design software to calculate cerebral blood flow of region-of-interestsplaced at the middle and posterior cingulate gyri.
21
22
23
24
25
Results: Right and left hippocampal volumes and middle and posterior cingulate gyri cerebral blood flows were significantlylower in the patients than in the controls (independent-samples t-tests, p < 0.05), and prediction accuracies of 89.3%, 82.1%,75.0% and 71.4% were achieved for each of the above parameters, respectively. In distinguishing patients from controls usingcorresponding optimized cut-off values, various combinations of these parameters were used to create the Receiver OperatingCharacteristic curves. The highest area under curve value was 0.944, by combining cerebral blood flow at the middle cingulategyrus, normalized right and left hippocampal volumes.
26
27
28
29
30
31
Conclusions: A ‘one-stop-shop’ magnetic resonance study of combined hippocampal volumetry and cerebral perfusion hasimproved efficacy in discriminating AD patients from cognitively normal older adults.
∗Correspondence to: Henry Ka-Fung Mak, Department of Diag-nostic Radiology, The University of Hong Kong, Room 406, BlockK, Queen Mary Hospital, 102 Pokfulam Road, Hong Kong. Tel.:+852 28170373; Fax: +852 28174013; E-mail: [email protected].
Fig. 1. Single parameter: CBF of middle and posterior cingulate gyriand the normalized volume ratios (percentage*100) of right and lefthippocampi.
The highest AUC value was 0.944, by combining275
normalized RH, LH, and MC CBF; AUC value of indi-276
vidual parameter such as normalized RH, normalized277
LH, MC CBF, and PC CBF was 0.928, 0.872, 0.749278
and 0.744, respectively. (Figs. 1–3, Table 3)279
Using corresponding optimized cut-off values, pre-280
diction accuracies of 89.3%, 82.1%, 75.0% and 71.4%281
for RH volumetry, LH volumetry, MC CBF, and PC282
CBF, respectively, were achieved in distinguishing AD283
from CN. Combination of 2, 3 or 4 of these parame-284
ters achieved accuracies ranging from 82.1% to 89.3%285
(Table 3).286
In the entire cohort of subjects plus controls, there287
were statistically significant correlations (p < 0.05)288
of the clinically observed severity measurements289
Fig. 2. Combining two parameters. MR: M cingulate with rVol RH;ML: M cingulate with rVol LH; PR: P cingulate with rVol RH; PL:P cingulate with rVol LH.
(MMSE, ADAS-cog, and SI) with brain scan param- 290
eters (normalized RH and LH volumes, MC and PC 291
CBF). For example, Pearson correlation between SI 292
and the brain scan measurements ranged from 0.502 293
to 0.646 (Table 4). Such strong correlations were also 294
demonstrated across diagnostic groups (Fig. 4). 295
Applying multiple regression analysis to the entire 296
cohort, we showed that combination of brain scan 297
parameters such as MC CBF, RH and LH volumes 298
achieved higher significance in prediction of clinically 299
observed severity measurements (Table 4). 300
DISCUSSION 301
Our current study evaluated the diagnostic per- 302
formance of manual hippocampal volumetry versus 303
Unc
orre
cted
Aut
hor P
roof
6 H.K.-F. Mak et al. / MRI Volumetry & ASL in AD and Normal Controls
Fig. 3. Combining three or more parameters. MRL: M cingulatewith rVol RH and rVol LH; PRL: P cingulate with rVol RHand rVol LH; MPR: M&P cingulate with rVol RH; MPL: M&Pcingulate with rVol LH; MPRL: M&P cingulate with rVol RH andrVol LH.
pulsatile ASL derived CBF, and their combination in304
discriminating a cohort of AD from CN. Our findings305
confirmed the diagnostic roles of these MRI techniques306
and highlighted the emerging role of ASL CBF.307
First, our current study employed a smaller cohort 308
of 13 AD and 15 CN [11] than our published study 309
[4] to allow comparison between manual hippocampal 310
volumetry and ASL CBF methods. In a prior study of 311
a larger cohort of 19 AD and 20 CN, the prediction 312
accuracy and AUC were 87.2% and 0.94 and 84.6% 313
and 0.93 for normalized RH and LH volume, respec- 314
tively. Similar diagnostic efficacies were achieved in 315
current study, with prediction accuracy and AUC being 316
89.3% and 0.928 and 82.1% and 0.872 for normalized 317
RH and LH volume, respectively. These findings vali- 318
dated the robust discriminatory power of hippocampal 319
morphometry. 320
Second, pulsatile ASL MRI was found in our study 321
to be inferior to hippocampal volumetry in diagnos- 322
tic accuracy, with prediction accuracy and area under 323
ROC curve being 75% and 0.749 and 71.4% and 0.744 324
for MC and PC CBF. A previous study using QUASAR 325
ASL [10] showed significant differences of CBF in the 326
PC and precunei. Based on the mean CBF in these 327
ROIs, the study achieved 73.9% sensitivity, 80.0% 328
specificity, and 76.7% accuracy in discriminating 20 329
AD and 23 CN, similar to our respective results of 330
61.5%, 86.7% and 74.9% (based on CBF in the MC 331
gyrus). A recent study [18] employed ‘qualitative’ 332
evaluations of continuous ASL and high resolution 333
T1W images by four blinded experienced readers and 334
found that perfusion abnormalities were superior to 335
structural volume losses in early detection of AD, 336
Fig. 4. Scatter-plot of middle and posterior cingulate CBF, and normalized right and left hippocampal volumes against the severity index.
Unc
orre
cted
Aut
hor P
roof
H.K.-F. Mak et al. / MRI Volumetry & ASL in AD and Normal Controls 7
Table 3The sensitivity, specificity, and accuracy results of the single parameter versus combined parameter models, optimal cut-offs of various parameters
based on Youden index, and AUC of the Receiver Operating Characteristic (ROC) curves of various combinations of parameters
cut-off Sensitivity Specificity Youden index Accuracy AUCmean 95% CI
confirmed that the diagnostic probability of AD, its411
phenotypes, and cognitively normal elderly controls412
lie along a continuum and the brain scan measurements413
added certainty to the diagnostic process.414
Finally, in real-life clinical situation, we found that415
hippocampal volumetry is practical in differentiating416
early AD from normal aging as it can be implemented417
within 20 minutes in experienced hands [4]. The cut-418
off values of normalized RH and LH volumes in our419
Chinese cohort could be employed as reference for420
differentiation of the two groups. However, MRI struc-421
tural changes in AD occurred later than FDG-PET422
changes [24]. In combining both techniques in the423
same setting, the additional discriminatory effect of424
decreased CBF can be exploited for improvement in425
radiologic diagnosis [23]. The color-coded maps of426
CBF of a normal control and an AD patient (being427
an average in each diagnostic group in our study) were428
shown (Fig. 5), and the quality of the ASL MRI images429
allowed visual (qualitative) evaluation in the clinical430
setting [18].431
The major limitation of our study is a relatively432
small cohort. Also, we were unable to demonstrate433
in our cohort the superior sensitivity of the parietal434
Fig. 5. The CBF color-coded maps of a healthy control (72F) andan AD patient (78F, MMSE = 15), selected at two typical transversesections: level of basal ganglia and level of centrum semiovale.
association cortices in detecting CBF abnormalities, 435
as published in prior studies [22, 23]. Our unpub- 436
lished data (scatter-plot of CBF in the precuneus 437
of AD and healthy controls plotted against the SI 438
and MMSE) showed that a few controls demonstrate 439
impaired precuneus CBF, and one even had lower 440
CBF than AD subjects. These outliers, probably due 441
to technical CBF measurement variability or early 442
AD, might explain the insignificant result of precuneus 443
CBF. The pulsatile ASL technique has the drawback 444
of limited signal-to-noise, thereby affecting data qual- 445
ity and required cleaning in a recent study [25]. The 446
new pseudo-continuous approach holds promise of a 447
high efficiency and broad compatibility MRI perfusion 448