Cerebral perfusion correlates of domain-specific cognitive function in Alzheimer's disease and dementia with Lewy bodies

Abstract

Background

Differentiating Alzheimer's disease (AD) from dementia with Lewy bodies (DLB) remains challenging because of overlapping symptoms and frequent co-pathology. Although cerebral perfusion abnormalities are well described in both disorders, how domain-specific cognitive decline relates to regional perfusion patterns remains incompletely understood.

Objective

To investigate disease-specific associations between regional cerebral perfusion and domain-specific cognitive performance in patients with AD and DLB.

Methods

Amyloid-positive patients with AD and probable patients with DLB, each with at least one abnormal neuroimaging biomarker, underwent ¹²³I-iodoamphetamine single-photon emission computed tomography and Addenbrooke's Cognitive Examination III (ACE-III). Regional cerebral perfusion was compared among 57 patients with AD, 72 with DLB, and 44 healthy controls. Voxel-wise regression analyses examined associations between perfusion and ACE-III total and domain scores.

Results

Compared with controls, patients with AD showed reduced cerebral perfusion in the posterior cingulate, precuneus, and temporo-parietal cortices, whereas patients with DLB showed posterior cortical hypoperfusion with relatively preserved posterior cingulate perfusion. Direct comparisons revealed limited differences, including posterior cingulate and hippocampal hypoperfusion in patients with AD, and occipital hypoperfusion in patients with DLB. Cognitive decline in patients with AD correlated with left-lateralized temporo-parietal hypoperfusion across multiple domains, with cerebellar and sensorimotor hyperperfusion. Patients with DLB showed limited perfusion–cognition coupling, except for robust associations between bilateral posterior regions and visuospatial performance.

Conclusions

Distinct patterns of cerebral perfusion-cognition relationships characterize AD and DLB, reflecting differential regional vulnerabilities underlying cognitive decline. These findings may contribute to improved understanding of the pathophysiology underlying cognitive decline and support disease monitoring.

Introduction

Dementia affects an increasing proportion of the global population, with an estimated 46.8 million individuals affected in 2015 and projections reaching 131.5 million by 2050. ¹ Among these disorders, Alzheimer's disease (AD) is the most common cause of dementia, accounting for 60–80% of cases.^2–4 Clinically, AD is characterized by progressive impairment in memory, executive function, language, and visuospatial abilities, ⁵ and neuropathologically by amyloid-β plaques, tau pathology, and frequent co-occurrence of vascular or α-synuclein pathology.^2,6,7

Dementia with Lewy bodies (DLB) is the second most common neurodegenerative dementia. DLB presents with fluctuating cognition, visual hallucinations, REM sleep behavior disorder, and parkinsonism, driven by widespread cortical and subcortical α-synuclein deposition.^8,9 Although AD and DLB have distinct pathological underpinnings, they share several overlapping clinical features—including cognitive decline, memory impairment, and neuropsychiatric symptoms—which contribute to frequent misdiagnosis.^4,6,7 Co-pathology further complicates clinical judgment: many clinically diagnosed patients with AD have Lewy pathology at autopsy, whereas a substantial portion of patients with DLB are initially labeled as having AD. ¹⁰

Differentiating AD from DLB is clinically essential because therapeutic responses and prognostic trajectories differ substantially. Patients with DLB exhibit marked neuroleptic sensitivity and derive greater benefit from cholinesterase inhibitors, whereas AD follows a more amnestic-dominant course with distinct management considerations. ⁸ Early and accurate discrimination is therefore critical for patient safety, appropriate treatment, and counseling.

Neuroimaging techniques, such as cerebral perfusion single-photon emission computed tomography (SPECT) and fluorodeoxyglucose (FDG) positron emission tomography (PET), play a central role in differential diagnosis. AD typically shows hypoperfusion or hypometabolism in the posterior cingulate, precuneus, and lateral temporoparietal cortices, ⁴ whereas DLB is characterized by posterior cortical and occipital involvement, including the so-called “cingulate island sign.”^9,11 However, these group-level patterns alone do not capture the heterogeneity of cognitive impairment across domains, and the relationship between cognitive performance and regional perfusion remains incompletely understood—particularly when comparing AD and DLB within the same analytical framework. Cognitive composites such as the Addenbrooke's Cognitive Examination III (ACE-III) provide domain-specific assessment across attention, memory, fluency, language, and visuospatial ability, making them well suited to investigate structure–function relationships relevant for differential diagnosis.^12,13

Given these diagnostic challenges and the need to understand how domain-specific cognitive decline relates to regional perfusion in each disorder, the present study aimed to (1) identify characteristic perfusion patterns in AD and DLB using voxel-wise univariate analysis and (2) determine how reductions in ACE-III total and subdomain scores relate to cerebral perfusion across the brain.

This approach enables direct comparison of disease-specific neurocognitive networks and may clarify perfusion markers that contribute to distinguishing AD from DLB in clinical practice.

Methods

Participants

This retrospective study included patients with AD and DLB who underwent cerebral perfusion SPECT with N-isopropyl-p-iodoamphetamine (¹²³I-IMP) at the Department of Neurology, Chiba University Hospital. Patients with AD were clinically diagnosed as probable AD according to the 2011 NIA-AA criteria, ¹⁴ and amyloid-β burden was confirmed either by cerebrospinal fluid examination ¹⁵ or amyloid PET using ¹⁸F-Flutemetamol or ¹⁸F-Florbetapir. Patients with DLB were diagnosed with probable DLB according to the 2017 consensus report of the DLB Consortium, ¹⁶ which included dopamine transporter (DAT) SPECT with ¹²³I-FP-CIT and/or myocardial scintigraphy with ¹²³I-metaiodobenzylguanidine (MIBG) as indicative biomarkers. In addition, it was confirmed that at least one of these imaging biomarkers showed abnormal findings. Among the 72 patients with DLB, DAT-SPECT was performed in 57 patients and MIBG scintigraphy in 61 patients. Abnormal findings were observed in 53/57 (93.0%) and 42/61 (68.9%), respectively. Patients who did not undergo DAT SPECT or MIBG were excluded. Patients treated with a cholinesterase inhibitor were excluded. Healthy controls (HC) were prospectively recruited and underwent ¹²³I-IMP SPECT and magnetic resonance imaging (MRI). The inclusion criteria were age between 50 and 80 years and no history of neurological, neuropsychiatric disorders, or severe head trauma. Board-certified neurologists confirmed normal findings through neurological examinations. The exclusion criteria were abnormal MRI findings and cognitive impairment, defined as an ACE-III score <80.

The recruitment of patients and healthy participants was approved by the Institutional Review Board (No. 2052 and No. 0068-23), and all protocols were conducted in accordance with the Declaration of Helsinki.

Clinical and cognitive data collection

For patients with AD and DLB, demographic and clinical information, including sex, age, years of education, disease duration, medical history, and neurological and neuropsychiatric symptoms, was retrospectively obtained from electronic medical records. HC were prospectively enrolled, and the same demographic variables were recorded at the time of study entry.

All participants underwent the Mini-Mental State Examination (MMSE). ACE-III data were available in 27/57 patients with AD, 54/72 patients with DLB, and 28/44 HCs. The ACE-III evaluates five cognitive domains: Attention, Memory, Fluency, Language, and Visuospatial abilities. MMSE scores were available for all participants, whereas ACE-III data contained missing values for both patient groups. For patients with AD and DLB, cognitive assessments were performed within 3 months of the ¹²³I-IMP SPECT scan. For HC, cognitive testing and SPECT imaging were conducted within approximately 1 month of each other. Cognitive assessments were conducted by certified clinical psychologists who were blinded to the clinical diagnosis and imaging findings.

Image acquisition and preprocessing

Perfusion SPECT was performed using ¹²³I-IMP. After intravenous administration of 111 MBq of ¹²³I-IMP, the participants rested for approximately 5 min, followed by a 24-min acquisition. Two SPECT/CT systems from the same vendor (GE Healthcare, USA) were used during the study period: Infinia Hawkeye 4 (used until December 2020) and Discovery NM/CT 870 DR (introduced in January 2021). Both scanners employed an extended low-energy general-purpose collimator and an identical continuous-rotation protocol (8 cycles × 3 min, 4° angular steps), yielding a 64 × 64 matrix with an axial slice thickness of 4.42 mm. When clinically appropriate, antiparkinsonian medications were withheld for at least 12 h before imaging.

Reconstruction parameters were kept constant across scanners and included a Butterworth filter (cutoff frequency 0.5 cycles/cm; order 8). Attenuation correction was performed using Chang's method with a linear attenuation coefficient of 0.07.

Raw DICOM data were converted to the NIFTI format using dcm2nii (MRIcroGL), and preprocessing was performed with SPM12 (Wellcome Centre for Human Neuroimaging, UK) implemented in MATLAB R2020a (MathWorks, USA). Images were spatially normalized using the OldNorm procedure and smoothed with a 14-mm full-width at half-maximum Gaussian kernel. All datasets were visually inspected to confirm successful preprocessing. To minimize interindividual variability in tracer uptake, voxel intensities were normalized by proportional scaling to the global mean count.

Statistical analysis

Demographic and clinical variables were summarized as proportions or as mean ± standard deviation (SD). Group differences were evaluated using Fisher's exact test for categorical variables and one-way analysis of variance (ANOVA) for continuous variables, followed by Tukey's HSD post hoc test (α = 0.05).

Voxel-wise comparisons of perfusion images were conducted in SPM12 using two-sample t-tests for each pairwise contrast among the three diagnostic groups (AD versus HC, DLB versus HC, and AD versus DLB). Associations between regional perfusion and ACE-III total or subdomain scores were examined separately within the AD and DLB groups using multiple regression models. As ACE-III data contained missing values in both patient groups, all voxel-wise regression analyses examining the associations between cerebral perfusion and ACE-III scores were performed using participants with available data. No imputation was performed. Age, sex, and scanner type were included as nuisance covariates in all imaging analyses. A gray matter mask derived from the SPM12 tissue probability map (thresholded at 95% probability) was applied as an explicit mask. A voxel-level threshold of p < 0.001 (uncorrected), combined with an extent threshold of >30 voxels, was primarily used to identify clusters of interest. This threshold was selected because the analyses were exploratory and intended to characterize broad perfusion patterns rather than test a priori region-specific hypotheses. To ensure robustness and avoid overinterpretation of threshold-dependent findings, voxel-wise and cluster-wise family-wise error (FWE)-corrected p-values were also reported for all clusters that survived the initial threshold but were not used as the primary criterion for statistical inference.

Results

Participant characteristics

A total of 57 patients with AD, 72 patients with DLB, and 44 HC were included in the analysis. Demographic and neuropsychological characteristics are summarized in Table 1. The proportion of women did not differ among the groups (AD 54.4%, DLB 41.7%, HC 50.0%; p = 0.335). Years of education were comparable across the groups (AD 13.4 ± 2.5, DLB 13.2 ± 2.8, HC 13.3 ± 2.1; p = 0.87). Age differed among groups (AD 70.9 ± 9.9 years; DLB 76.4 ± 7.4 years; HC 66.8 ± 9.3 years; p < 0.001), with post hoc tests showing that patients with DLB were older than patients with AD (p = 0.002) and HC (p < 0.001). Disease duration did not differ between AD and DLB (3.1 ± 3.0 versus 4.0 ± 3.4 years; p = 0.11). MMSE scores were higher in HC (29.5 ± 0.7) than in patients with AD (20.2 ± 5.7) and DLB (21.9 ± 5.9) (p < 0.001). MMSE scores did not differ between the AD and DLB groups (p = 0.14). The HC group showed higher ACE-III total scores (93.7 ± 4.7) than the AD (62.4 ± 18.9) and DLB (63.4 ± 20.1) groups (p < 0.001), with no difference between the two patient groups (p = 0.97). Across ACE-III domains, HC consistently outperformed AD and DLB in Attention (p = 0.007), Memory (p < 0.001), and Fluency (p < 0.001). No differences were observed between AD and DLB in any subdomain (p > 0.05). The proportion of participants scanned with the newer SPECT system differed across groups (AD 35.1%, DLB 51.4%, HC 63.6%; p = 0.018), with post hoc testing showing a difference between AD and HC (p = 0.016).

Table 1.

Patient demographics.

	Alzheimer's disease (n = 57)	Dementia with Lewy bodies (n = 72)	Healthy controls (n = 44)	Group comparison (p-value)	Post hoc p-value a: AD vs. DLB b: AD vs. HC c: DLB vs. HC
Women (%)	54.4	41.7	50.0	0.335
Age (years)	70.9 ± 9.9	76.4 ± 7.4	66.8 ± 9.3	<0.001*	a = 0.002*, b = 0.061, c < 0.001*
Education (years)	13.4 ± 2.5	13.2 ± 2.8	13.3 ± 2.1	0.869
Disease duration (years)	3.1 ± 3.0	4.0 ± 3.4	-	0.112
MMSE	20.2 ± 5.7	21.9 ± 5.9	29.5 ± 0.7	<0.001*	a = 0.138, b < 0.001*, c < 0.001*
ACE-III total	62.4 ± 18.9	63.4 ± 20.1	93.7 ± 4.7	<0.001*	a = 0.971, b < 0.001*, c < 0.001*
Attention	12.6 ± 3.6	12.6 ± 3.8	17.7 ± 0.8	0.007*	a = 1.00, b = 0.008*, c = 0.005*
Memory	10.1 ± 5.3	13.3 ± 7.2	24.8 ± 2.0	<0.001*	a = 0.111, b < 0.001*, c < 0.001*
Fluency	7.6 ± 3.2	5.8 ± 3.3	12.5 ± 1.3	<0.001*	a = 0.050, b = 0.003*, c < 0.001*
Language	20.7 ± 7.2	21.1 ± 5	25.8 ± 0.4	0.131
Visuospatial	11.0 ± 4.8	10.4 ± 4.3	15.0 ± 0.9	0.058
New Scanner (%)	35.1	51.4	63.6	0.018*	a = 0.227, b = 0.016, c = 0.745

Values are shown as the mean ± standard deviation or as percentages. Group comparisons were tested using one-way ANOVA with Tukey's HSD post hoc tests or Fisher's exact test with Bonferroni correction for post hoc pairwise tests. The ACE-III comprises five subdomains: Attention, Memory, Fluency, Language, and Visuospatial. A p-value <0.05 was considered statistically significant. AD: Alzheimer's disease; DLB: dementia with Lewy bodies; HC: healthy controls; MMSE: Mini-Mental State Examination; ACE-III: Addenbrooke's Cognitive Examination III.

Group comparison of cerebral perfusion images

Group comparisons showed broadly overlapping posterior cortical hypoperfusion and cerebellar hyperperfusion in AD and DLB compared with HC, whereas direct comparison between AD and DLB revealed only limited differences. Full cluster details are listed in Supplemental Table 1.

AD versus HC. Compared with HC, patients with AD showed widespread cortical hypoperfusion, mainly in the bilateral temporo-parietal and occipital association cortices (Figure 1A). These bilateral temporoparietal clusters, together with the left frontal cluster, survived cluster-level FWE correction (p < 0.001). The largest cluster was located in the right angular gyrus, extending into the middle and superior temporal gyri, middle occipital gyrus, and superior parietal lobule (peak T = 8.93; MNI 52, −48, 34; 6406 voxels). A similarly large cluster was observed in the left hemisphere, encompassing the middle temporal gyrus, angular gyrus, superior parietal lobule, superior temporal gyrus, and inferior temporal gyrus, with additional bilateral involvement of the precuneus and cingulate region (peak T = 8.89; MNI −42, −62, 26; 18,952 voxels). Patients with AD also showed frontal hypoperfusion, including the left middle, inferior, and superior frontal gyri and the precentral gyrus (peak T = 6.31; MNI −40, 22, 42; 3979 voxels), and a smaller right frontal cluster involving the middle and superior frontal gyri. Several smaller clusters were detected in deep and medial structures, including the caudate nucleus, thalamus, cingulate region, medial frontal gyrus, and the uncus and parahippocampal gyrus.

Figure 1.

Group comparisons of cerebral perfusion among Alzheimer's disease, dementia with Lewy bodies, and healthy controls. Voxel-wise two-sample t-tests were performed in SPM12 to compare regional cerebral perfusion between groups (voxel-level uncorrected p < 0.001, cluster size >30 voxels). (A) Alzheimer's disease (AD) vs. healthy controls (HC): regions of hypoperfusion in AD are indicated by blue-cyan, and regions of hyperperfusion are indicated by red-yellow. (B) Dementia with Lewy bodies (DLB) vs. HC: decreases relative to HC appear in blue-cyan and increases appear in red-yellow. (C) AD vs. DLB: hypoperfusion in AD relative to DLB is displayed in yellow, and reductions in DLB relative to AD are displayed in red. Age, sex, and scanner type were included as nuisance covariates. Numbers below each image indicate the z-coordinate of the axial slice in the Montreal Neurological Institute space. L: left.

Hyperperfusion relative to HC was observed in the cerebellum, brainstem, and medial and lateral occipitotemporal regions (peak T = 9.10; MNI −10, −62, −22; 20,248 voxels), as well as in the right-dominant sensorimotor cortices (peak T = 7.48; MNI 22, −20, 56; 3852 voxels). Both clusters survived cluster-level FWE correction (p < 0.001).

DLB versus HC. Compared with HC, patients with DLB showed posterior cortical hypoperfusion, prominently involving the angular gyrus, middle temporal gyrus, middle occipital gyrus, superior parietal lobule, and precuneus (Figure 1B). The posterior cluster and both frontal clusters survived cluster-level FWE correction (all p ≤ 0.004). The largest cluster encompassed the bilateral angular and middle occipital regions as well as the precuneus (peak T = 7.61; MNI −62, −44, 34; 10,273 voxels). These posterior deficits were symmetrical and involved higher-order association cortices that were strongly linked to visuospatial and attentional processing. Prominent reductions were also observed in frontal regions, including the left middle and inferior frontal gyri, bilateral superior frontal and medial frontal gyri, and the cingulate region (peak T = 6.23; MNI −44, 22, 22; 5279 voxels). A right-dominant frontal cluster involved the middle, superior, and inferior frontal gyri and the precentral gyrus (peak T = 4.66; MNI 36, 10, 62; 2209 voxels). Smaller clusters appeared in the left temporal and occipitotemporal cortices and in the left caudate nucleus and thalamus.

Relative hyperperfusion was observed in the cerebellum, brainstem, and medial and lateral occipitotemporal cortices (peak T = 9.37; MNI −8, −64, −24; 15,991 voxels). Additional clusters of hyperperfusion were noted bilaterally in the precentral and postcentral gyri and the superior and medial frontal gyri (peak T = 5.59; MNI −10, −26, 62; 4885 voxels). Both clusters survived cluster-level FWE correction (p < 0.001). A smaller right-lateralized cluster was observed in the putamen, insula, and globus pallidus (peak T = 3.88; MNI 30, 2, −8; 436 voxels).

AD versus DLB. Compared with patients with DLB, patients with AD showed bilateral hypoperfusion involving the cingulate region and superior parietal lobule, extending into the right precuneus (peak T = 4.68; MNI 4, −42, 30; 793 voxels) (Figure 1C). This peak survived FWE correction (p = 0.012). Additional clusters showed left temporal–limbic reductions, including the hippocampal formation, parahippocampal gyrus, superior temporal gyrus, and middle temporal gyrus (peak T = 4.37; MNI −30, −8, −26; 597 voxels), whose peak survived FWE correction (p = 0.035). A symmetric right-sided temporal–limbic cluster was also identified, involving the inferior and middle temporal gyri, superior temporal gyrus, parahippocampal gyrus, hippocampal formation, and lateral occipitotemporal cortex (peak T = 4.25; MNI 36, −6, −28; 537 voxels).

Compared with patients with AD, patients with DLB showed right cuneus hypoperfusion (peak T = 3.34; MNI 20, −70, 18; 52 voxels).

Cerebral perfusion correlates with ACE-III scores

Regression analyses showed distinct regional perfusion correlates of ACE-III total and domain scores in AD and DLB. Positive correlations are shown in Figure 2, negative correlations are shown in Supplemental Figure 1, and full cluster details are listed in Supplemental Tables 2 and 3.

Figure 2.

Associations between Addenbrooke's cognitive examination III domain scores and regional cerebral perfusion in Alzheimer's disease and dementia with Lewy bodies. Voxel-wise multiple regression analyses were used to evaluate the correlations between cerebral perfusion and ACE-III scores (voxel-level uncorrected p < 0.001, cluster size >30 voxels). Only positive correlations are displayed. Panels show: (A) Total score, (B) Attention, (C) Memory, (D) Fluency, (E) Language, and (F) Visuospatial performance. Binary maps of positive correlations are shown in yellow for AD and in red for DLB. Age, sex, and scanner type were included as nuisance covariates. Numbers below each image indicate the z-coordinate of the axial slice in the Montreal Neurological Institute space. L: left.

Total score

AD

Declines in ACE-III total performance were associated with left lateral temporal hypoperfusion, including the middle, superior, and inferior temporal gyri and the lateral occipitotemporal gyrus (peak T = 4.95; MNI −48, −6, −12; 1962 voxels), which survived cluster-level FWE correction (p = 0.011). Another hypoperfusion area involved the left middle occipital gyrus, angular gyrus, and superior parietal lobule (peak T = 4.17; MNI −40, −66, 28; 576 voxels).

In contrast, lower total scores were associated with bilateral cerebellar hyperperfusion (peak T = 5.17; MNI −14, −42, −32; 1108 voxels) and in the right sensorimotor–parietal region encompassing the precentral and postcentral gyri and the angular gyrus (peak T = 4.96; MNI 46, −22, 54; 648 voxels). The bilateral cerebellar cluster survived cluster-level FWE correction (p = 0.049).

DLB

In DLB, worsening total scores were accompanied by a small hypoperfusion cluster in the left lateral occipitotemporal and middle temporal gyri (peak T = 3.75; MNI −38, −58, 0; 184 voxels) and the right middle and superior temporal cortex (peak T = 3.70; MNI 50, −42, 4; 505 voxels).

No clusters showed hyperperfusion associated with score decline.

Attention

AD

Attention decline was related to left temporal hypoperfusion, involving the middle, superior, and inferior temporal gyri (peak T = 4.59; MNI −48, −6, −14; 1043 voxels).

Score deterioration was also associated with hyperperfusion in the posterior fossa and occipitotemporal structures, including the cerebellum, medial occipitotemporal cortex, lingual gyrus, and cuneus (peak T = 5.45; MNI 6, −68, −38; 1923 voxels), and in a right sensorimotor cluster spanning the precentral and postcentral gyri (peak T = 4.68; MNI 42, −18, 52; 519 voxels). The peak voxel of the posterior fossa/occipitotemporal cluster survived peak-level FWE correction (p = 0.039).

DLB

In DLB, worse attention performance showed small hypoperfusion in the left lateral occipitotemporal cortex (peak T = 3.58; MNI −38, −58, 0; 104 voxels) and the left middle and inferior frontal gyri (peak T = 3.52; MNI −38, 18, 22; 43 voxels).

No clusters survived correction, and no hyperperfusion was observed.

Memory

AD

Memory decline was weakly associated with hypoperfusion in the right superior parietal lobule and precuneus (peak T = 4.02; MNI 8, −54, 40; 84 voxels), and in the right middle frontal gyrus (peak T = 3.63; MNI 28, 34, 38; 84 voxels).

Worse memory scores were also linked to modest hyperperfusion in the right precentral and postcentral gyri (peak T = 4.18; MNI 42, −22, 44; 187 voxels).

DLB

Memory decline was associated with hypoperfusion in the right lateral occipitotemporal cortex, extending into the hippocampal formation and parahippocampal gyrus (peak T = 4.46; MNI 42, −30, −16; 651 voxels).

No hyperperfusion was associated with memory decline.

Fluency

AD

Worsening fluency was associated with hypoperfusion in the left perisylvian–fronto-insular network involving the superior, middle, and inferior temporal gyri and the insula (peak T = 4.56; MNI −46, −8, −14; 446 voxels). Additional hypoperfusion areas included the left caudate nucleus, insula, middle and inferior frontal gyri, precentral gyrus, and thalamus (peak T = 4.54; MNI −28, 26, 14; 930 voxels).

Fluency decline was also accompanied by hyperperfusion in the bilateral medial occipitotemporal regions, lingual gyrus, cuneus, cerebellum (peak T = 5.45; MNI 4, −74, 4; 2939 voxels), and right sensorimotor–parietal cortex (peak T = 4.48; MNI 36, −22, 50; 496 voxels). The bilateral occipitotemporal/cerebellar cluster survived cluster-level FWE correction (p = 0.002).

DLB

DLB showed a small region in which fluency decline corresponded to left middle and inferior frontal cortical hypoperfusion (peak T = 3.43; MNI −38, 18, 24; 32 voxels).

No hyperperfusion was identified.

Language

AD

Language decline was associated with hypoperfusion in the left temporal and occipitotemporal cortices, including the middle and superior temporal gyri, lateral occipitotemporal gyrus, inferior temporal gyrus, and inferior occipital regions (peak T = 5.03; MNI −60, −42, −18; 2607 voxels). Another hypoperfusion area extended into the superior parietal lobule, angular gyrus, and middle and inferior occipital gyri (peak T = 4.34; MNI −38, −70, 26; 815 voxels). The left temporal–occipitotemporal cluster survived cluster-level FWE correction (p = 0.004).

Language deterioration also corresponded to hyperperfusion in the cerebellum (peak T = 4.72; MNI −14, −42, −30; 599 voxels) and the right precentral and postcentral gyri (peak T = 4.39; MNI 40, −24, 56; 364 voxels).

DLB

DLB showed a region in which language decline was associated with left temporal hypoperfusion (peak T = 3.91; MNI −62, −54, 16; 122 voxels).

No hyperperfusion was linked to language performance.

Visuospatial

AD

Worse visuospatial performance corresponded to hypoperfusion in a large left posterior temporo-parietal–occipital network (peak T = 7.28; MNI −48, −62, −14; 8503 voxels), involving the lateral occipitotemporal gyrus, superior parietal lobule, middle temporal gyrus, and middle and inferior occipital cortices. This cluster survived cluster-level FWE correction (p < 0.001).

Visuospatial decline was also associated with hyperperfusion in the bilateral cerebellum and in the right mesial temporal and subcortical structures, including the parahippocampal gyrus, uncus, putamen, inferior temporal gyrus, hippocampal formation, right orbitofrontal cortex, and thalamus (peak T = 6.38; MNI 6, −46, −38; 1342 voxels). This also survived cluster-level FWE correction (p = 0.030).

DLB

In DLB, worsening visuospatial performance was associated with large hypoperfusion in the bilateral posterior temporo-parietal-occipital cluster (peak T = 5.42; MNI 44, −60, 24; 13,578 voxels), encompassing the middle temporal gyrus, superior parietal lobule, superior temporal gyrus, middle occipital gyrus, angular gyrus, lateral occipitotemporal gyrus, and precuneus. This cluster survived cluster-level FWE correction (p < 0.001).

A small region of the left orbitofrontal cortex showed hyperperfusion with visuospatial decline (peak T = 4.17; MNI −18, 58, −20; 117 voxels).

Discussion

The present study demonstrated overlapping but partially different cerebral perfusion abnormalities in AD and DLB, together with disease-specific relationships between regional perfusion and domain-specific cognitive performance. Both disorders showed posterior cortical hypoperfusion relative to HC, whereas direct comparison revealed more limited differences involving the posterior cingulate, medial temporal, and occipital regions. In AD, cognitive decline was associated with widespread left-lateralized temporo-parietal hypoperfusion across multiple cognitive domains, whereas DLB showed more selective perfusion–cognition associations, most prominently in the visuospatial domain. These findings suggest that integrating cerebral perfusion imaging with domain-specific cognitive assessment may help characterize disease-related network dysfunction in AD and DLB.

The current results show that HC outperformed patients with AD and DLB in attention, memory, and fluency, with no significant differences between AD and DLB in these domains. This is in line with studies indicating that, although DLB is typically associated with greater deficits in attention, executive function, and visuospatial abilities, and AD with more pronounced memory impairment, these distinctions can blur in moderate to advanced stages.^17,18 For example, Kang et al. found that patients with DLB had worse attention, visuospatial, and executive function but better memory than patients with AD in early stages, but these differences diminished as dementia severity increased. ¹⁸ The lack of significant differences in disease duration and cognitive scores between AD and DLB highlights the diagnostic challenge. Recent research emphasizes the importance of early and comprehensive neuropsychological assessment, as well as the use of biomarkers (e.g., SPECT, PET, CSF analysis) to improve diagnostic accuracy.^19–21

This imaging analysis identified differences in cerebral perfusion among AD, DLB, and HC and showed how domain-specific cognitive decline, as measured by the ACE-III, corresponds to distinct regional perfusion abnormalities. AD showed widespread hypoperfusion in the bilateral temporo-parietal association cortices and precuneus, whereas DLB was characterized by prominent posterior cortical involvement with preserved cingulate perfusion. In AD, lower scores across multiple domains were strongly associated with hypoperfusion in the left temporal, temporoparietal, and occipitotemporal regions, with disease-related cerebellar and sensorimotor hyperperfusion. In contrast, DLB showed weaker and spatially restricted associations, with meaningful perfusion–cognition coupling emerging primarily in the visuospatial domain, which involved a large bilateral posterior temporo-parietal–occipital network.

In the group comparison, patients with AD showed marked hypoperfusion in the posterior cingulate cortex, precuneus, temporoparietal association cortices, and frontal regions compared with HC. This distribution is consistent with FDG-PET and cerebral perfusion SPECT studies reporting early and prominent functional decline in the posterior cingulate and precuneus along the AD continuum,^22–24 with meta-analytic evidence confirming their sensitivity as early imaging markers.^25,26 In addition, cerebellar and sensorimotor hyperperfusion was observed in regions typically spared in early AD. Recent studies with arterial spin labeling have shown similar cerebellar and subcortical hyperperfusion alongside characteristic hypoperfusion in the default mode and temporoparietal networks,^27,28 interpreted as compensatory activity or altered metabolic demand as the association cortices fail.^29,30 The emerging recognition of cerebellar contributions to cognitive processing further supports this interpretation.

These findings align with the Braak framework, in which pathology progresses from transentorhinal and limbic areas to multimodal association cortices, including the posterior cingulate and temporoparietal regions, before affecting primary cortices.^22,31 The combination of hypoperfusion in association cortices and relative sparing or compensatory hyperperfusion in primary cortices and cerebellar territories is therefore consistent with hierarchical network vulnerability as AD pathology advances.^27,28

Patients with DLB predominantly showed parietal, temporal, and frontal hypoperfusion, together with relative increases in the cerebellum, brainstem, basal ganglia, and sensorimotor regions. These findings accord with prior evidence of posterior cortical involvement in DLB,^32,33 including occipital and parietotemporal reductions and relative preservation of the posterior cingulate cortex—the cingulate island sign.^9,11 Although occipital hypoperfusion is a traditional hallmark, several studies show that its prominence can vary with disease stage, phenotype, and co-pathology.^34,35 The relative cerebellar and subcortical hyperperfusion, increasingly reported in arterial spin labeling studies, may reflect neurovascular compensation within cortico-striato-thalamo-cortical and cerebello-thalamo-cortical loops disrupted by α-synuclein pathology, ³⁶ or altered baseline activity related to attentional fluctuation. ³⁷ Frontal and parietal hypoperfusion is consistent with vulnerability of higher-order attentional and visuospatial networks in DLB, reflecting α-synuclein–related degeneration. ³⁸

Direct comparison of AD and DLB revealed posterior cingulate hypoperfusion in AD and right occipital hypoperfusion in DLB,^39,40 consistent with the cingulate island sign.^11,22,40 In addition, AD showed reduced hippocampal perfusion compared with DLB, which likely reflects the greater degree of hippocampal atrophy in AD rather than a primary functional abnormality. ⁴¹ Given that AD and DLB frequently coexist with mixed pathology, cerebral perfusion findings should be interpreted with this background in mind. ⁴² Nevertheless, the observed relative perfusion differences suggest differential vulnerability of limbic–posterior cingulate regions in AD and posterior visual systems in DLB, providing a clinically relevant contrast between the two disorders. ⁴²

In AD, global cognitive decline was linked to hypoperfusion in the left lateral temporal and posterior temporoparietal association cortices—regions supporting semantic processing, language, and multimodal integration, which are known to be vulnerable early in the disease.^43,44 This pattern is consistent with extensive evidence showing that posterior cingulate–precuneus and temporo-parietal dysfunction are major contributors to global cognitive deterioration in AD. ⁴⁵ In DLB, total score decline was associated with small clusters of bilateral occipitotemporal hypoperfusion, with no accompanying hyperperfusion. This weak perfusion–cognition coupling likely reflects substantial inter-individual variability, with different cognitive deficits contributing to total score decline across patients.^16,46,47

In AD, attention decline was associated with left temporal association hypoperfusion and cerebellar, occipitotemporal, and right sensorimotor hyperperfusion, closely mirroring the pattern observed in the ACE-III total score. This similarity suggests that decreased attention exerts a pervasive influence on performance across multiple cognitive domains and may partly underlie the clinical efficacy of cholinesterase inhibitors in global cognition.^48,49 In DLB, attention decline was associated with small clusters in the left frontal and occipitotemporal regions. This limited association suggests that attentional disturbances in DLB reflect cognitive fluctuations arising from large-scale network instability rather than regional hypoperfusion, consistent with the interpretation for the total score.^46,47

In AD, memory decline showed only weak associations with cerebral perfusion, limited to small clusters in the right precuneus, middle frontal gyrus, and right sensorimotor cortex, and did not involve the hippocampus. This absence is consistent with known methodological and biological considerations: hippocampal perfusion is strongly confounded by atrophy, partial volume effects, and low spatial resolution in SPECT, and the ACE-III Memory domain incorporates verbal and executive components that rely on broader frontoparietal networks rather than hippocampal-dependent episodic memory alone.^43,50 Importantly, although medial temporal and limbic perfusion was significantly reduced in AD compared with DLB at the group level, no significant association between regional perfusion and memory performance was observed in AD. This apparent dissociation suggests that once neurodegeneration is advanced, memory impairment in AD may become partially decoupled from regional perfusion measures. In this context, hypoperfusion may reach a floor effect, such that interindividual perfusion variability no longer directly translates into variability in cognitive performance. More broadly, memory decline in AD may be driven by downstream network disintegration beyond regional perfusion deficits alone. ⁵¹ In contrast, medial temporal structures are relatively preserved in DLB, ⁴¹ which may allow changes in regional perfusion to remain more directly linked to memory performance. This interpretation is supported by the observation that in DLB, memory decline was associated with right lateral occipitotemporal hypoperfusion extending into the hippocampal and parahippocampal regions. Notably, the precuneus cluster identified in the perfusion–memory analysis in AD overlapped with a region that also showed hypoperfusion in AD relative to DLB in the group comparison. As a core hub of the default mode network, the precuneus reflects a characteristic site of AD-related vulnerability, and its appearance in both analyses suggests posterior cingulate–precuneus network dysfunction as a key substrate of memory decline in AD. ⁵¹ In contrast, this region is relatively preserved in DLB, consistent with the cingulate island sign.^9,11 These findings support a model in which AD-related memory impairment is more tightly linked to default mode network hub dysfunction, whereas DLB-related memory decline emerges when posterior associative and medial temporal regions within a broader posterior network are affected.

In AD, declines in fluency were associated with hypoperfusion in the left frontal and subcortical regions, including the middle and inferior frontal gyri, insula, and caudate nucleus, together with anterior temporal areas. This pattern aligns with the left frontal–striatal and anterior temporal networks supporting phonemic and semantic fluency, which are known to be vulnerable early in AD. ⁵² In DLB, fluency decline was associated with only small clusters of hypoperfusion in the left middle and inferior frontal gyri, which are regions known to support executive and phonemic aspects of verbal fluency. ⁵³

Language decline in AD was linked to temporo-parietal hypoperfusion, including the middle and superior temporal gyri, lateral occipitotemporal cortex, and angular region. These areas constitute core posterior perisylvian language networks that support lexical–semantic processing and are frequently affected in early AD.^54–56 In DLB, language decline was associated with small clusters of hypoperfusion in the left middle and superior temporal gyri, regions supporting lexical–semantic and auditory language processing. ³⁵

AD and DLB exhibited strong associations between visuospatial performance and perfusion in posterior temporo-parietal–occipital regions, highlighting the centrality of these areas for visuospatial cognition.^43,46,47 However, the laterality and extent differed markedly between disorders. In AD, correlations were left-lateralized and localized to a posterior temporo-parietal–occipital network supporting integrative and constructive visuospatial functions. In DLB, visuospatial decline was associated with a much larger, right-biased bilateral posterior cluster encompassing the angular gyrus, precuneus, superior parietal lobule, lateral occipitotemporal cortex, and visual association areas. These findings align with the known posterior cortical vulnerabilities of the two disorders but reflect distinct network emphases.^46,47,57 The small orbitofrontal cluster with hypoperfusion in DLB suggests additional involvement of frontal–visual integration pathways.

Our results show that AD and DLB differ not only in perfusion patterns but also in the relationship between regional perfusion and ACE-III domains. In AD, strong, left-lateralized coupling between global, language, fluency, and visuospatial scores and hypoperfusion in the temporoparietal and perisylvian regions suggests that quantitative perfusion–cognition mapping could help monitor disease burden and progression. In DLB, the selective association between visuospatial scores and right-predominant posterior perfusion supports the use of visuospatial–posterior network coupling as a functional marker complementary to the cingulate island sign. Overall, combining cerebral perfusion imaging with domain-specific cognitive testing may improve differential diagnosis and phenotyping in routine clinical practice.

This study is cross-sectional and cannot determine whether perfusion–cognition coupling predicts future decline or treatment response. In addition, we did not perform explicit cross-validation across scanners, which may limit the generalizability of the findings. The spatial resolution of SPECT and partial volume effects limit sensitivity in small structures such as the hippocampus, and ACE-III subscores are not domain-pure; therefore, the observed associations likely reflect mixed cognitive processes. Residual confounding by factors such as vascular burden, medications, or mixed AD and DLB pathology cannot be excluded, and autopsy- or biomarker-confirmed cohorts are needed to validate and refine these findings. In addition, ACE-III data were incomplete in both patient groups, which reduced the sample size available for voxel-wise regression analyses.

Conclusion

AD and DLB differ in how cognitive decline maps onto regional perfusion: AD shows robust, left-lateralized associations between multiple ACE-III domains and hypoperfusion in temporoparietal and perisylvian regions, whereas DLB exhibits limited coupling, except for a prominent right-posterior visuospatial network. These patterns reflect distinct network vulnerabilities and cognitive architectures in the two disorders. Cerebral perfusion–cognition mapping using widely available SPECT may therefore provide useful, complementary information for differential diagnosis and understanding of disease-specific network pathology.

Data availability statement

The datasets generated and analyzed during the current study are not publicly available due to institutional restrictions and the nature of patient-derived data. However, de-identified data may be provided by the corresponding author upon reasonable request.