Diagnostic Performance of ASL-MRI and FDG-PET in Frontotemporal Dementia: A Systematic Review and Meta-Analysis

Abstract

BACKGROUND:

While the diagnosis of frontotemporal dementia (FTD) is based mostly on clinical features, [¹⁸F]-FDG-PET has been investigated as a potential imaging standard in ambiguous cases, with arterial spin-labeling (ASL) MRI gaining recent interest.

PURPOSE:

The purpose of this study is to conduct a systematic review and meta-analysis on the diagnostic performance of ASL MRI in patients with FTD and compare it with that of [¹⁸F]-FDG-PET.

DATA SOURCES:

A systematic search of PubMed, Scopus, and Embase was conducted until March 13, 2024.

STUDY SELECTION:

Inclusion criteria were original articles, patients with FTD and/or its variants, use of ASL MR perfusion imaging with or without [¹⁸F]-FDG-PET, and presence of sufficient diagnostic performance data. Exclusion criteria were meeting abstracts, comments, summaries, protocols, letters and guidelines, longitudinal studies, and overlapping cohorts.

DATA ANALYSIS:

The quality of eligible studies was assessed by using the Quality Assessment of Diagnostic Accuracy Studies-2. Pooled sensitivity, specificity, and diagnostic odds ratio (DOR) for [¹⁸F]-FDG-PET and ASL MRI were calculated, and a summary receiver operating characteristic curve was plotted.

DATA SYNTHESIS:

Seven eligible studies were identified, which included a total of 102 patients with FTD. Aside from some of the studies showing, at worst, an unclear risk of bias in patient selection, index test, flow, and timing, all studies showed low risk of bias and applicability concerns in all categories. Data from 4 studies were included in our meta-analysis for ASL MRI and 3 studies for [¹⁸F]-FDG-PET. Pooled sensitivity, specificity, and DOR were 0.70 (95% CI: 0.59–0.79), 0.81 (95% CI: 0.71–0.88), and 8.00 (95% CI: 3.74–17.13) for ASL MRI and 0.88 (95% CI: 0.71–0.96), 0.89 (95% CI: 0.43–0.99), and 47.18 (95% CI: 10.77–206.75) for [¹⁸F]-FDG-PET.

LIMITATIONS:

The number of studies was relatively small, with a small sample size. The studies used different scanning protocols as well as a mix of diagnostic metrics, all of which might have introduced heterogeneity in the data.

CONCLUSIONS:

While ASL MRI performed worse than [¹⁸F]-FDG-PET in the diagnosis of FTD, it exhibited a decent diagnostic performance to justify its further investigation as a quicker and more convenient alternative.

Frontotemporal dementia (FTD) is a spectrum of clinical syndromes characterized by progressive neurodegeneration and atrophy in the frontal and temporal lobes. More specifically, its most common behavioral subtype (behavioral variant of frontotemporal dementia [bvFTD]) is characterized by impairments in behavior and language.^1-3 With an estimated prevalence of 15 per 100,000 in people between 45 and 64 years of age,⁴ it is the second most common cause of dementia in patients under the age of 65.⁵ In patients older than 65, FTD is the third most common cause of dementia.

Given the rapid functional decline and relatively short average survival time of 5 to 10 years for patients with FTD, early and reliable diagnosis remains critical.^6,7 The premortem diagnosis of FTD is based mostly on clinical features, standardized cognitive tests, and consensus criteria.⁸ Unfortunately, symptoms can often mimic those of other neurodegenerative diseases like Alzheimer disease (AD) or even psychiatric conditions with prominent behavioral features, such as bipolar disorder and schizophrenia.⁹

[¹⁸F]-FDG-PET has been found to help with not only the diagnosis of FTD,^10-12 but also its differentiation from other dementias like AD.¹² [¹⁸F]-FDG-PET identifies metabolically active tissues throughout the body. As such, it can leverage the distinct atrophic and hypometabolic patterns of different dementias to achieve reliable diagnosis and differentiation.¹³ Unfortunately, [¹⁸F]-FDG-PET carries several disadvantages, including high costs and radiation exposure. Arterial spin-labeling (ASL) MRI has thus been investigated as an alternative to [¹⁸F]-FDG-PET, requiring no radiation exposure nor injection for the diagnosis of several dementias.¹⁴ Performing ASL MRI is convenient as patients with neurocognitive symptoms or suspicion of dementia are already likely to undergo structural MRI and it can thus be added as part of the protocol. Such investigation is based on the fact that areas of hypoperfusion and hypometabolism follow similar patterns.^15,16 Further endeavors have been made to compare the performance of ASL MRI and [¹⁸F]-FDG-PET in the diagnosis of FTD within the same population with comparable yet variable results.^17-23

While some studies have tried pooling and systematically reviewing the available literature on the use of ASL MRI in several dementias,^24,25 the same has not been done for FTD specifically. Therefore, the purpose of this systematic review and meta-analysis is to assess the diagnostic performance of ASL MRI versus [¹⁸F]-FDG-PET in patients with FTD.

MATERIALS AND METHODS

Search Strategy and Eligibility Criteria

The systematic review was conducted following an apriori-defined protocol (available from the authors upon request). It was designed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines.²⁶ Two independent reviewers (R.D. and B.B.O., with more than 1–2 years of research experience, respectively) systematically searched the PubMed, Scopus, and Embase databases from inception to March 13, 2024, to identify studies investigating the use of ASL MR perfusion in patients with FTD and its variants. The detailed search string can be found in the Online Supplementary Data. The search was supplemented by citation checking and contacting the author of the eligible articles if necessary.

All search results were imported to EndNote 21 (Clarivate), where duplicates were automatically screened and deleted. The remaining references were imported into the Rayyan online platform,²⁷ where further duplicates were removed. Screening of titles and abstracts was performed by 2 independent reviewers (R.D. and B.B.O.). A third reviewer (L.P.L., neuroradiologist, 14 years of experience) examined all screening results, and discordant evaluations were resolved by discussion to reach a consensus. All screening decisions are documented and outlined in the Online Supplemental Data.

Inclusion Criteria

Studies were included if they met all the following criteria: a) original articles published in English, b) inclusion of patients with FTD and/or its variants, c) studies that provided the diagnostic criteria for FTD and its variants used, d) used ASL MR perfusion imaging with or without [¹⁸F]-FDG-PET, and e) presence of sufficient data on the performance of ASL. There were no limitations regarding the age, nationality, race, sex, and disease stage of the participants.

Exclusion Criteria

Studies were excluded if they met any of the following criteria: a) duplicates; b) animal studies; c) did not provide qualitative or quantitative information about the diagnostic performance of ASL; d) longitudinal studies; e) studies with overlapping cohorts; and f) meeting abstracts, comments, summaries, protocols, letters and guidelines. In case 2, studies had overlapping cohorts; the most recent study was included.

Data Extraction

Data extraction was performed independently by 2 investigators (R.D. and P.A., 2 years of research experience). For each article, we extracted the following variables: a) authors, b) journal, c) year of publication, d) study design, e) target condition, f) sample size, g) patient demographics such as age and sex, h) ASL sequence and protocol, i) reference standard, j) study conclusions, k) sensitivity, l) specificity, and m) accuracy. If true-positive (TP), true-negative (TN), false-positive (FP), and false-negative (FN) were not explicitly provided, they were calculated by using the provided diagnostic numbers. The numbers for only patients with bvFTD were extracted.

Quality Evaluation

The quality assessment was conducted independently by 2 investigators (R.D. and P.A.) by using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2).²⁸ Four domains were evaluated: patient selection, index test, reference standard, and flow and timing. The questions pertaining to each domain were answered with Yes, No, or Unclear, with a final score for each domain reached via consensus between the 2 investigators. Concerns about applicability and bias risk were graded as low, high, or unclear. All discordant evaluations were settled through discussion or by a third independent reviewer (L.P.L.).

Meta-Analysis

All statistical analyses were performed on STATA 17 (Stata). When needed, TP, FP, FN, and TP were calculated by using the sample sizes and diagnostic metrics provided by the studies. The pooled sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), diagnostic odds ratio (DOR), and the heterogeneity were calculated by a random-effects model with corresponding 95% CI by using the Clopper-Pearson method.²⁹ In addition, we summarized the joint distribution of TP and TN rates in a summary receiver operating characteristic (SROC) curve. The area under the curve represents an analytical summary of test performance and displays the trade-off between sensitivity and specificity. The pooled sensitivity, specificity, PLR, NLR, DOR, and SROC curves were calculated and plotted by using the metadta STATA command.³⁰ Statistical heterogeneity was quantified by using the I² statistic with the following interpretation: 0%–25% indicates low heterogeneity, 50%–75% indicates high heterogeneity, and >75% indicates large heterogeneity.³¹ Heterogeneity was also assessed visually on the SROC curve, with high heterogeneity suspected if individual studies substantially deviated from the SROC line.³² The same analysis was performed for ASL and [¹⁸F]-FDG-PET separately.

RESULTS

Eligible Studies

The study selection process is summarized in Fig 1. After excluding duplicates (n = 492) and articles unrelated to our topic of interest based on the title and abstract (n = 1718), the full text of 23 articles was ultimately acquired and retrieved. Sixteen articles were excluded because of the following reasons, which are further detailed in the Online Supplemental Data: no diagnostic performance data (n = 6), meeting abstracts (n = 3), longitudinal studies (n = 2), presymptomatic genetic FTD (n = 2), multiple dementias lumped together (n = 1), overview/not ASL (n = 1), and overlapping cohorts (n = 1), leading to a total of 7 included studies.^17-23 Because of a lack of diagnostic metric reporting, 3 studies^19,22,23 were included in the review but not in the meta-analysis.

FIG 1.

PRISMA flow diagram of the study selection process.

Study Characteristics

The Online Supplemental Data summarize the main characteristics of the included studies. The studies were published between 2006 and 2018. The studies were all cross-sectional, with all but one²⁰ happening at a single-center. All studies took place at specialized memory/dementia clinics, with other possibly overlapping diseases being excluded from all of them. In addition to FTD, some of the studies included other diseases like AD, progressive supranuclear palsy, and limbic encephalitis.^18-23 One study²³ used a multidisciplinary team and follow-up approach as the standard for FTD diagnosis, whereas the rest used published consensus criteria. A total of 102 patients with FTD were included across all 7 studies, with a mean age of 60.8 years. Three studies^17,20,23 relied on the visual assessment of their images for diagnosis, one¹⁹ on both visual and quantitative assessments, whereas the other 3 relied on quantitative measurements. The studies by Verfaillie et al,²² Fällmar et al,¹⁹ and Weyts et al²³ did not provide diagnostic performance metrics to calculate TP, TN, FP, and FN, while Du et al¹⁸ did not study [¹⁸F]-FDG-PET. Therefore, 83 patients were included in the subsequent meta-analysis on the performance of ASL, while 62 patients were included in the meta-analysis on the performance of [¹⁸F]-FDG-PET.

The Online Supplemental Data summarize the ASL MRI scanning details of each study. One study¹⁸ used a 1.5T scanner, while the rest used 3T scanners. Two studies^18,21 used pulsed ASL sequences, while the rest used 3D pseudocontinous ASL (3DPCASL) sequences. Scanning parameters and voxel sizes were variable and heterogeneous across studies.

Quality Assessment

A summary of the quality assessment of the included studies by using the QUADAS-2 tool is illustrated in Fig 2. There was no suspicion of a high risk of bias or high applicability concerns regarding any category in all included studies. Three studies^17,18,20 had an unclear risk of bias regarding patient selection due to unclear reporting of randomization/consecutiveness during the selection, while the rest had a low risk. Two studies^18,21 showed an unclear risk of bias relating to the index test due to unclear reporting on the blinding of the index test readers, with the rest having a low risk. Three studies^17,19,20,23 had an unclear risk of bias regarding flow and timing because not all patients were included in the analysis, while the rest had a low risk. No risk of bias was found regarding the reference standard in any of the studies, since all used published criteria or clinical follow-up for the diagnosis of FTD. There was also a low concern for applicability for the study population, index test, and reference standard since all 3 matched the review question in all the included studies.

FIG 2.

Summary of the QUADAS-2 results.

Meta-Analysis

ASL MRI.

Four articles reported enough diagnostic metrics to be included in our analysis. Fig 3A shows the forest plot of the calculated sensitivity and specificity of each study. Sensitivities ranged between 0.55 (95% CI: 0.32–0.77) and 0.78 (95% CI: 0.60–0.91). Specificities ranged between 0.60 (95% CI: 0.26–0.88) and 0.93 (95% CI: 0.68–1.00). The pooled sensitivity was 0.70 (95% CI: 0.59–0.79), while the pooled specificity was 0.81 (95% CI: 0.71–0.88). Fig 4A shows the forest plot for the DOR of each study. DOR values for each study ranged from 3.50 (95% CI: 0.55–22.30) to 50.00 (95% CI: 5.57–449.03), with a pooled DOR of 8.00 (95% CI: 3.74–17.13). Visual assessment of the SROC curve (Fig 5A) showed no prominent departures/outliers.

FIG 3.

Forest plot of the studies’ sensitivity and specificity for (A) ASL MRI and (B) [¹⁸F]-FDG-PET.

FIG 4.

Forest plot of the studies’ DOR for (A) ASL MRI and (B) [¹⁸F]-FDG-PET.

FIG 5.

SROC plot of the studies for (A) ASL MRI and (B) [¹⁸F]-FDG-PET.

[¹⁸F]-FDG-PET.

Three articles reported enough diagnostic metrics to be included in our analysis. Fig 3B shows the forest plot of calculated sensitivity and specificity of each study. Sensitivities ranged between 0.78 (95% CI: 0.60–0.91) and 0.95 (95% CI: 0.75–1.00). Specificities ranged between 0.55 (95% CI: 0.38–0.71) and 1.00 (95% CI: 0.78–1.00). The pooled sensitivity was 0.88 (95% CI: 0.71–0.96) while the pooled specificity was 0.89 (95% CI: 0.43–0.99). Fig 4B shows the forest plot for the DOR of each study. DOR values for each study ranged from 23.47 (95% CI: 2.85–193.61) to 105.40 (95% CI: 5.62–1976.82), with a pooled DOR of 47.18 (95% CI: 10.77–206.75). Visual assessment of the SROC curve (Fig 5B) showed no prominent departures/outliers.

DISCUSSION

FTD is one of the major causes of neurodegenerative dementia in adulthood, accounting for 20% of all degenerative dementia cases in adults younger than 65.⁵ Presenting symptoms can mimic other neurodegenerative and psychiatric diseases, potentially leading to misdiagnosis and delays in intervention in a highly progressive disease.⁹ [¹⁸F]-FDG-PET has been found to be a useful tool in the diagnosis and differentiation of dementias, including FTD.^10-12 ASL MRI is being investigated as a more convenient option for the diagnosis of FTD,^17-23 requiring no radiotracer injection and given that it could be acquired during the same session as that of a standard structural MRI for the evaluation of cognitive symptoms. To our knowledge, our study represents the first systematic review and meta-analysis on the use of ASL MRI versus [¹⁸F]-FDG-PET in the diagnosis of FTD specifically.

Our search of the available literature yielded a total of 7 articles covering 102 patients. While all were assessed, evaluated, and reviewed, only 4 of them (and thus 83 patients) were included in the meta-analysis of ASL MRI performance, while only 3 (and thus 62 patients) were included in that of [¹⁸F]-FDG-PET. Our quality assessment showed an unclear risk of bias in several of the included studies regarding several of the categories. Nevertheless, visual and quantitative assessments showed the included data to be mostly in agreement.

Virtually all studies individually found that [¹⁸F]-FDG-PET still holds an advantage in diagnosing FTD, with its performance being at least as good as that of ASL MRI. Our analysis further substantiates this claim, with a pooled sensitivity of 0.88 versus 0.70, pooled specificity of 0.89 versus 0.81, and pooled DOR of 47.18 versus 8.00 for [¹⁸F]-FDG-PET versus ASL MRI, respectively. Such findings fall in line with what previous studies have found regarding the diagnostic performance of [¹⁸F]-FDG-PET and ASL MRI in all-type dementias. Dolui et al³³ found that ASL MRI and [¹⁸F]-FDG-PET hold comparable diagnostic power in AD and mild cognitive impairment, while Haidar et al²⁵ also found that [¹⁸F]-FDG-PET performed better in the diagnosis of all-type dementias. We suspect multiple reasons behind this finding. First, basi-frontal cortices physiologically experience relative hyperperfusion compared with the rest of the brain, normally leading to hyperperfusion artifacts and false suspicion of pathologically increased perfusion.³⁴ Given that the frontal and temporal lobes are primarily affected in FTD, the pathologic decrease in frontal lobe perfusion might not be enough to overcome its baseline relative hyperperfusion, leading to difficulty in its identification. Second, all studies used a fixed postlabel delay (PLD) for all of their patients during imaging. Because of the variable blood velocity from patient to patient depending on vessel patency, smearing or blurring of the signal could occur in patients where PLD is not optimal, a phenomenon referred to as PLD artifact.^34,35 Investigations have suggested using variable PLDs, depending on patient blood velocity, to minimize such artifact.³⁶ Third, glucose metabolism is one of the first markers of neuronal activity change,³⁷ preceding any atrophy or changes in regional perfusion. As such, in diseases with rapid progression, hypoperfusion might lag behind hypometabolism,³⁸ at least in part explaining the difference in performance of their respective imaging techniques. Nevertheless, our analysis showed ASL MRI possessing a decent diagnostic power, making it a potential alternative for [¹⁸F]-FDG-PET in cases where clinical symptoms were too ambiguous.

ASL MRI presents itself as an alternative to [¹⁸F]-FDG-PET in cases were contrast injection is contraindicated, such as in diabetic patients. It can also be performed in addition to structural MRI as part of the work-up for neurocognitive symptoms, saving time and cost. For example, it can help rule in or out FTD where symptoms overlap with other diseases like AD, enabling early and more certain therapy, especially in light of emerging disease-modifying anti-amyloid therapies. As a final step, and in cases where ASL did not yield satisfactory results, we suspect that [¹⁸F]-FDG-PET can be added to further increase sensitivity and specificity, especially in the case of double positives or double negatives. What remains to be overcome are the limitations inherent in the current state of ASL MRI. Notably, ASL suffers from a lack of standardization across scanner vendors,³⁹ leading to difficulties in consistent clinical applications. ASL also has a higher failure rate compared with both structural MRI and [¹⁸F]-FDG-PET. Finally, there are cases where both ASL and [¹⁸F]-FDG-PET cannot help easily distinguish between diseases with similar presentations to FTD. For example, the behavioral/dysexecutive variant of AD shares overlapping clinical and imaging features with FTD.⁴⁰ The clinical utility of both ASL MRI and PET in such cases remains to be studied.

There are several limitations to our study. First, the number of patients is still too small to draw any final conclusions. The fact that the studies did not show heterogeneity does not rule out the potential risks that arise from including low numbers in our statistical analysis. The cross-sectional design of included studies could have also introduced some bias, as FTD is best diagnosed via longitudinal assessments in addition to published criteria. Second, different scanner protocols were used in the study for both ASL MRI and [¹⁸F]-FDG-PET. More specifically, ASL acquisition was heterogeneous across papers, with some using 3DPCASL, others using pulsed ASL (PASL), and none using 2DPCASL, introducing an added layer of variability in the data. This could, however, be considered a strength given that ASL was still found to be reliable. It is worth noting that PASL slightly outperformed 3DPCASL (sensitivity/specificity 0.75/0.82 versus 0.6/0.79), though the sample size is too small to draw any meaningful conclusions. Third, the studies used a mix of quantitative and visual assessment of the obtained images for diagnosis. While all studies assessed brain regions that were shown to be affected by FTD, different papers focused on different brain regions, with some leaving it up to the reader’s impression. Fourth, most studies included studies that were assumed to address patients with bvFTD based on the provided numbers and definitions. As such, the diversity of included FTD subtypes remains limited, and generalization for all FTD, let alone frontotemporal lobar degeneration, cannot be certainly made. Therefore, a more extensive network of studies is still needed to address all of these issues and allow subgroup analyses. Finally, it remains unclear how our findings directly translate into the real world, as all patients with FTD were selected with a predetermined diagnosis, and the clinicians evaluating the images were only diagnosing FTD versus normal instead of giving a differential diagnosis, including other possible diseases.

CONCLUSIONS

Our study shows that despite being lower compared with [¹⁸F]-FDG-PET, the diagnostic performance of ASL MRI was still good enough to consider it as a potential safer alternative, with a pooled sensitivity of 0.70, specificity of 0.81, and DOR of 8.00. Further studies are still needed, however, to eliminate potential confounders and sources of variability in the data, allow for targeted subgroup analyses, and eventually explore its relevance in differentiating FTD from other neurodegenerative dementias.

ABBREVIATIONS:

3DPCASL

3D pseudocontinuous ASL

AD

Alzheimer disease

ASL

arterial spin-labeling

bvFTD

behavioral variant of frontotemporal dementia

DOR

diagnostic odds ratio

FN

false-negative

FP

false-positive

FTD

frontotemporal dementia

NLR

negative likelihood ratio

PASL

pulsed ASL

PLD

postlabel delay

PLR

positive likelihood ratio

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-Analysis

QUADAS-2

Quality Assessment of Diagnostic Accuracy Studies-2

SROC

summary receiver operative characteristic

TN

true-negative

TP

true-positive