Keywords Brain hypometabolism - dementia - neurodegeneration - positron emission tomography/magnetic
resonance imaging - semi-quantitative analysis
Introduction
Hybrid positron emission tomography/magnetic resonance imaging (PET/MRI) provides
an effective modality in the clinical evaluation of cognitively impaired patients.
Dedicated high-resolution brain MR sequences provide quantifiable information that
can be used to aid in diagnosis. Decreased radiation exposure with PET/MRI and improved
motion correction functionality result in improved PET image quality and more accurate
anatomic localization compared to PET/CT. In addition, simultaneous PET/MRI scanners
allow for images to be obtained in a single, convenient session, minimizing patient
discomfort. These advances in technology provide newer and more effective ways to
evaluate patients with cognitive impairment and suspected underlying neurodegenerative
disease.[1 ],[2 ],[3 ]
Neurodegenerative disorders typically manifest with a myriad of clinical symptoms,
including memory loss, impaired higher-order cognitive functioning, and some degree
of corresponding brain volume loss – all of which progress following diagnosis.[4 ] The most common form of dementia is Alzheimer's disease (AD) followed by frontotemporal
lobar degeneration (FTLD), dementia with Lewy bodies (LBD), and other rare neurodegenerative
disorders such as corticobasal degeneration (CBD), which are far less common.[5 ],[6 ] Early diagnosis is essential in patients with suspected dementia as early intervention
can slow progression and improve quality of life. As per the National Institute for
Neurologic and Communicative Disorders and Related Disorders Association, when tissue
sampling is not feasible, imaging and CSF biomarkers should be used for the diagnosis
of neurodegenerative disorders.[7 ] Utilizing 18 F-fluorodeoxyglucose 18 F-FDG PET brain imaging aids in diagnosis as lobar-specific, characteristic patterns
of hypometabolism may be evident in these patients.[8 ] FDG-PET brain imaging, and subsequently, metabolic analysis, is influenced by the
volume of brain matter present within each area of interest, for example, partial
volume averaging. Therefore, regions of atrophy may underestimate the degree of FDG
uptake in select regions, depending on neurodegenerative disease subtype. Correcting
for this is difficult and limited by the current technology. However, advanced hybrid
imaging such as simultaneous PET/MRI may decrease this underestimation, thereby increasing
accuracy of metabolic data interpretation.
Utilizing novel hybrid and molecular imaging techniques provides an avenue for improved
early and accurate diagnosis of underlying neurodegenerative disorder. Multiple previous
publications have extensively studied characteristic lobar and sublobar regional patterns
of hypometabolism in the various dementia subtypes. Specifically, AD characteristically
involves the precuneus, posterior cingulate gyrus, medial temporal, and posterior
parietal lobes with corresponding hypometabolism in these regions.[8 ],[9 ],[10 ],[11 ] Frontotemporal lobar dementia encompasses multiple disorders, with the behavioral
variant being the most common. Other subtypes include primary progressive aphasia
variants, including semantic dementia (svPPA), logopenic variant (lvPPA), agrammatic
variant or progressive nonfluent aphasia (PNFA); and also, the progressive supranuclear
palsy (PSP) variant of FTLD. Each of these disorders has slightly different patterns
of hypometabolism, but the typical underlying lobar distribution primarily involves
the frontal and temporal lobes.[8 ],[12 ],[13 ],[14 ] Finally, dementia with Lewy bodies classically demonstrates hypometabolism in the
occipital lobes, including the mesial occipital lobes and primary visual cortex in
a symmetric pattern, especially if presenting with visual hallucinations, a feature
distinguishing it from Alzheimer's disease. There is also involvement of the parietal,
mesial temporal regions, and posterior cingulate gyrus in DLB.[8 ],[15 ],[16 ],[17 ]
CBD is a rare neurodegenerative disorder that classically presents with movement abnormalities,
including myoclonus and alien limb phenomenon. This disorder is not as well understood
as the other dementias, and previous studies have reported asymmetric, unilateral
hypometabolism in the posterior frontal and parietal lobes typically involving the
primary sensorimotor cortex and also in the ipsilateral basal ganglia and thalamus.[18 ]
Currently, there is limited data assessing the relationship between semi-quantitative
changes in lobar-specific gray matter volumes and corresponding regions of brain hypometabolism
in the various subtypes of dementia and neurodegenerative disease. In this retrospective
study, we aim to identify the presence and strength of the relationship between cortical
gray matter volume loss utilizing NeuroQuant morphometric analysis and decreased FDG
uptake as per calculated hypometabolism z-scores using MIMneuro software in patients
undergoing 18 F-FDG PET/MR brain imaging as part of their routine clinical dementia workup.
Materials and Methods
Patients
This was an institutional review board and HIPAA-compliant retrospective study. We
identified 175 patients who were referred from January 2015 to February 2019 for brain
FDG PET/MRI examination as part of their routine clinical care. We excluded: patients
younger than 18 years old, clinical indication other than dementia, pregnant women,
and patients with a fasting blood glucose level >150 mg/dL; we also excluded all patients
with PET/MRI scans without dedicated brain sequences, PET/MRI scans with technical
issues such as missing sequences, or non-FDG PET/MRI studies. Of the 175 patients
identified, 89 patients meeting inclusion criteria (average age 71.4 years, range
53–85 years) received a combined clinical and radiological diagnosis of a specific
subtype of dementia (AD, FTLD, DLB, CBD) and were included in this study. Common presenting
symptoms (must have been reported in >50% of patients) by dementia subtype were collected
based on the referring clinician's documentation. Neurocognitive testing varied based
on the referring neurologist, and as such multiple different testing tools were used
by private referrers as well as our staff academic neurologists. Due to this lack
of consistency in the type of test and charting, standardized information was limited.
The symptoms listed in [Table 1 ] were obtained directly from the history of present illness in the patient's medical
charts. Additional available clinical information is summarized in [Table 1 ].
Table 1 Common clinical symptoms reported by dementia subtype
The final imaging diagnosis was made by combining the pattern of FDG distribution
and corresponding volume loss, as determined by our staff neuroradiologists and nuclear
medicine physician, in communication with the referring clinician to correlate with
the patient's symptoms and cognitive testing results. For all dementia patients, semi-quantitative
calculated z-scores <−1.65 (the threshold for statistically significant hypometabolism
using MIMneuro software) in the selected regions of interest (ROIs) were considered
abnormal and supportive of a specific diagnosis. For MRI volumetric analysis, the
lobar and sublobar cortical volume loss was considered abnormal if gray matter volumes
were less than the 5th percentile for age (>1.65 standard deviations [SDs] below the
mean), as provided by NeuroQuant software.
Image acquisition
Before imaging, an injection of approximately 5 mCi of FDG was given. Allowing for
40 min of uptake, patients were positioned for brain imaging in a Siemens mMR 3T PET/MRI
(Siemens Healthcare, Erlangen, Germany) scanner with a standard 12-channel head coil.
The PET and MRI data were obtained simultaneously. A dual-echo T1-weighted gradient-recalled
echo sequence was completed to acquire an MRI attenuation-correction map based on
Dixon segmentation (air, fat, soft tissue, and lungs). MRI data included images from
the skull vertex to the foramen magnum. Standard high-resolution 3D sagittal MPRAGE
and 3D fluid-attenuated inversion recovery (FLAIR) sequences were used for brain anatomy.
Afterward, routine diagnostic MRI sequences including T2 Turbo Spin Echo in the axial
and coronal planes, axial susceptibility-weighted imaging, diffusion tensor imaging,
proton density axial imaging, and diffusion-weighted imaging were performed while
PET data were simultaneously obtained for a total of 45 min.
Along with qualitative assessment, 3D MPRAGE image data were additionally evaluated
by NeuroQuant (2019 CorTechs Labs, Inc., San Diego, California) for semi-quantitative
volumetric analysis. The automated analysis provided by NeuroQuant compares lobar
and sublobar cortical volumes to their standardized atlas. NeuroQuant is the Food
and Drug Administration-cleared software for the utilization of parenchymal volumetric
data, providing volumetric measurements of brain structures and comparing these volumes
to a normative database adjusted for age, sex, and intracranial volume. Per the manufacturer,
NeuroQuant's normative database is founded on a population-based set which collected
data from several thousand subjects from 3 to 100 years of age with an equivalence
of gender.[19 ] Regions of parenchymal volume loss >1.65 SDs from normal controls in the standardized
atlas were flagged as abnormal. Quantitative percentiles were assigned to lobar and
sublobar areas to quantify the extent of decreased parenchymal volume loss.
Further postprocessing of PET images was performed utilizing MIMneuro version 6.9.5
(MIM Software, Inc., Cleveland, Ohio). MIMneuro software runs a region-based analysis
that calculates z-scores (number of SDs from the mean) and asymmetry measurements
for individual brain regions defined by the Single Brain Atlas and MIM Probabilistic
anatomical atlas to provide semi-quantitative analysis of brain hypometabolism. Per
the manufacturer, the Single Brain Atlas and MIM Probabilistic anatomical atlas are
composed of 43 individuals (19 females and 24 males), ages 41–80 years. The breakdown
into age ranges are as follows: six subjects ages 40–49, eight subjects ages 50–59,
14 subjects ages 60–69, 14 subjects ages 70–79, and one subject with age between 80
and 89. The mean age and SD of these 43 individuals are 63.8 ± 9.98 years. The automated
z-scores are calculated by comparing the patient to the selected age-matched set of
normal controls. Both NeuroQuant and MIMneuro software age-match the standard atlas
within 5 years of the patient's age. Please note, both software only provide semi-quantitative
data regarding the aforementioned variables and do not offer a dementia diagnosis.
Image interpretation
Two fellowship-trained neuroradiologists, one with dedicated PET/MRI training and
6 years research experience in the field, assessed scans first qualitatively and then
in conjunction with NeuroQuant and MIMneuro z-score data. Each neuroradiologist reader
interpreted the images independently, followed by adjudication with a board-certified
nuclear medicine physician with 25 years of experience in brain PET imaging. The imaging
diagnosis was then discussed with the referring clinician to correlate with the patient's
symptoms and neuropsychological testing.
Statistical analysis
NeuroQuant quantitative volumetric percentiles and MIMneuro software brain hypometabolism
z-scores were correlated using Pearson's correlation coefficient (r) in a lobar specific
pattern for each subtype of dementia (Alzheimer's dementia, frontotemporal lobar dementia,
dementia with Lewy bodies and CBD). All statistical tests were conducted with a two-tailed
t-test with a P < 0.05 significance level. In addition, ratio of z-score to volume
percentile (M/V index) was calculated as a way to further compare regions by providing
assessment of metabolic activity within respective corresponding regions of volume
loss.
Results
The distribution of the dementia subtypes within the 89 individuals was as follows:
29 with Alzheimer's dementia, 34 with frontotemporal lobar degeneration, 14 with dementia
with Lewy bodies, and 12 with CBD. Of the 34 patients with FTLD, subtypes were as
follows: 16 patients with behavioral variant (bvFTLD), 10 with semantic dementia,
4 with logopenic variant primary progressive aphasia, 2 with the agrammatic variant
primary progressive aphasia, and 2 with the PSP.
The average patient age was 71.3 (SD = 7.7), and the majority (54.1%) were female.
Mean z-scores were generally negative across all corresponding ROIs in this study
[results detailed in [Table 2 ]. Associations between z-scores and volumetric percentiles were moderate in strength
across all patients, specifically in the parietal lobe, temporal pole, and occipital
lobe and weak in the frontal and temporal lobes. These findings were significant in
the frontal lobe, parietal lobe, temporal pole, occipital lobe, and temporal lobe
(P < 0.05). The ratio of the mean z-score to mean volumetric percentile (M/V index)
was determined and provided in [Table 3 ] and [Table 4 ] for the various neurodegenerative disorders, respectively.
Table 2 Descriptive characteristics for shared MIMneuro and NeuroQuant regions of interest
across all patients included in this study
Table 3 Alzheimer's dementia and frontotemporal lobar degeneration: Average z-scores, lobar
volumetric percentiles, and degree of correlation
Table 4 Dementia with Lewy bodies and corticobasal degeneration: Average z-scores, lobar
volumetric percentiles, and degree of correlation
Alzheimer's dementia
Twenty-nine patients (10 males, 19 females, average age 70.6 years) with suspected
AD had an average parietal lobe z-score of −1.76 with a corresponding mean lobar volume
in the 21st percentile. The mean z-score for the temporal lobe was −1.67, with a corresponding
mean lobar volume in the 7th percentile. There was a weak positive relationship that
approached statistical significance between the Z-scores and degree of volume loss
for patients with suspected Alzheimer's disease in the parietal lobe, r = 0.3 (P = 0.120). A similar yet statistically significant, weakly positive relationship was
noted for the temporal lobe, r = 0.38 (P < 0.05). These results are presented in [Table 3 ] accordingly.
Frontotemporal lobar degeneration
Thirty-four patients (19 males, 15 females, average age 72.8 years) with suspected
frontotemporal lobar degeneration demonstrated a mean frontal lobe z-score of −1.10
and corresponding lobar volume in the 22.8th percentile. The mean temporal lobe z-score
was −0.59, with a corresponding volume in the 13.8th percentile. There was a weakly
positive relationship that approached statistical significance between the z-scores
and degree of volume loss for patients with suspected FTLD in the frontal lobe, r
= 0.35 (P = 0.051). No relationship between volume and metabolic z-score was suggested by the
data for the temporal lobe, r = 0.02 (P = 0.916). These results are outlined in [Table 3 ] accordingly.
[Figure 1a ], [Figure 1b ], [Figure 1c ], [Figure 1d ] demonstrates the cortical surface maps, NeuroQuant output, and MIMneuro z-score
output images of a patient with PET/MRI findings suggestive of bvFTLD. Cortical surface
maps [Figure 1a ] demonstrated significantly decreased uptake in the frontal lobes, most evident in
the right frontal lobe. These findings were confirmed on the z-score data overlay
on cortical surface maps [Figure 1b ]. NeuroQuant morphometric results [Figure 1c ] demonstrate a corresponding decrease in the volume of the frontal lobes, which was
in the 1st percentile when compared to age-matched controls. MIMneuro software z-score
tabular output [Figure 1d ] for this patient with FTLD further confirmed decreased metabolism with diffusely
decreased z-scores in the frontal lobes (total frontal lobe z-score = −4.98).
Figure 1 (a) 71-year-old female with behavioral variant behavioral variant frontotemporal
lobar degeneration: cortical surface maps. (b) 71-year-old female with behavioral
variant frontotemporal lobar degeneration: z-score overlay on cortical surface maps.
(c) 71-year-old female with behavioral variant frontotemporal lobar degeneration:
NeuroQuant morphometric results. (d) 71-year-old female with behavioral variant frontotemporal
lobar degeneration: MIMneuro z-score output
[Figure 2a ], [Figure 2b ], [Figure 2c ], [Figure 2d ] is an example of a patient with PET/MRI findings consistent with semantic dementia
(semantic variant, primary progressive aphasia) subtype of FTLD. Fusion FDG-PET and
FLAIR axial images [Figure 2a ] provide insight into the anatomy and distribution of hypometabolism with striking
decreased tracer uptake and corresponding volume loss in the left temporal lobe. Corresponding
z-score overlay data [Figure 2b ] confirm these findings and shows markedly decreased metabolism, particularly in
the left temporal pole. NeuroQuant morphometric analysis [Figure 2c ] and MIMneuro z-score tabular data [Figure 2d ] demonstrated concordance of decreased grey matter volume (1st percentile) and hypometabolism
(z-score = −3.13) in the left temporal pole.
Figure 2 (a) 82-year-old male with semantic PPA: fusion fluorodeoxyglucose positron emission
tomography and fluid-attenuated inversion recovery magnetic resonance imaging axials.
(b) 82-year-old male with semantic PPA: z-score overlay on cortical surface maps.
(c) 82-year-old male with semantic PPA: NeuroQuant morphometric analysis. (d) 82-year-old
male with semantic PPA: MIMneuro software z-score output
Dementia with Lewy bodies
Fourteen patients (7 males, 7 females, average age 66.7 years) with suspected DLB
demonstrated an average occipital lobe z-score of −2.88 and corresponding mean lobar
volume in the 16th percentile. Mean middle occipital gyrus z-score was −4.23, which
corresponded to a sublobar volume in the 44th percentile. Mean z-scores of the parietal
lobes were −2.15 with corresponding mean lobar volumes in the 17th percentile. A moderate
positive trend that approached statistical significance was seen in the occipital
lobe and middle occipital gyrus r = 0.42 (P = 0.130) and 0.5 (P = 0.067), respectively. The parietal lobe demonstrated a weakly positive, however
not statistically significant, relationship between the metabolic z-score and lobar
volume with r = 0.22 (P = 0.447). These results are outlined in [Table 4 ].
Corticobasal degeneration
Twelve patients (3 males, 9 females, average age 74.3 years) with clinical history
and FDG uptake pattern suggestive of underlying CBD including unilateral hemispheric
hypometabolism with decreased FDG uptake including in the primary sensorimotor cortex
and ipsilateral basal ganglia and thalami, demonstrated an average superior parietal
lobule z-score of −1.65 and corresponding mean volume in the 16th percentile. A moderately
positive relationship between these findings was seen in the superior parietal lobule
r = 0.58. This relationship was found to be statistically significant (P < 0.05). These results are outlined in [Table 4 ].
Discussion
Neurodegenerative disorders encompass a broad range of diseases initially presenting
with mild cognitive impairment, which, oftentimes, then directs the neurologist to
pursue further neuroimaging workup following cognitive screening tests such as the
Montreal Cognitive Assessment or the Mini-Mental State Exam.
Neuroradiologists are then charged with the responsibility of determining the presence,
and subtype, of the underlying neurodegenerative condition. Since there is relatively
little guidance about the utility of semi-quantitation reported by different software
programs commonly utilized in clinical practice such as NeuroQuant and MIMneuro software,
we assessed the feasibility of an approach relying on these semi-quantitative tools
with a focus on the strength of expected positive correlation between lobar-specific
patterns of hypometabolism on metabolic imaging and volumetric patterns of cortical
volume loss on structural MRI brain scans. When compared among all dementia subtypes,
we found a statistically significant, positive relationship between grey matter volume
loss and corresponding regional hypometabolism, as summarized in [Table 2 ]. Within the various subtypes of dementia, however, there was some variability in
the strength of this relationship.
For our subset of patients with clinical and imaging findings consistent with underlying
AD, the parietal and temporal lobe semi-quantitative data demonstrated a positive
correlation between the degree of hypometabolism and volume loss in the parietal lobes,
including the precuneus and posterior cingulate gyrus. These findings were statistically
significant for the temporal lobe and approached statistical significance in the parietal
lobe. In addition, the average z-score data and cortical volume percentiles, when
assessed independently of each other, were consistent with prototypical findings seen
in AD patients. The calculated M/V index for the temporal and parietal lobe in this
subset of AD patients demonstrated a larger, more negative value in the temporal lobe
compared to the parietal lobe indicating that the remaining neurons in the temporal
lobe exhibited higher amounts of volume loss per degree of hypometabolism. This is
as opposed to the parietal lobe, which, albeit, still demonstrated some correlation,
although not approaching statistical significance.
Furthermore, positive relationship between semi-quantitative data for cortical volume
loss and corresponding hypometabolism was elucidated in the frontal lobes in our patients
with suspected FTLD, which nearly reached statistical significance. This finding is
important as it suggests this correlation may be used to strengthen a radiologist's
confidence in the diagnosis of FTLD in patients displaying frontal lobe hypometabolism
and volume loss. Interestingly, no relationship between the two variables was seen
in the temporal lobe of patients with FTLD, which may be explained by a predominance
of bvFTLD cases in our dataset as compared to the primary progressive aphasia subtypes
including semantic dementia and logopenic PPA of which there were 10 and 4 cases,
respectively. Of note, the M/V index for both the frontal and temporal lobe demonstrated
nearly identical values of −4.82 versus −4.34, respectively, which likely represent
similar degrees of hypometabolism per unit of volume loss in the frontal and temporal
lobes of FTLD patients in our dataset.
In our DLB cohort, we noted moderately positive relationship for semi-quantitative
data assessing hypometabolism and volume loss, specifically in the occipital lobes
including the middle occipital gyrus which is characteristic for DLB as previously
shown by Higuchi et al .[16 ] In this instance the middle occipital gyrus, which is part of the occipital lobe
had a higher volume percentile compared to the occipital lobe (44 vs. 16). This is
due to the nature of the NeuroQuant software's semi-quantitative analysis providing
an average percentile for the entire occipital lobe. This includes the middle occipital
gyrus but also other regions which were likely more disproportionately affected with
volume loss thus decreasing the overall average of the occipital lobe to the 16th
percentile in our study. In addition, a weakly positive relationship was seen for
the parietal lobes including in the precuneus and posterior cingulate (although to
a lesser degree than with AD cases), regions also typically affected in DLB. However,
the latter relationships did not reach statistical significance, likely due to the
size of our study sample and variable disease severity. The M/V index for DLB demonstrated
increased values in the occipital and parietal lobes compared to the middle occipital
gyrus, which again may represent higher rates of hypometabolism per region of volume
loss in these regions.
In our CBD cohort, in addition to the characteristic FDG uptake pattern with markedly
hypometabolic single hemisphere including the ipsilateral thalami and basal ganglia
and importantly involving the primary sensorimotor cortex, a statistically significant,
moderately positive, relationship on semi-quantitative analysis was observed in the
ipsilateral superior parietal lobule which is a region thought to be typically involved
in CBD according to a study by Grijalvo-Perez and Litvan.[18 ] This again resonates with our earlier statements that radiologists may have increased
clinical confidence in their diagnosis of underlying CBD when evaluating a patient
with suggestive hypometabolism and volume loss patterns upon semi-quantitative review
of images, particularly focusing on the superior parietal lobule, along with history
and symptoms supportive of underlying CBD. In addition, the M/V index for the superior
parietal lobule was more negative than regions seen in the other disorders (e.g.,
frontal/temporal lobe in FTLD). This also likely corresponds to increasing degrees
of hypometabolism per unit of volume loss, similar to lobar-specific regions typical
for previously mentioned disorders.
Of note, there are several limitations inherent to our study. First, the retrospective
nature of the research and selection of cases, which were dependent on our referring
clinicians, may skew the cohort composition. The sample size and heterogeneity of
the patient groups are a limiting factor, as described above, within each disorder
and may have contributed to the lack of statistically significant metabolic and volumetric
correlation in some lobar-specific regions including in the parietal lobe in AD, temporal
lobe in FTLD and parietal lobe in DLB; regions shown to be affected in prior studies.[4 ],[5 ],[13 ],[17 ],[18 ]
Furthermore, the volumetric analysis and the calculated hypometabolism z-scores, as
provided by NeuroQuant and MIMneuro software, respectively, are both semi-quantitative
tools and as such are meant to be representative of hypometabolism and brain volume
loss patterns and are not quantitative data. Therefore, although the z-score can potentially
be strikingly low, cortical volume loss cannot decrease below the 1st percentile.
This limited our ability to draw relationships for patients with the most severe disease
as they would appear to have the same degree of volume loss as subjects with less
severe hypometabolism. As part of our data analyses, we also noted regions with hypometabolism
without significant brain volume loss, which may reflect decreases in neuronal activity
or other effects such as diaschisis. Alternatively, in cases where there was volume
loss with maintained levels of metabolism, findings may reflect a compensatory increased
activity of residual neurons due to the surrounding neurodegenerative process. These
observations warrant further delineation by carefully planned longitudinal clinical
studies.
Conclusion
Semi-quantitative 18 F-FDG PET/MRI analysis demonstrated a positive relationship between structural volume
loss and hypometabolism within certain lobar-specific regions, depending on neurodegenerative
disorder subtype. We believe our findings add new information to the limited existing
literature specifically regarding the clinical utility of hybrid brain PET/MRI in
patients presenting with cognitive impairment due to suspected neurodegeneration.
With the use of computer assisted semi-quantitative tools such as NeuroQuant morphometric
analysis and MIMneuro software hypometabolism z-scores; neuroradiologists, nuclear
medicine physicians, and clinicians interpreting these images can gain increased confidence
in their diagnoses and provide more accurate and precise information in the assessment
of patients with cognitive impairment due to underlying dementia and related neurodegenerative
disorders.
Declaration of patient consent
The authors certify that they have obtained all appropriate patient consent forms.
In the form the patient(s) has/have given his/her/their consent for his/her/their
images and other clinical information to be reported in the journal. The patients
understand that their names and initials will not be published and due efforts will
be made to conceal their identity, but anonymity cannot be guaranteed.
