Neuroimaging alterations in dementia with Lewy bodies and neuroimaging differences between dementia with Lewy bodies and Alzheimer's disease: An activation likelihood estimation meta‐analysis

Abstract Aims The aim of this study was to identify brain regions with local, structural, and functional abnormalities in dementia with Lewy bodies (DLB) and uncover the differences between DLB and Alzheimer's disease (AD). The neural networks involved in the identified abnormal brain regions were further described. Methods PubMed, Web of Science, OVID, Science Direct, and Cochrane Library databases were used to identify neuroimaging studies that included DLB versus healthy controls (HCs) or DLB versus AD. The coordinate‐based meta‐analysis and functional meta‐analytic connectivity modeling were performed using the activation likelihood estimation algorithm. Results Eleven structural studies and fourteen functional studies were included in this quantitative meta‐analysis. DLB patients showed a dysfunction in the bilateral inferior parietal lobule and right lingual gyrus compared with HC patients. DLB patients showed a relative preservation of the medial temporal lobe and a tendency of lower metabolism in the right lingual gyrus compared with AD. The frontal‐parietal, salience, and visual networks were all abnormally co‐activated in DLB, but the default mode network remained normally co‐activated compared with AD. Conclusions The convergence of local brain regions and co‐activation neural networks might be potential specific imaging markers in the diagnosis of DLB. This might provide a pathway for the neural regulation in DLB patients, and it might contribute to the development of specific interventions for DLB and AD.


| INTRODUC TI ON
Dementia with Lewy bodies (DLB) is characterized by fluctuating cognition, recurrent visual hallucinations, rapid eye movement sleep behavior disorder, and spontaneous parkinsonism, 1 accounting for 15%-20% of the total dementia cases at autopsy. 2,3 Although DLB is the second most common neurodegenerative disorder after AD, the sensitivity of its diagnosis in clinical practice is suboptimal. The widely spread pathologies related to Lewy bodies and coexisting AD-type pathologies 4-6 make the clinical manifestations complex and highly variable, increasing the difficulty of the differential diagnosis between DLB and AD, especially in the early stages.
Multimodal neuroimaging is widely used in clinical practice. For example, the role of DAT imaging in distinguishing DLB from AD is well established, with a sensitivity of 78% and specificity of 90%. 7 A neuropathologically confirmed study showed that DAT imaging can distinguish between DLB and AD more accurately than the consensus clinical criteria. 8 However, broader structural and functional studies provided conflicting results. Therefore, stable and consistent indicators that provide a theoretical basis for the diagnosis and differential diagnosis of DLB are still lacking.
Structural imaging can reflect changes in brain volume at voxelwise level. 9,10 Some reports showed the cortical atrophy of the frontal lobe, 11 temporal lobe, 11,12 parietal lobe, and occipital lobe 12 in DLB. However, other studies found a relatively concentrated pattern of atrophy in the subcortical brain, including midbrain, hypothalamus/thalamus, basal ganglia, 13,14 and substantia innominate. [15][16][17] DLB patients with a similar level of dementia have relatively better preservation of the hippocampus, temporal lobe, 12,14,18 and amygdala. 11,19 This aspect means that they are more likely to develop subcortical atrophy than AD patients. 13,14 A functional imaging report showed a hypoperfusion in the frontal, insular, and temporal cortexes of DLB patients, as well as the hypoperfusion in the parietal and temporoparietal cortexes of AD patients. 20 Another article revealed that the temporal cerebral blood flow in DLB patients remained unchanged. 21 Additionally, a reduced metabolic activity in the frontal and occipital lobes is observed in both DLB and AD, although more reduced in the former 22,23 Therefore, it is necessary to focus on these different findings to better understand the relatively uniform damage of brain regions.
Growing evidence suggests that neurodegenerative diseases are caused by brain network dysfunction rather than the dysregulation of an isolated brain region. 24,25 Local brain regions that are selectively damaged act as "nodes" in functional networks, representing the basis of the network degradation hypothesis. 26 Brain network abnormalities detected in patients with DLB are predominantly described in the default mode network (DMN), 27 frontal-parietal network (FPN), 28,29 basal ganglia network, 30,31 and visual network (VIS). 32 Therefore, functional meta-analytic connectivity models (fMACM) 33 should be further constructed based on locally convergent brain regions. This might allow to test the network degradation hypothesis in DLB and evaluate whether the regional degeneration in DLB reflects distinct human neural network architecture. Patterns of the involved neural networks might be used as predictors of disease-related changes, thus providing a reference for the development of novel therapies, such as transcranial magnetic stimulation for network regulation.
Anatomical/activation likelihood estimation (ALE) is a powerful coordinate-based meta-analysis allowing to quantify consistent imaging findings across studies. 34 The fMACM can be used to determine which brain regions are co-activated above chance, with a particular seed region. The whole-brain co-activation pattern can be regarded as a surrogate for functional connectivity (FC). 35,36 A previous meta-analysis investigated gray matter atrophy in DLB, but this investigation was limited to structural imaging. 37 Currently, there is no consensus on brain structure and function damage in DLB patients, and whether the functional neural networks are dependent on the affected brain regions.
In this work, a quantitative meta-analysis was performed to delineate the most affected brain regions in DLB patients to highlight the differences in imaging findings between DLB and AD. The fMACM technique was then used to identify the neural networks involved in the affected brain regions in DLB. According to previous studies, our hypothesis was that DLB is characterized by a convergent damaged brain regions compared with HCs and AD. Our specific expectation is to observe that the co-activated neural networks prominently include the DMN, FPN, and VIS. Finally, the applications of some of the promising novel imaging modalities in DLB were reviewed, which may provide further insights into DLB pathophysiology.

| Literaturesearchandstudyselection
The meta-analysis was preregistered on Prospero (registration number: CRD42020162018) and was conducted according to the PRISMA statement. 38 A systematic search was conducted on March 27, 2021, using PubMed, Web of Science, OVID, Science Direct, and Cochrane Library database using the following keywords: "Magnetic Resonance Imaging" [Mesh] OR "Positron-Emission Tomography" [Mesh] OR "Tomography, Emission-Computed, Single-Photon" [Mesh] OR MRI OR "magnetic resonance imaging" OR "imaging" OR "neuroimaging" OR "brain imaging" OR "gray matter" OR "white matter" OR "voxel-based morphometry" OR "VBM" OR "voxelwise" K E Y W O R D S dementia with Lewy bodies, neuroimaging, anatomical/activation likelihood estimation, coordinate-based meta-analysis, functional meta-analytic connectivity modeling OR "positron emission tomography" OR PET OR "single photon emission computed tomography" OR SPECT AND Lewy OR "Lewy Body Disease" [Mesh] (Table S1). The reference lists of the eligible articles and relevant review articles were also screened to find potential additional studies. Authors not providing the necessary data were contacted to obtain clarification regarding the missing or unclear information.
The original studies included in this work were based on the following criteria: (1) they were published in English with peer review; (2) they report structural and functional neuroimaging changes related to the comparison between DLB patients and HCs (DLB-HCs), or comparison between DLB and AD (DLB-AD); (3) they report the whole-brain results in three-dimensional coordinates (x, y, z) in standard reference space (Talairach/Montreal Neurological Institute, MNI); and (4) they report the statistical significance. Structural imaging refers to the whole-brain analysis using Voxel-based morphometry (VBM). The functional imaging included the fludeoxyglucose positron emission tomography (FDG-PET) and single-photon emission computed tomography (SPECT). If the data from one study overlapped with those of another study, the largest group was selected for our meta-analysis.
Studies with one of the following parameters were excluded: (1) the necessary data could not be obtained; (2)

| Dataextraction
The three-dimensional coordinates, literature basic information, demographic data, and the experimental and imaging details were extracted from the eligible articles. Then, any coordinate (focus) reported in Talairach

| Qualityassessment
The quality of the included studies was assessed using a 12-point checklist ( Table S2). The checklist focused on three aspects in each study: (1) clinical and demographic characteristics of the samples; (2) imaging-specific methodology; and (3) standardization of the results and conclusions. This checklist was based on previous metaanalysis studies, 40

| Jackknifesensitivityanalysis
After the ALE analysis, a jackknife sensitivity analysis was performed by iteratively repeating the same analyses, but one dataset each time was excluded to test the replicability of the results across studies. 45 driven by specific studies that were ignored, thus compromising the robustness against spurious findings.

| Fail-safeNanalysis
Traditional detection methods including size meta-analysis are not suitable for the ALE method in order to consider the possibility of publication bias. 48 Therefore, the potential publication bias in this study was evaluated by a post hoc noise simulation, which was referred to a modified version of the fails-safe N (FSN) method. 49 It was applied for the estimation of the robustness against unpublished neuroimaging findings. A recent study using the data from BrainMap provides evidence for the existence of a file drawer effect, with the rate of missing contrasts estimated as at least 6 per 100 reported. 50 Therefore, the convergence meta-analysis was retested starting with an additional 6% noise to evaluate the robustness of the identified clusters. The surviving clusters were then retested, with a noise rate of up to 30%, as in the previous study. 51 A flowchart providing a visual interpretation of the data extraction, ALE meta-analysis, and FMACM analysis is shown in

| FMACManalysis
The fMACM analysis used data derived from the BrainMap database (screened on April 16, 2021). 33,52 The key idea of fMACM is to identify co-activation patterns of each specific ROI. 53 In our fMACM, each significant ROI is derived from the above ALE meta-analysis. All experiments in the BrainMap database that reported group analyses of task-based activations of healthy subjects were first identified, and which featured at least one focus of neural activation in the respective seed. According to this, the ALE meta-analysis over the experiments was carried out yielding the whole-brain co-activation patterns for each ROI. The significance was evaluated using 1000 permutations, with a cluster-forming threshold of p < 0.001, and corrected with a cluster-level FWE threshold of p < 0.05. 54

F I G U R E 2 Anatomical/activation likelihood estimation (ALE) and FMACM flowchart. Pipeline showing the process of ALE and FMACM
analyses and the related software, Ginger ALE and Sleuth, leading to the brain converging regions and their co-activation regions. ① Data Extraction: Literature basic information, demographic data, experimental and imaging details and the 3D coordinates were extracted from eligible articles. ② ALE analysis: The main contrasts of interest were performed ALE analysis in MNI space using the Ginger ALE software, leading to the brain converging regions. ③ Sleuth: Create spherical ROIs of nodes using peak foci coordinates of the corrected results from ALE analysis. Then, seed individual ROIs in BrainMap's Sleuth to search functional database. Use MNI brain space. ④ FMACM Analysis: Ginger ALE software was used to perform FMACM analysis with appropriate and consistent thresholds to identify ALE meta-analysis-coactivated brain regions 3 | RE SULTS

| Studyinclusionandcharacteristics
The The general information of the eligible studies, image data acquisition equipment and parameters, statistical threshold, standard space, and quality scores is summarized in Table 1. The study information including sample size, demographic characteristics of the subjects, evaluation of the cognitive function, movement disorder, and diagnostic criteria is reported in Table 2.

| Regions with structural changes between DLB and HCs
Based on the structural analysis of DLB <HCs, no converging brain area was found after FWE correction. Atrophy of the right parahippocampal gyrus tended to converge in DLB patients (uncorrected, p < 0.001; Figure 3A, Table 3).

| Regions with functional changes between DLB and HCs
The functional analysis based on DLB <HCs showed that the reduced functional activity in DLB patients was mainly located in the bilateral inferior parietal lobule and right lingual gyrus ( Figure 3B, Table 3).

| Regions with structural changes between DLB and AD
The structural analysis based on AD <DLB showed that the local brain atrophy in the left medial temporal lobe (MTL) was more severe in AD patients compared with that in DLB patients. Peak cluster was found in the left parahippocampal gyrus ( Figure 3C, Table 3). No enough experiments were available to analyze DLB <AD (n = 35; 3 foci; 2 experiments).

| Regions with functional changes between DLB and AD
Based on the DLB <AD functional analysis, DLB patients had a tendency of lower metabolism in the right lingual gyrus compared with that in AD patients (uncorrected, p < 0.001; Figure 3D, Table 3). No enough experiments were available to analyze AD <DLB (n = 93; 11 foci; 3 experiments).

| Jackknifesensitivityanalysis
In this study, jackknife sensitivity analysis was performed on the corrected ALE results. To this end, 14 and 6 different ALE meta-analyses of "functional changes between DLB and HCs" and "structural changes between DLB and AD", respectively, were conducted. The sensitivity analysis revealed that the reduced functional activity of DLB patients in the right inferior parietal lobule was the most robust result, replicable in all the 14 datasets. The reduced functional activity in the right lingual gyrus and left inferior parietal lobule remained relatively highly replicable. This was due to the still significant value in the combination of at least 9 combinations of the datasets (Table S3). However, the less atrophy of the left parahippocampal gyrus in DLB patients compared with AD was a replicable result in only three studies (Table S4).

| Fail-safeNanalysis
The last column of Table 3 shows the fail-safe percentage of the additional noise that must be added to each meta-analysis to cause the convergence failure of previously determined clusters. Overall, the FSN assessment results were consistent with the jackknife sensitivity analysis. The most stable result was a decreased functional activity of the right inferior parietal lobule in patients with DLB. The reduction of functional activity in the right lingual gyrus and the left inferior parietal lobule remained relatively highly stable. This was due to the still significant value in the combination of more than 10% noise datasets. Moreover, the less atrophy in the left parahippocampal gyrus in DLB patients compared with AD remained a significant result with the addition of 33% noise (Table 3).  The co-activation brain regions of the left parahippocampal gyrus consisted of the bilateral parahippocampal gyrus, thalamus, and left hippocampus ( Figure 4D, Table 5).

| DISCUSS ION
This ALE meta-analysis is the first quantification of the location of cerebral changes across different imaging modalities in DLB.
In addition, it is the first application of fMACM to characterize co-activated neural networks associated with the damaged brain areas in DLB. This study found that the right parahippocampal gyrus atrophy in DLB patients tended to converge. Moreover, the functional activity of the bilateral parietal lobe and right occipital lobe significantly decreased compared with those in HC patients.
Structural differences between DLB and AD were preferentially concentrated in the left parahippocampal gyrus, and functional differences tended to converge to the right lingual gyrus. Furthermore, these convergent brain regions co-activated with extensive brain regions, covering multiple neural networks. These local convergent brain regions might be potential image markers of DLB damage or differentiation from AD. Moreover, they might be key "nodes" in those co-activated neural networks, forming the basis of the network degradation hypothesis. Represents the ratio to the number of experiments included in the meta-analysis.
These findings suggested that the co-activation patterns of these regions could be attributed to some recognized neural networks.
However, the fact that the neuromodulation of these neural networks can improve the cognitive and mental disorders in DLB group is yet to be explored.

| Localchangesandco-activation patternsofthedifferencesbetweenDLBandAD
The left parahippocampal atrophy was less in DLB patients than in AD patients, supporting the idea that the MTL of DLB was relatively

F I G U R E 4
Results of all fMACM co-activated brain areas. All results were FWE corrected with a cluster-forming threshold of p < 0.001 and cluster-level inference of 0.05. Results were superimposed on a brain template using MRIcron software in MNI space. Color bars represent anatomical/activation likelihood estimation scores. DLB, dementia with Lewy body, HCs, healthy controls, AD, Alzheimer's disease preserved. The parahippocampal gyrus is responsible for high-level neurological activities such as emotion, learning, and memory. It is also an important structure to ensure the normal hippocampal function. Its structural damage may cause abnormal emotional and cognitive behaviors. The volume of the MTL structure such as the parahippocampal gyrus was significantly reduced in AD patients due to the large amount of AD-type pathological deposition. 72 The loss of MTL gray matter is associated with memory impairment, even at a prodromal stage. 73 ALE meta-analysis studies reveal that AD structurally affects the (trans-) entorhinal, hippocampal regions and the TA B L E 5 Functional meta-analytic connectivity models (fMACM) co-activated brain areas FDG-PET occipital hypometabolism correlates with visual cortex neuropathology in DLB. 22 In addition, an autopsy-confirmed study suggested that the above correlation could distinguish DLB from AD with high accuracy. 81 The ALE meta-analysis of AD functional images showed that the hypoperfusion and hypometabolism in the parietal lobe (angular gyrus, supramarginal gyrus, and precuneus) 74 and posterior cingulate gyrus 75 were convergent compared with HCs. Our study found that patients with DLB had functionally affected bilateral inferior parietal lobules and right lingual gyrus.
Overall, these findings further demonstrated that decreased occipital activity is more frequently seen in DLB, while decreased temporal parietal activity is common in both AD and DLB. 79 This allowed the distinction between DLB and AD with a sensitivity of 90% and a specificity of 80%. 81 Furthermore, hypoperfusion in the right lingual gyrus precedes the hypoperfusion in the frontal and temporal cortices, underlining the changes in the early stages of the disease. 82 These aspects suggest that the measurement of the occipital metabolism/perfusion, even in the early stages of the disease, might be an informative diagnostic aid to distinguish DLB from AD. Thus, the combination of hippocampal volumes and occipital activity allows the distinction of DLB patients from those with AD with a higher level of accuracy. 83 The co-activated brain areas of the left parahippocampal gyrus involve the bilateral parahippocampal gyrus, hippocampus, and thalamus, which are mainly located in the DMN. 64 The DMN has an important role in several cognitive functions and includes the prefrontal cortex, bilateral parahippocampal gyrus, hippocampus, thalamus, inferior-lateral-parietal lobule, and precuneus. 84 Reduced DMN connectivity is associated with decreased memory performance, slower processing speed, and decreased executive function. [85][86][87] Alterations in the DMN are involved in a range of neurodegenerative disorders such as AD, Parkinson's disease, and frontotemporal dementia. [88][89][90][91] 106 Positive studies have been reported in premotor DLB with reduced uptake manifesting prior to reduced DAT uptake, indicating that 123I-MIBG scintigraphy may have an even greater role in early disease. 107 123I-MIBG scintigraphy was given an increased diagnostic weighting in the 2017 DLB consortium and is now considered an indicative biomarker.
In some cases, DLB pathology is characterized by amyloid protein (Aβ) and tau deposition in addition to α-synuclein aggregation. 108,109 Studies have shown significant increase in Aβ load in more than 80% of DLB patients. 110  have shown reduced FC in the extensive brain network of DLB subjects, and dyssynchrony of cortical and subcortical regions is associated with cognitive fluctuations. 120 In fact, one study, from a modeling perspective, detected significant differences in DFC in vision-related networks (ie, occipito-parietal lobe-frontal and medial occipito-frontal) and attention network (ie, right fronto-parietal control networks) in DLB patients compared with HC, suggesting that the interdependence between networks is reduced. These temporarily disconnected networks may be related to the pathogenesis of DLB. 121 Previous work of our research group found that DLB's dynamic functional connectivity variability and time allocation of clustering state sequences changed, which may lead to complex brain network dynamics disorder, and may make the brain lack integration and flexibility, resulting in ineffective brain function. 122 Overall, DFC is a promising approach to better understand the neurodegenerative process of DLB and to investigate new biomarkers for disease diagnosis and prognosis. At present, studies on DFC in patients with DLB are limited and it is difficult to draw consistent conclusions.
This report has some limitations. First, the heterogeneity of the study characteristics, including different data acquisition, preprocessing protocols, statistical methods, and threshold settings, could not be entirely ruled out. Second, the number of experiments included in each analysis was small. The coordinate-based metaanalysis was limited to the primary studies that convey all information in the format required for statistical processing. This means that the included literature was not comprehensive. However, the quantitative meta-analysis provides the most reliable results when performed correctly, because it provides statistically testable evidence for the convergence of the current literature. In addition, the sensitivity analysis, publication bias, and quality evaluation were carried out as a reference for the reliability and stability of the conclusions.
Our cautious idea was that the brain abnormalities of DLB should be included, but not be limited to the results reported in this work.

| CON CLUS ION
Overall, the present meta-analysis suggests that the alterations of the brain structure and function in DLB might be specific and significantly different from AD. Co-activated neural networks correspond to the FPN, VIS, and SN of HCs, suggesting that DLB might be abnormal in these networks. The integrity of the DMN in DLB patients provides a new observation to help in the clinical distinction between AD and DLB.
The identified brain regions or networks might serve as a framework for future quantitative analysis of per-subject image data. Such customized imaging indices might help the development of diagnosis, prognostic judgment, and targeted network regulation, thus improving the clinical management. However, a further study of the phenotype of DLB is necessary in order to comprehensively evaluate the neuroimaging features of DLB and its physiological significance. In addition, future early diagnosis and in-depth understanding of DLB, AD, and other types of dementia will likely rely on multimodal approaches, through a combination of the mature imaging and some of the promising novel imaging modalities, such as molecular imaging and novel functional imaging.

ACK N OWLED G M ENTS
This study was supported by the National Natural Science

CO N FLI C TO FI NTE R E S T
There are no conflicts of interest that need to be disclosed.

DATAAVA I L A B I L I T YS TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.