The global prevalence of diabetes has increased by 31% in the past decade, and now affects 435 million people.1 A major macrovascular complication of diabetes is peripheral arterial disease (PAD), which affects >200 million people worldwide, leading to significant morbidity and healthcare costs.2–4 PAD results from atheroma plaque build-up in the lower limb arteries, restricting blood flow and causing intermittent claudication, lower limb wound ulcers/infections, chronic limb-threatening ischaemia and even gangrene and lower extremity amputation.5 Effective new treatments are needed to mitigate plaque inflammation and progression to prevent these complications.6
In diabetes, hyperglycaemia accelerates atherosclerosis by promoting endothelial dysfunction and vascular inflammation, impairs microvascular blood flow and drives atheroma formation.7–9 This contributes to earlier onset, faster progression and greater PAD severity.10–12 Understanding the interplay between arterial inflammation, plaque burden and arterial and microvascular blood flow provides insights into the mechanisms behind PAD and helps identify novel therapeutic targets.
Advances in imaging allow non-invasive assessments of PAD pathology. 18F-fluorodeoxyglucose (FDG) PET quantifies lower limb arterial inflammation; vascular ultrasound evaluates plaque burden and arterial blood flow; and MRI-based methods, such as blood oxygen level dependent (BOLD) and intravoxel incoherent motion (IVIM) imaging, assess microvascular perfusion in lower limb muscles.13–15 BOLD MRI serves as a functional surrogate for microvascular oxygenation by measuring changes in T2* signals, which reflect the ratio of oxyhaemoglobin to deoxyhaemoglobin in the capillary bed during reactive hyperaemia.13,14 Complementing this, IVIM MRI acts as a structural and perfusion surrogate that uses the pseudo-diffusion effect of blood to quantify microcapillary blood volume and velocity without the need for exogenous contrast agents.15 Together, these modalities offer a multidimensional view of both macrovascular and microvascular changes in PAD.
Despite these advances, gaps remain in how arterial inflammation, plaque burden and microvascular dysfunction interact in PAD.16 As such, the aim of the present study was to address this knowledge gap by using multimodality imaging to investigate the relationship between lower limb arterial inflammation, plaque formation, arterial stenoses and microvascular perfusion in patients with diabetes and PAD. This is the first multimodality imaging study to investigate such interplay in this patient population. We hypothesise that femoral arterial inflammation drives atheromatous plaque formation, leading to compromised arterial blood flow and impaired microvascular perfusion in the lower limbs.
Methods
Study Design
This cross-sectional study recruited nine stable patients with diabetes and PAD from three Singapore medical centres (Singapore General Hospital, National University Hospital and Khoo Teck Puat Hospital) between September 2023 and August 2024. Eligible participants were aged ≥21 years with stable PAD, defined by an ankle–brachial index (ABI) of <0.9 and/or toe–brachial index (TBI) of <0.7, and on stable type 2 diabetes therapy for ≥6 weeks before enrolment.
Following recruitment, all PAD participants underwent bilateral ABI and TBI measurements for arterial blood flow, lower limb duplex ultrasound for plaque burden and stenoses and lower limb 18F-FDG PET/MRI for arterial inflammation and microvascular perfusion. Ten healthy participants, recruited between February and May 2022, underwent lower limb MRI for assessment of microvascular perfusion. ABI and TBI were not measured in the healthy controls, but were deemed unnecessary based on professional health screening confirming the absence of PAD symptoms.
Peripheral Arterial Duplex Ultrasound Scan
All patients with diabetes and PAD underwent lower limb B-mode colour Doppler duplex ultrasound scans, performed by sonographers with >3 years of experience, to assess lower limb atherosclerotic lesions, atheroma burden and arterial blood flow (Figure 1 ). The scans focused on identifying arterial narrowing in the common and superficial femoral arteries of the more diseased leg, determined by the lower TBI and ABI measurements (Figure 1 ).
For non-occluded lesions with stenosis >50% (peak systolic velocity ratio >2.0), lesion length and diameter stenosis were measured.17,18 Fully occluded lesions were documented based on lesion length, because peak systolic velocity was not measurable. ABI and TBI were measured bilaterally unless clinically contraindicated (n=2; Supplementary Table 1 ).
18F-fluorodeoxyglucose PET/MRI Imaging
All participants with diabetes and PAD underwent simultaneous 18F-FDG PET/MRI scans using a 3-T Biograph mMR PET/MR scanner (Siemens Healthineers; Figure 1). MRI and PET scans were acquired in the lower limbs, following the circulation time and dosage recommendations from the Cardiovascular Committee of the European Association of Nuclear Medicine (Supplementary Methods).19 To ensure high-quality PET imaging and minimise glucose competition, all participants were required to have a blood glucose level <11.1 mmol/l before tracer injection.19
Two rounds of T1-weighted black blood sampling perfection with application-optimised contrasts using different flip angle evolution (T1w SPACE) sequences were acquired. The first provided the anatomical reference for femoral artery segmentation in 18F-FDG PET image analysis, whereas the second round focused on the mid-calf, where a 10 mm 2D slice was reconstructed at the calf’s widest diameter for the microvascular perfusion analysis.
Microvascular perfusion was assessed using IVIM and BOLD in sequence (Figure 1 ) because IVIM does not interfere with the sequential BOLD measurements. Images were acquired at the widest section of the mid-calf (Figure 1 ). For BOLD imaging, reactive hyperaemia was induced in the more diseased leg, if tolerable, by inflating an MRI-compatible cuff (Ima-x) to 50 mmHg above the participant’s systolic blood pressure for 5 minutes, followed by rapid deflation (Figure 2A). Detailed scan parameters are provided in the Supplementary Methods.
Healthy participants underwent a mid-calf T1w SPACE, BOLD and IVIM protocol unless intolerable (Figure 1 ). For consistency, BOLD imaging was performed on the left leg. The imaging results, particularly perfusion metrics, were reviewed by health professionals and compared with published normal ranges to validate control data.15
Image Analysis
18F-FDG uptake in the common and superficial femoral arteries was quantified for each axial segment using the maximum target-to-background ratio (TBRmax), calculated by normalising the maximum standardised uptake value (SUV) of the femoral arteries to the blood pool SUV in the femoral veins.19
Regions of interest (ROIs) encompassing femoral arteries and veins were delineated semiautomatically on whole-leg T1w SPACE scans using 3D Slicer software (version 5.2.2; https://www.slicer.org/) by an investigator (XW) and reviewed by a consultant physician (DJH) and an experienced radiographer (Figure 3A). PET images were aligned to T1w SPACE images, and corresponding ROIs and SUV measurements were performed using in-house MATLAB scripts (R2022a; MathWorks).
Segment-wise TBRmax values were aggregated to provide three participant-level metrics: whole-vessel mean TBRmax; (2) mean TBRmax of the 15 most-diseased segments (MDS) with the highest TBRmax (TBRmax [MDS]); and the number of active segments (AS), with a TBRmax ≥1.6 (Figure 3B).19 To adjust for differences in artery length, the number of AS was normalised to the total number of arterial segments.
For BOLD and IVIM scans, ROIs were drawn around five muscle groups (tibialis anterior, peroneal, soleus, gastrocnemius and deep posterior) on the 2D T1w SPACE slices, as in a previous study (Figure 2A).20 T2* signals from BOLD scans were normalised to baseline values and BOLD parameters were calculated using established methods (Figure 2A).20 IVIM perfusion fraction (fp), diffusivity (D) and pseudo-diffusivity (D*) were calculated using a two-step modelling approach.21 Additional methodological details are provided in the Supplementary Methods.
Statistical Analysis
Continuous variables are reported as the mean ± SD, whereas categorical variables are presented as absolute counts and percentages. Group differences were assessed using Fisher’s exact test for categorical variables and either Student’s t-test or the Wilcoxon signed-rank test for continuous variables, depending on data distribution. Univariate correlations were examined using Spearman’s correlation coefficients. Two-tailed p<0.05 was considered statistically significant.
Statistical analyses were performed using GraphPad Prism 10.2.0 (GraphPad Software; www.graphpad.com), Stata version 17.0 (StataCorp) and R version 4.4.1 (R Foundation for Statistical Computing).
Results
Participant Characteristics
Nine participants with diabetes and PAD and 10 healthy controls were recruited to the study (Table 1). Two PAD participants and one healthy control were excluded from the BOLD MRI scans due to cuff intolerance. Duplex ultrasound identified 10 lesions with diameter stenosis ≥50% in PAD participants, with no significant negative correlations between plaque burden and arterial blood flow (Supplementary Results, Supplementary Table 2 and Supplementary Figure 1).
Lower Limb Microvascular Perfusion
Both BOLD and IVIM imaging demonstrated compromised lower limb microvascular perfusion in PAD participants compared with healthy controls. BOLD scans revealed a longer time to peak (TTP), more attenuated maximum T2* (T2*max; both before and after normalisation), prolonged time to reach half ischaemic minimum and a greater gradual T2* gradient in participants with PAD compared with healthy controls (p<0.05 for all parameters; Figures 2B–2G). There were no significant differences in baseline T2* or normalised minimum T2* (T2*min) between participants with PAD and the healthy controls (p>0.05 for both; Table 2 and Figures 2B–2G ).
IVIM scans also revealed impaired lower limb microvascular perfusion, with lower fp and D*, and water molecule diffusion (D) in participants with diabetes and PAD (p<0.05; Table 2; Figure 2H–J). In pairwise comparisons, no significant differences were observed in fp or D* between the more and less diseased legs in participants with PAD (Supplementary Figure 2 ). Perfusion deficits demonstrated by both modalities were consistent across all calf ROIs (Supplementary Figures 3 and 4).
The BOLD perfusion parameter TTP was negatively correlated with IVIM-derived fp and D* in all participants (ρ=−0.36 for both) and PAD participants alone (ρ=−0.36 and ρ=−0.40, respectively; p<0.05 for both; Figure 2K ). A significant positive correlation was observed between T2*max and fp (ρ=0.24; p<0.05; Figure 2K), further validating IVIM imaging for the detection of microvascular deficits.
Interplay Between Lower Limb Arterial Inflammation, Plaque Volume and Blood Flow in Diabetic PAD
PET-derived inflammation parameters of bilateral common and superficial femoral arteries are summarised in Table 2. There were no significant correlations between PET inflammation parameters and duplex ultrasound-derived atheroma plaque characteristics (Figure 3C). Similarly, PET inflammation parameters did not significantly differ between the more and less diseased legs (Supplementary Results and Supplementary Figure 5 ).
However, TBRmax (MDS) showed a significant negative correlation with TBI (ρ=−0.57; p<0.05), indicating an inverse relationship between lower limb arterial inflammation and blood flow. In addition, BOLD-derived T2*min was negatively correlated with all PET-derived inflammation parameters, suggesting that PAD participants with greater arterial inflammation had lower T2*min values during the ischaemic phase of the BOLD scans (Supplementary Figure 6). There were no significant correlations between PET-derived parameters and other BOLD (T2*max and TTP) or IVIM (fp and D*) parameters.
Interplay Between Lower Limb Arterial Blood Flow and Microvascular Tissue Perfusion in Diabetic PAD
Lower limb arterial blood flow (ABI/TBI) was compared with microvascular perfusion assessed by BOLD and IVIM imaging in participants with diabetes and PAD. Bilateral ABI/TBI values confirmed at least moderate PAD in all participants with diabetes and PAD (Supplementary Table 1). There were significant positive correlations between normalised T2*max and both ABI and TBI (ρ=0.37 and ρ=0.43, respectively; p<0.05 for both) and a significant negative correlation between TTP and TBI (ρ=−0.37; p<0.05; Figure 2L). These results suggest that reduced lower limb arterial blood flow was associated with impaired microvascular perfusion measured by BOLD imaging. However, there were no significant correlations between IVIM-derived microvascular perfusion parameters and ABI/TBI (Figure 2L).
Discussion
This study is the first to use multimodality imaging to investigate the interplay between lower limb arterial inflammation, plaque formation, blood flow and microvascular perfusion in participants with diabetes and PAD. By integrating PET, BOLD, IVIM and duplex ultrasound, we uncovered significant correlations between microvascular perfusion and arterial blood flow, identifying the contributions of each imaging modality in understanding PAD pathophysiology.
Perfusion MRI modalities, BOLD and IVIM, effectively differentiated participants with diabetes and PAD from the healthy controls, underscoring the utility of these imaging modalities for detecting impaired lower limb perfusion. BOLD imaging captured oxygenation dynamics with prolonged TTP and reduced normalised T2*max in participants with PAD, reflecting lower perfusion reserve and microvascular capacity.15,22 This functional response represents the ability of the microvascular bed to reoxygenate tissue following a period of ischaemia. IVIM imaging provides complementary insights, demonstrating lower perfusion fraction and pseudo-diffusivity, indicating reduced capillary blood volume and impaired microvascular flow.23 BOLD imaging provides a functional assessment of oxygen delivery, whereas IVIM imaging models the intrinsic physical properties of the microcapillary network, specifically the volume fraction of blood (fp) and its velocity (D*). These findings align with those of previous studies on microvascular dysfunction in PAD, reinforcing the role of perfusion deficits in PAD pathophysiology.15,24
Our novel IVIM protocol, which reduced scan time by one-third while maintaining its correlation with BOLD measurements, enhances the feasibility of microvascular perfusion assessment in clinical settings.21 This innovation may improve patient compliance, offering a more efficient means for perfusion assessment.
BOLD-derived perfusion metrics showed significant correlations with arterial blood flow, measured by ABI/TBI, highlighting a link between artery and microvascular function in participants with diabetes and PAD.24–26 However, IVIM parameters did not correlate with ABI/TBI, potentially due to the difference in the physiological processes captured by each modality: BOLD captures haemoglobin oxygenation that depends on macrovascular flow, whereas IVIM models local water molecule diffusion and capillary perfusion, which is an intrinsic microvascular property less influenced by macrovascular flow.22,23 This distinction underscores the complementary roles of BOLD and IVIM imaging in detecting PAD-related perfusion deficits by simultaneously evaluating both macrovascular-dependent oxygenation and independent microcapillary integrity. Future research should determine the most clinically helpful modality and parameters for evaluating microvascular dysfunction in patients with diabetes and PAD.
Although PET imaging showed no significant differences in arterial inflammation between more and less ischaemic legs or had correlation with MRI perfusion parameters, TBRmax (MDS) was negatively correlated with TBI, indicating a potential association between increased arterial inflammation and reduced blood flow. This suggests that arterial inflammation could have an indirect effect on blood flow, even if it does not directly correlate with perfusion metrics. The stronger correlation with TBI than ABI may suggest an altered resistance in small vessels, influenced by upper leg arterial inflammation, that disproportionately affects distal pressures. The non-significant association of TBRmax (MDS) with ABI may also be due to the lower sensitivity of proximal pressure measurements, such as ABI, to PAD severity and systemic inflammation effects compared with TBI.27,28
The findings of this study partially support our study hypothesis, suggesting that although inflammation alone may not explain blood flow differences in diabetic PAD, compromised microvascular perfusion is effectively captured by BOLD and IVIM imaging. Non-significant correlations between PET-measured arterial inflammation and ultrasound-derived plaque measurements seem not to agree with findings in carotid and coronary artery atherosclerosis.29 This may reflect a unique relationship between arterial inflammation and plaque development in PAD, because an earlier small-scale radiopathological study reported no correlation between 18F-FDG PET signal and histology-derived plaque inflammation.30 This may be due, in part, to our inability to perform site-specific PET analysis on individual stenotic plaques.
As one of the first studies to simultaneously use PET/MRI and ultrasound in patients with PAD, our results underscore the importance of a multimodality approach, highlighting its potential for assessing the complex pathology of diabetic PAD and supporting both diagnostic and therapeutic strategies. Combining PET, ultrasound and MRI-based techniques allows a broad assessment of arterial inflammation, plaque burden and microvascular perfusion. This is especially relevant in diabetic PAD, where both inflammation and vascular dysfunction contribute to disease progression.10–12
Our findings suggest that BOLD and IVIM imaging are practical, non-invasive tools for evaluating microvascular perfusion deficits, with BOLD parameters significantly correlating with arterial blood flow. The novel IVIM protocol, with reduced scan time, enhances patient compliance, supporting its use in routine clinical practice. By integrating these imaging techniques into individualised patient profiles, clinicians can more effectively stratify PAD severity, monitor disease progression and tailor interventions targeting both inflammation and perfusion deficits. This multidimensional approach offers a promising pathway towards the personalised management of diabetic PAD, potentially improving patient outcomes.
This study has several strengths. Conducting all imaging in a single centre using a single-vendor platform minimised variability in scan acquisition and ensured consistency. However, broader validation in larger, more diverse cohorts is needed through multicentre collaborations or alternative recruitment strategies. Future research should focus on identifying imaging parameters most strongly associated with symptoms and prognostic outcomes to enhance clinical utility. In addition, investigations using novel molecular imaging tracers (e.g. 68Ga-DOTATATE or 18F-NaF) or assessing longitudinal dynamics and therapeutic response of imaging parameters could further support personalised management strategies.31–34 Further investigations into the interactions among arterial inflammation, plaque characteristics and perfusion in diabetic PAD are also needed to refine our understanding of PAD pathophysiology and improve targeted interventions.
Study Limitations
First, the small sample size limits the generalisability of our findings, primarily due to stringent inclusion criteria, logistical constraints and the complexity of advanced multimodality imaging. The technical challenges of IVIM, BOLD and PET imaging, particularly when combined, contributed to recruitment difficulties, a limitation common in studies using these techniques.15,35–37 Excluding some of the patients with PAD from BOLD imaging due to cuff intolerance may have introduced selection bias, reducing the power of group comparisons. Second, although ABI/TBI was used to recruit participants with PAD, it was not measured in the healthy controls. Although the healthy controls were carefully screened and had no comorbidities, the lack of ABI/TBI data limits direct comparisons. Despite this, the perfusion metrics of the healthy controls aligned with published normative values, supporting the validity of the control group.15 Third, the advanced age, compared with the healthy controls, and the comorbidities of patients with PAD may affect arterial inflammation and blood flow, potentially confounding the results and complicating data interpretation. Fourth, PET imaging sensitivity also varies, particularly in patients with diabetes, where systemic inflammation may obscure localised differences and altered glucose metabolism could affect 18F-FDG uptake and, thus, arterial inflammation estimates.19,38,39 In the present study, we strictly followed European Association of Nuclear Medicine guidelines by ensuring in-scan blood glucose levels remained below 11.1 mmol/l (7.6 ± 2.0 mmol/l) to mitigate the effects of hyperglycaemia on 18F-FDG uptake. Although this controlled for acute glycaemic impact, the long-term duration of diabetes and chronic glycaemic control (e.g. HbA1c) were not explicitly correlated with plaque burden, which remains an area for future investigation. Fifth, although cardiovascular comorbidities were recorded, specific cardiac function metrics, such as ejection fraction, were not measured. Given that cardiac output can influence peripheral microcirculation, this remains a potential confounder. Sixth, the present study focused on stable atheroma and chronic inflammation rather than acute or chronic thrombotic processes, which also contribute to PAD progression. Seventh, the potential impact of medial arterial calcification on ABI measurements and imaging signals was not explicitly evaluated, although TBI was used to mitigate calcification-related inaccuracies in patients with diabetes. Finally, due to the limitation of MRI spatial resolution and insufficient blood signal suppression, we could not delineate the artery segments distal to the knee or identify specific segments containing atherosclerotic lesions, and did not characterise infrapopliteal arterial stenosis or collateral circulation. Although granular spatial data were collected for each axial slice of the femoral artery images, these were used to derive vessel-level inflammatory metrics (MDS and AS) rather than for direct lesion-to-lesion coregistration, because precise plaque boundaries could not be reliably established on the fusion images. Therefore, our inflammatory assessments reflect patient-level or vessel-level metrics rather than lesion-specific activity, which may obscure correlations between inflammation and focal stenotic severity. Furthermore, our ultrasound measurements focused on luminal stenosis and flow velocity rather than outer vessel wall morphology. Consequently, we did not quantitatively assess arterial positive remodelling and plaque characteristics, such as echogenicity or instability, which may also influence both PET signal intensity and local haemodynamics independently of the degree of stenosis.
Conclusion
This study demonstrates the utility of multimodality imaging in capturing the complex interplay of arterial inflammation, plaque burden, arterial blood flow and microvascular perfusion in patients with diabetes and PAD. Correlations among the imaging modalities partially support the hypothesis that lower limb arterial inflammation drives atheroma plaque progression, compromising both macrovascular and microvascular blood flow. PET, BOLD and IVIM imaging provided complementary insights, with BOLD metrics exhibiting strong correlations with arterial blood flow and the novel IVIM protocol enabling efficient perfusion assessment. Although no direct relationship between arterial inflammation and microvascular perfusion was observed, combining PET and MRI-based imaging with ultrasound offers a comprehensive approach to understanding the pathology of diabetic PAD. These findings highlight the potential for multimodal imaging to improve diagnostic precision, enhance disease monitoring and inform personalised therapeutic strategies.
Clinical Perspective
- Advanced MRI techniques (BOLD and IVIM) can non-invasively detect and quantify impaired microvascular perfusion in patients with diabetes and PAD, offering a deeper assessment of tissue health that is not fully captured by standard arterial blood flow measurements like ABI/TBI.
- The finding that arterial inflammation (measured by PET) correlates with reduced arterial blood flow (TBI) and that microvascular perfusion (measured by BOLD) correlates with ABI/TBI highlights that these are distinct but interconnected pathological processes that can be assessed simultaneously.
- This multimodal imaging approach supports a move towards personalised PAD management. By identifying a patient’s dominant pathological profile, clinicians could better stratify disease and select targeted therapies, such as anti-inflammatory agents, revascularisation or treatments to improve microvascular function.