Imaging evaluation of vascular complications in patients with diabetes mellitus. A systematic review

 

 

Nº de DOI: 10.34896/RSI.2026.56.39.002

 

 

AUTHORS

1. Marlon Leonardo Ordoñez Ramos. General Practitioner. Affiliated with Instituto Ecuatoriano de Seguridad Social. Graduate of Universidad de Cuenca. Based in Esmeraldas, Ecuador. https://orcid.org/0009-0002-7959-2987
2. Dayra Michell Tuz Tapia. General Practitioner with a Master’s Degree in Hospital Management and Hospital Administration. Affiliated with Clínicas Privadas del Ecuador. Graduate of Universidad Regional Autónoma de Los Andes. Based in Tulcán, Ecuador. https://orcid.org/0009-0009-9769-7281
3. Luis Santiago Vilca Moreno. General Practitioner. Affiliated with Patronato Municipal de Inclusión Social del Gobierno Autónomo Descentralizado Municipal de Santo Domingo. Graduate of Universidad de Guayaquil. Based in Santo Domingo, Ecuador. https://orcid.org/0009-0004-0528-2745
4. Ericka Alejandra Cevallos Ortega. General Practitioner. Affiliated with Clínicas Privadas del Ecuador. Graduate of Universidad Central del Ecuador. Based in Santo Domingo, Ecuador. https://orcid.org/0009-0005-1728-6759
5. Karla Gissel Ipiales Iles. General Practitioner. Affiliated with Clínicas Privadas del Ecuador. Graduate of Universidad UTE. Based in Ibarra, Ecuador. https://orcid.org/0009-0009-1779-5413

ABSTRACT

Objective: To synthesise the journal literature on imaging evaluation of the microvascular and macrovascular complications of diabetes, and to identify where the evidence is strong and where it is thin.

Methodology: Structured search of the peer-reviewed literature on 4 August 2026, restricted to journal articles with a verifiable digital object identifier (DOI), with extraction by vascular territory and imaging modality and a narrative synthesis without pooling. Thirty articles were included: eight systematic reviews or meta-analyses, four narrative or consensus reviews, three trials (two randomised) and fifteen observational or diagnostic-accuracy studies.

Results: In the retina, optical coherence tomography angiography (OCTA) shows capillary dropout, enlargement of the foveal avascular zone and microaneurysms, and in several studies detects changes before retinopathy is visible on ophthalmoscopy; an autonomous artificial-intelligence system has been validated for screening in primary care^1,2,3,4. In the coronary circulation, coronary artery calcium (CAC) scoring is the best supported stratification tool in asymptomatic patients, since about 40% have a score below 10 and an excellent short-term prognosis, whereas routine screening with myocardial perfusion imaging or coronary CT angiography did not significantly reduce cardiac events in randomised trials^5,6,7. Cardiovascular magnetic resonance (CMR) detects reduced myocardial perfusion reserve and altered myocardial mechanics in diabetes without obstructive coronary disease⁸. In the diabetic foot, magnetic resonance imaging (MRI) is the most accurate test for osteomyelitis, with pooled sensitivity of 96.4% and specificity of 83.8%⁹. Evidence for carotid, peripheral arterial and renal imaging specific to diabetes is thinner.

Conclusions: Imaging should be selected by vascular territory and clinical question, not applied as a uniform screening battery. The main gaps are outcome data for newer modalities, standardised quantitative metrics, and diabetes-specific evidence for peripheral, renal and cerebral imaging.

KEY WORDS

Diabetes mellitus, diabetic angiopathies, diagnostic imaging, optical coherence tomography angiography, coronary artery calcium, cardiovascular magnetic resonance, computed tomography angiography, diabetic foot, diabetic retinopathy, diabetic kidney disease.

RESUMEN

Objetivo: Sintetizar la literatura científica sobre la evaluación mediante técnicas de imagen de las complicaciones microvasculares y macrovasculares de la diabetes, e identificar dónde la evidencia es sólida y dónde es escasa.

Metodología: Búsqueda estructurada de literatura revisada por pares realizada el 4 de agosto de 2026, limitada a artículos de revistas con un identificador de objeto digital (DOI) verificable; se realizó una extracción de datos según territorio vascular y modalidad de imagen, seguida de una síntesis narrativa sin combinación estadística de resultados (metaanálisis). Se incluyeron treinta artículos: ocho revisiones sistemáticas o metaanálisis, cuatro revisiones narrativas o de consenso, tres ensayos (dos aleatorizados) y quince estudios observacionales o de precisión diagnóstica.

Resultados: En la retina, la angiografía por tomografía de coherencia óptica (OCTA) muestra pérdida de capilares, agrandamiento de la zona avascular foveal y microaneurismas; varios estudios detectan cambios antes de que la retinopatía sea visible mediante oftalmoscopia, y se ha validado un sistema autónomo de inteligencia artificial para el cribado en atención primaria. En la circulación coronaria, la puntuación de calcio en las arterias coronarias (CAC) es la herramienta de estratificación con mayor respaldo en pacientes asintomáticos, dado que alrededor del 40 % presenta una puntuación inferior a 10 y un pronóstico excelente a corto plazo, mientras que el cribado sistemático mediante estudios de perfusión miocárdica o angiotomografía coronaria no redujo significativamente los eventos cardíacos en ensayos aleatorizados. La resonancia magnética cardiovascular (RMC) detecta una reducción de la reserva de perfusión miocárdica y alteraciones en la mecánica miocárdica en pacientes con diabetes sin enfermedad coronaria obstructiva. En el pie diabético, la resonancia magnética (RM) es la prueba más precisa para la osteomielitis, con una sensibilidad combinada del 96,4 % y una especificidad del 83,8 %. La evidencia sobre técnicas de imagen específicas para la diabetes en los territorios carotídeo, arterial periférico y renal es más limitada.

Conclusiones: La elección de la técnica de imagen debe basarse en el territorio vascular y la pregunta clínica, en lugar de aplicarse como una batería de cribado uniforme. Las principales lagunas de conocimiento incluyen la falta de datos sobre resultados clínicos con modalidades más recientes, la ausencia de métricas cuantitativas estandarizadas y la escasez de evidencia específica de la diabetes para las técnicas de imagen periféricas, renales y cerebrales.

PALABRAS CLAVE

Diabetes mellitus, angiopatías diabéticas, diagnóstico por imagen, angiografía por tomografía de coherencia óptica, calcio en arterias coronarias, resonancia magnética cardiovascular, angiotomografía computarizada, pie diabético, retinopatía diabética, nefropatía diabética.

INTRODUCTION

Vascular disease is the main cause of disability and death in people with diabetes mellitus, and many of its lesions are silent until damage is irreversible. Imaging can show these lesions directly, but the evidence for each modality is spread across ophthalmology, cardiology, nephrology, vascular medicine and foot-care journals.

OBJECTIVE

The objective of this review was therefore to synthesise the peer-reviewed journal literature on the imaging evaluation of vascular complications in people with diabetes mellitus and to answer three questions: (1) which imaging modalities have been evaluated in each vascular territory, and what do they detect; (2) what is the diagnostic or prognostic performance of these modalities, and where do they change clinical decisions; and (3) where are the main gaps in the evidence.

METHODOLOGY

Design and reporting:

This is a systematic-style review with a narrative synthesis. Its reporting follows the logic of the PRISMA statement (explicit question, eligibility criteria, search, selection, extraction and synthesis), but it was not conducted to the full standard of a protocol-registered, multi-database systematic review. The departures from that standard are stated below because they determine how far the conclusions can be trusted.

Search strategy:

The literature was searched on 4 August 2026 with a web search engine that indexes PubMed Central, publisher platforms, the Directory of Open Access Journals and institutional repositories. Queries combined a diabetes term (diabetes, diabetic, type 1, type 2) with a vascular territory and an imaging term. The territory terms were retinopathy, coronary artery disease, diabetic cardiomyopathy, carotid atherosclerosis, peripheral artery disease, diabetic kidney disease and diabetic foot osteomyelitis. The imaging terms were optical coherence tomography angiography, artificial intelligence, coronary artery calcium, coronary CT angiography, myocardial perfusion imaging, cardiac magnetic resonance, carotid intima-media thickness, duplex ultrasound, CT angiography, magnetic resonance angiography, functional renal MRI and bone MRI. Queries were refined when a territory returned few results, and the reference lists of retrieved reviews were examined for primary studies. The search was not run directly in MEDLINE, Embase, Scopus or the Cochrane Library.

Eligibility criteria:

Articles were eligible when they met all of the following: (1) published in a peer-reviewed journal; (2) population of people with type 1 or type 2 diabetes, or a mixed population in which people with diabetes were the explicit target of the imaging question or were analysed as a subgroup; (3) an imaging modality evaluated for detecting, characterising, staging or predicting the outcome of a microvascular or macrovascular complication, including osteomyelitis of the diabetic foot as a neuro-vascular complication; (4) a systematic review, meta-analysis, trial, cohort or cross-sectional study, consensus statement or narrative review; (5) a DOI that could be verified on a publisher, indexing or reference-list page; and (6) full text or a detailed abstract in English. Book chapters, conference abstracts, preprints, theses, trial-registry entries, news items and articles without a DOI were excluded. Imaging of diabetic neuropathy and of cerebral small-vessel disease was searched for, but no eligible journal article could be verified in the time available, so these territories are not covered.

Selection and data extraction:

One reviewer screened titles and abstracts and assessed full texts or detailed abstracts against the criteria. No second reviewer was involved and agreement was not measured. The number of records screened at each step was not logged, so a PRISMA flow diagram with record counts is not provided. Thirty articles were included. For each, the first author, year, journal, design, population, imaging modality, main numerical findings and stated limitations were extracted. For ten primary studies the numerical data were taken from their description in a retrieved consensus document rather than from the original article, and the consensus document is cited alongside them where used^5,7,16,24.

Appraisal of evidence:

No formal risk-of-bias instrument was applied. Each article was instead labelled by design, and the hierarchy of designs was used to weight the narrative. Of the 30 included articles, eight were systematic reviews or meta-analyses, four were narrative or consensus reviews, three were trials, two of them randomized and one prospective with a pilot randomised substudy, and fifteen were observational or diagnostic-accuracy studies. Where a review reported its own appraisal, this was relayed: for example, a meta-analysis of imaging for diabetic foot osteomyelitis judged eight of its 36 included studies to be at high risk of bias.

Synthesis:

Results were synthesised narratively by vascular territory and modality. Numerical estimates are reported as published and were not pooled across articles, because the included reviews overlap in their primary studies and because populations, devices and reference standards differ. Where two sources covered the same evidence, the more specific or more recent was preferred.

Limitations of the method:

The main limitations are the single-reviewer process, the absence of direct bibliographic database searches, the absence of a registered protocol, the absence of a formal risk-of-bias tool, the reliance on a secondary description for ten primary studies, and the likely over-representation of open-access and widely indexed articles. The review should be read as a structured, source-verified synthesis, not as a substitute for a protocol-registered systematic review with full database searching.

RESULTS

Overview of the included evidence:

Thirty journal articles published between 2004 and 2025 were included. By territory, ten concerned the retina, 15 the coronary circulation and myocardium, two the carotid or peripheral arteries [26,28], one the lower-limb arteries in general, one the kidney and one the diabetic foot. Table 1 summarises the principal modalities and the strength of the supporting evidence; the sections below give the detail.

Retinal microvasculature:

Systematic evidence for OCTA:

Four systematic reviews address OCTA in diabetes. The largest, covering PubMed and Embase, included 135 studies and described the findings that OCTA can display: microaneurysms, changes in the size and morphology of the foveal avascular zone (FAZ), changes in vascular perfusion, retinal microvascular abnormalities and new vessels, diabetic macular oedema, and the application of deep learning¹¹. A review focused on the FAZ searched PubMed, Embase and the Cochrane Library, identified 358 records, screened 215 after removing duplicates and included 12 studies (nine cross-sectional and three retrospective). It concluded that OCTA can non-invasively identify foveal capillary non-perfusion as an early event in diabetic retinopathy, in some studies even in patients without clinically identifiable microvascular lesions, and that longitudinal studies are needed to establish whether OCTA findings predict long-term structural and functional outcome. A meta-analysis was not possible with the data available¹. A review of proliferative disease screened 1300 records, assessed 283 in full text and included 60 studies; it found OCTA more sensitive than fluorescein angiography for detecting microvascular changes indicating disease progression¹³. A more recent review of recent advances addressed grading, lesion recognition and artificial intelligence applied to OCTA in diabetic retinopathy²⁵.

Quantitative findings in primary studies:

Primary OCTA studies converge on reduced vessel density in the retinal capillary plexuses. In a cross-sectional study using wide-field swept-source OCTA over a 12 x 12 mm field, 38 eyes of 20 people with type 2 diabetes and no retinopathy and 39 eyes of 21 people with early retinopathy were compared with 42 control eyes. Vessel density in the inner retina, superficial capillary plexus (SCP) and deep capillary plexus (DCP) was lower in both diabetic groups, and the decrease was most pronounced in the outermost ring (9 to 12 mm from the fovea), a region not covered by conventional small-field scans². A study of 99 eyes from 55 participants introduced branching and fragmentation metrics. Branch node number and density fell, and fragmented vessel metrics rose, from controls to people with diabetes without retinopathy to mild retinopathy, and the changes were clearer in the deep than the superficial plexus; the authors suggested that foveal branching metrics in the deep plexus may mark preclinical disease¹². In 54 eyes with type 2 diabetes and moderate non-proliferative retinopathy without macular oedema, compared with 73 control eyes, vessel density in the SCP was significantly lower in all five regions analysed and the FAZ area of the SCP was significantly larger (p < 0.001). Peripheral disruption of the FAZ (83.3%), microaneurysms in both plexuses (79.6%) and flow changes in the DCP (81.5%) were the commonest anatomical alterations, and a longer duration of diabetes correlated with a larger SCP FAZ²⁹. In type 1 diabetes without or with only mild retinopathy, parafoveal vessel density in the DCP was reduced compared with non-diabetic controls³. Other studies evaluated retinal and choroidal vessel density across stages of retinopathy and their relation to disease severity³⁰.

Limits of the retinal evidence:

The retinal OCTA literature is dominated by small cross-sectional studies, single devices and varying scan sizes. The authors of one study noted that findings on vessel density, FAZ area and fractal dimension vary considerably between studies, with some reporting reduction, some no change and some increase in early disease¹². The review of the FAZ also did not address heterogeneity among included studies in detail¹. Study populations are often small and ethnically homogeneous^2,12.

Artificial intelligence for screening:

The pivotal trial of an autonomous AI diagnostic system enrolled 900 people with diabetes and no history of retinopathy at ten primary-care sites. The system was compared with a reading-centre reference standard consisting of widefield stereoscopic fundus photography and macular OCT, graded on the Early Treatment Diabetic Retinopathy Study (ETDRS) severity scale. On this basis the regulator authorised the system to detect more than mild retinopathy and diabetic macular oedema, making it the first autonomous AI diagnostic system authorised in any field of medicine; the system was operated by existing primary-care staff⁴.

Coronary circulation:

Heterogeneity of risk and the role of CAC:

A consensus document of the Imaging Council of the American College of Cardiology systematically reviewed noninvasive risk assessment in asymptomatic diabetes. It highlighted that diabetes should no longer be treated as an automatic coronary risk equivalent, because a high proportion of patients with diabetes have a CAC score of zero or very low. In its summary, about 40% of adults with diabetes have a score below 10 and a very low mortality rate, whereas rapid progression of CAC identifies patients at higher risk⁵. A separate review likewise stated that more than 30% of asymptomatic patients with diabetes have no evidence of coronary atherosclerosis¹⁰. The mortality data come from a cohort of 9,474 people without diabetes and 903 with type 2 diabetes followed for five years after CAC screening: mortality was 44% higher in people with diabetes in each CAC category, but people with diabetes and CAC below 10 had the same mortality as controls with CAC below 10¹⁶. In 1,343 patients with type 2 diabetes from two population cohorts, CAC outperformed the Framingham risk score and the UKPDS risk engine for predicting incident coronary events (areas under the curve of 0.76, 0.70 and 0.69), and it improved reclassification⁵. In a prospective cohort of 510 asymptomatic patients with type 2 diabetes, there were 20 cardiovascular events over 2.2 years, none in patients with a CAC score below 10¹⁷.

The ACC consensus document attributes a Class IIa recommendation to CAC screening in asymptomatic diabetes in United States guidance, while carotid intima-media thickness and routine functional testing were given Class III⁵.

Myocardial perfusion imaging:

The DIAD study first estimated how common silent ischaemia is. Of 522 asymptomatic patients aged 50 to 75 years with type 2 diabetes who underwent adenosine stress perfusion imaging, 113 (22%) had inducible ischaemic changes and 33 (6%) had moderate to large ischaemia¹⁸. The randomised phase then assigned 1123 participants with type 2 diabetes and no symptoms to screening with adenosine-stress myocardial perfusion imaging (561 participants) or no screening (562). Over a mean follow-up of 4.8 years the cardiac event rate (non-fatal myocardial infarction or cardiac death) was 2.7% in screened and 3.0% in unscreened participants (hazard ratio 0.88; 95% CI 0.44 to 1.88; p = 0.73), and revascularisation rates were 5.5% and 7.8% (hazard ratio 0.71; 95% CI 0.45 to 1.1; p = 0.14); event rates were low and were not significantly reduced by screening⁶. In the BARDOT study of 400 asymptomatic high-risk patients, baseline perfusion imaging was abnormal in 22%, and patients with an abnormal baseline study had more cardiac deaths, myocardial infarctions and revascularisations (9.8% versus 2.9%) than those with a normal study¹⁹. The ACC document concluded that routine perfusion screening of all asymptomatic patients has a low yield, and that yield improves by selecting higher-risk patients, for example those with symptoms, peripheral vascular disease, chronic kidney disease, an abnormal electrocardiogram or a high CAC score above 400⁵. Combining CAC with selective perfusion imaging has been proposed on this basis¹⁷. Inducible ischaemia is more likely in patients with metabolic abnormalities at a given CAC level: in a cohort of 1043 asymptomatic patients, ischaemia was found in 13% versus 3.6% of patients with a CAC score of 100 to 399 with and without metabolic abnormalities, and in 23.4% versus 13.6% of those with a score above 400⁵.

Coronary flow reserve measured by positron emission tomography also separates risk. Diabetic patients without known coronary disease and with a flow reserve above 1.6 had low cardiac mortality similar to non-diabetic patients, whereas those with a reserve below 1.6 had a risk similar to non-diabetic patients with established coronary disease²⁰.

Coronary CT angiography:

The ACC document summarised that approximately 25% to 30% of asymptomatic patients with diabetes have no demonstrable plaque on coronary CT angiography, 24% to 32% have a stenosis of 50% or more (one outlying study reported 17%), and plaque with features of instability is more common than in non-diabetic subjects⁵. The FACTOR-64 trial randomised 900 high-risk asymptomatic patients with type 1 or type 2 diabetes to CT angiography screening or standard care. CT angiography showed no coronary disease in 31%, mild stenosis in 46%, moderate in 12% and severe in 11%. The primary outcome (all-cause death, non-fatal myocardial infarction or hospitalisation for unstable angina) occurred in 6.2% versus 7.6% (hazard ratio 0.80; 95% CI 0.49 to 1.32; p = 0.38), so the difference was not significant, although screening was followed by more intensive risk-factor modification⁷. The same document concluded that CT angiography is not recommended as a risk stratification tool in asymptomatic diabetic patients, while accepting that it might become appropriate in a subset defined by age, risk factors or biomarkers; another review reached the same view for asymptomatic patients^5,10.

Myocardium and coronary microcirculation:

Diabetes also affects the myocardium independently of epicardial coronary stenosis. CMR combines assessment of structure, function, perfusion and tissue characterisation in one examination, and a review of its role described its ability to detect occult myocardial lesions and reduced microvascular perfusion in early type 2 diabetes, while noting that myocardial spectroscopy is not yet part of routine assessment¹⁴.

In a CMR study of 65 patients whose coronary angiography showed no stenosis above 30%, 19 had diabetes. Left ventricular mass was higher in diabetes than in non-diabetes (112.8 versus 91.5 g, p = 0.01), as was left ventricular torsion (9.65 versus 8.59 degrees, p = 0.047), while myocardial perfusion reserve was lower (2.10 versus 2.84 mL/g/min, p = 0.01). Perfusion reserve correlated inversely with torsion, suggesting a mechanistic link between microvascular disease and cardiac dysfunction. Patients with prediabetes did not differ from controls with the techniques used⁸.

Late gadolinium enhancement has prognostic value. In the ACC document, unsuspected myocardial scar on late gadolinium enhancement in diabetic patients was linked with a four-fold increase in major adverse cardiovascular events and a seven-fold increase in mortality, and scar was found in 4.3% of asymptomatic people with type 1 diabetes in the DCCT/EDIC cohort^5,21,22. T1 mapping detects diffuse fibrosis: extracellular volume was increased in diabetic patients and related to diastolic function, and diffuse fibrosis by T1 mapping was associated with subclinical myocardial dysfunction in a pilot study^5,23. Proton spectroscopy showed higher myocardial triglyceride content in people with type 2 diabetes without coronary disease than in matched controls, directly correlated with diastolic dysfunction²⁴. Adenosine stress CMR perfusion had a sensitivity of 88% and a specificity of 82% for obstructive coronary disease in the general population and in diabetic patients, but the two studies on which this rests included few patients with diabetes (8.6% and 13%)⁵.

Carotid and peripheral arteries:

Carotid ultrasound:

An individual-participant meta-analysis of 3,902 adults with type 2 diabetes from 21 population-based cohorts found that the hazard ratio for cardiovascular events was 1.22 (95% CI 1.12 to 1.33) per standard deviation of mean common carotid intima-media thickness after adjustment for age, sex and cardiometabolic risk factors. The authors concluded that their results did not support the use of intima-media thickness progression as a surrogate end point in clinical trials in diabetes²⁶. The ACC document noted that the recommendation to use carotid intima-media thickness in risk assessment was dropped in the 2013 American guidance, and that in an analysis of the MESA cohort diabetic patients in the top quartile of thickness did not show a significant increase in events (hazard ratio 1.7; 95% CI 0.7 to 4.3)⁵.

Lower-limb arteries:

Direct evidence in diabetic populations is limited. A recent review of the diagnosis of peripheral arterial disease in people with diabetes discussed the non-invasive haemodynamic tests (ankle-brachial index, toe pressure, transcutaneous oxygen pressure) and the imaging methods that define the anatomical site and severity of stenosis or occlusion: duplex ultrasound, CT angiography, magnetic resonance angiography and digital subtraction angiography. It noted that diagnosis is often made harder by the characteristics of diabetic foot disease itself [28]. A systematic review and economic evaluation in unselected patients with symptomatic lower-limb disease (not specific to diabetes) concluded that contrast-enhanced MR angiography had better overall diagnostic accuracy than CT angiography or duplex ultrasound, that duplex ultrasound and MR angiography were more accurate for lesions above the knee than below it, and that the only controlled trial found duplex ultrasound comparable with contrast angiography for surgical planning, a finding at odds with the diagnostic-accuracy studies²⁷. The extent to which these results apply to diabetic below-knee and pedal disease, which is often calcified and distal, was not resolved by the included sources.

Kidney:

A systematic review and meta-analysis examined functional MRI in diabetic kidney disease. It identified four main techniques used to assess renal structure and function: magnetic resonance elastography, arterial spin labelling, blood-oxygen-level-dependent (BOLD) imaging and diffusion-weighted imaging including diffusion tensor imaging (DTI) and intravoxel incoherent motion. It reported that DTI had an advantage in assessing renal cortical changes and that intravoxel incoherent motion had some value for identifying early disease from the cortex or medulla¹⁵. The ACC document also stressed the interaction between kidney disease and the coronary circulation: in the CRIC cohort patients with chronic kidney disease and diabetes had more than three-fold odds of a high CAC score (above 100) compared with those without diabetes (odds ratio 3.25; 95% CI 2.44 to 4.34), which makes CAC a potential marker of risk in advanced kidney disease⁵. No other eligible diabetes-specific renal imaging article could be verified.

Diabetic foot osteomyelitis

A systematic review and meta-analysis of imaging for osteomyelitis in people with diabetic foot ulcers included 36 studies in the meta-analysis, eight of which were at high risk of bias. MRI (22 studies) had a pooled sensitivity of 96.4% (95% CI 90.7 to 98.7) and specificity of 83.8% (95% CI 76.0 to 89.5). Scintigraphy (17 studies) had sensitivity of 84.2% (76.8 to 89.6) and specificity of 67.7% (56.2 to 77.4), and plain radiography (16 studies) 61.9% (50.5 to 72.1) and 78.3% (62.9 to 88.5). Positron emission tomography also diagnosed osteomyelitis reliably, evidence for single-photon emission CT was limited, and MRI was judged the preferable test in most cases given its availability and the absence of ionising radiation⁹.

DISCUSSION

This review brings together imaging evidence across six vascular territories in diabetes. The picture is uneven. The evidence is strongest where a well-defined clinical question meets a mature test: MRI for suspected osteomyelitis in a diabetic foot ulcer, CAC scoring for stratifying risk in asymptomatic patients, and structured retinal imaging for the detection of retinopathy^4,5,9. It is more tentative where the technology is new and the primary studies are small, as with OCTA^1,2,11,12, CMR tissue characterisation^8,14 and functional renal MRI^8,14,15. It is thin or indirect for the peripheral arteries and for carotid imaging, where a consistent association with events has not led to a recommended role^5,26,27,28.

The most important lesson from the cardiac literature is that detecting disease is not the same as improving outcomes. Silent ischaemia can be found in about one in five carefully selected asymptomatic patients, and coronary CT angiography finds plaque in most, yet two randomised trials did not show a significant reduction in cardiac events with screening^5,6,7,18. The explanations proposed by the investigators and by the ACC document are consistent: event rates in contemporary, well-treated diabetic populations are low, medical therapy improved in both arms, and revascularisation of silent lesions has not been shown to add to medical therapy^5,6,7. Both trials had limited power for rare outcomes. The confidence intervals around the hazard ratios are wide (0.44 to 1.88 for DIAD, 0.49 to 1.32 for FACTOR-64), so they do not exclude a clinically important benefit or harm. They do show that routine screening of unselected patients cannot be recommended on the evidence available.

CAC scoring offers a different logic. Rather than looking for an obstructive lesion to treat, it re-sorts patients by risk. A score of zero or near zero identifies roughly two fifths of people with diabetes whose short-term mortality is similar to that of people without diabetes^5,16, which questions the older doctrine that all people with diabetes carry a coronary risk equivalent. A high score identifies people in whom intensification of preventive therapy is more likely to be worthwhile, and possibly those in whom functional testing has higher yield^5,17. This is a more modest ambition than screening for ischaemia, and it is better aligned with the evidence, although the ACC document itself acknowledges that this approach improves stratification but has not been shown to improve outcomes, and calls for trials⁵.

The retinal literature shows a recurring pattern: capillary loss, particularly in the deep plexus, with enlargement of the foveal avascular zone, appears in people with diabetes before the clinical features that define retinopathy^1,2,3,12. Wide-field imaging and new morphological metrics extend this finding to the periphery and to branching and fragmentation of vessels^2,12. These results are biologically coherent with the known sequence of pericyte loss and capillary non-perfusion, and they are the main reason for interest in OCTA as a tool for earlier detection^11,13. Yet there are good reasons for caution. Cross-sectional comparisons of group means do not show that an individual’s OCTA metrics predict progression; the FAZ review explicitly called for longitudinal studies¹. Reported effects differ in direction and magnitude across studies, and the authors of one OCTA study listed the inconsistent findings on vessel density, FAZ area and fractal dimension as a reason to examine a broader range of metrics¹². Instrument and scan-size differences make thresholds hard to transfer, since a vessel density value from one device cannot be applied to another. Segmentation and projection artefacts, which are particularly relevant in the deep plexus, may also contribute. For these reasons OCTA currently complements, and does not replace, the fundus-photography-based grading on which screening programmes depend.

The autonomous AI trial illustrates the other path to impact: not a better image but a different delivery model. By moving the diagnosis to the point of care in primary care offices and using existing staff, the system aimed at the large number of people with diabetes who do not attend eye examinations⁴. The performance of such a system is tied to its intended use. It was authorised to detect more than mild retinopathy and macular oedema, in people without a prior diagnosis of retinopathy, with images acquired on a specific camera, and an editorial accompanying the trial noted that it excludes many patients with established disease⁴. Extrapolating its performance beyond that setting is not justified.

The myocardial and renal literatures raise a related point. CMR shows reduced perfusion reserve and increased torsion in people with diabetes and no epicardial stenosis, supporting a microvascular contribution to diabetic heart disease, and tissue characterisation (scar, extracellular volume, steatosis) adds prognostic and mechanistic information^5,8,21,23,24. But the available primary studies are small, single-centre and mostly cross-sectional. The perfusion-reserve study included only 19 patients with diabetes, the stress CMR accuracy data came from studies in which only 8.6% and 13% of patients had diabetes, and prospective data linking CMR microvascular markers to hard outcomes in diabetes are scarce^5,8. In the kidney, functional MRI sequences measure oxygenation, perfusion, diffusion and stiffness and offer a non-invasive complement to biopsy, which is rarely feasible; however, the consolidated evidence we could verify is a single meta-analysis whose findings concern individual sequences and compartments, and no outcome-based thresholds were identified¹⁵.

For osteomyelitis in a diabetic foot ulcer, the one meta-analysis included here found MRI to be the most accurate imaging test, plain radiography and scintigraphy less accurate, and PET a reliable alternative⁹. Even here, caveats apply. The high sensitivity of MRI contrasts with a specificity that is moderate (83.8% pooled), and roughly a fifth of the included studies (eight of 36) were at high risk of bias⁹. Imaging findings therefore need to be integrated with clinical and, ideally, microbiological or histological data. Because a single review carries this evidence, it should be read as consistent rather than replicated. The vascular side of the diabetic foot is less well served: the review of peripheral arterial disease diagnosis in diabetes described the tools but could not offer comparative accuracy data specific to the below-knee and pedal arteries typically affected in diabetes, and the one systematic review of duplex ultrasound, CT angiography and MR angiography was not restricted to diabetes^27,28.

The strengths of this review are its coverage of several vascular territories in one place, the inclusion of only journal articles with a verified DOI, the explicit separation of the designs of the included studies, and the transparent description of the places where data were taken from secondary sources. The limitations are those of the method. The search was run through a web search engine rather than directly in bibliographic databases; screening and extraction were done by a single reviewer; no risk-of-bias tool was applied; and the numbers of records screened were not recorded. Ten primary studies were described from a consensus document, not from their original reports, which carries a risk of propagating any error or selective emphasis of the secondary source. Territories such as neuropathy and cerebral small-vessel disease could not be covered. The included systematic reviews share primary studies, so counting them separately would overstate the weight of evidence, which is why no pooling was attempted. Finally, publication bias towards positive diagnostic accuracy results cannot be excluded, particularly for the newer modalities.

For clinicians, a practical reading of this evidence is as follows. In the retina, structured screening with fundus-based imaging remains the foundation, with AI-based systems as one way of widening access in primary care, and OCT-based imaging as the tool for macular oedema and, increasingly, for non-perfusion^4,11,13. In the heart, CAC scoring can help to individualise preventive therapy in asymptomatic patients, while routine perfusion or CT angiographic screening is not supported, and functional or anatomical testing should be reserved for symptoms or high-risk features^5,7. In the foot, MRI is the first advanced test for suspected osteomyelitis after plain films and clinical assessment⁹. In the legs, kidneys and carotid arteries, imaging should follow general indications, with the recognition that diabetes-specific evidence is limited^15,26,28.

For research, five priorities emerge. First, prospective cohorts linking OCTA and CMR microvascular metrics to clinical outcomes, using harmonised devices and metrics, are needed to turn associations into validated predictors^1,8,12. Second, trials should test whether CAC-guided intensification of therapy improves outcomes in diabetes, which the ACC document identifies as a knowledge gap⁵. Third, diabetes-specific diagnostic-accuracy studies are needed for below-knee arterial imaging and for functional renal MRI^15,28. Fourth, reporting of studies should state the proportion of participants with diabetes and the reference standard clearly, since several accuracy estimates rest on mixed populations^5,27. Fifth, evaluation of the safety and equity of AI-based screening in populations and devices different from those of the pivotal trial is required⁴.

CONCLUSIONS

Imaging evaluation of vascular complications in diabetes is effective when the modality is matched to the territory and to a defined clinical question. The strongest journal evidence supports MRI for osteomyelitis of the diabetic foot, CAC scoring for risk stratification in asymptomatic patients, and fundus-based and OCT-based imaging for retinopathy, including autonomous AI screening in primary care. Routine screening of unselected asymptomatic patients with myocardial perfusion imaging or coronary CT angiography has not been shown to reduce cardiac events in randomised trials. OCTA, CMR tissue characterisation and functional renal MRI reveal early microvascular abnormalities in people with diabetes, but their clinical value depends on longitudinal studies that link these findings to outcomes and on standardisation of devices and metrics. Evidence specific to diabetes is limited for peripheral arterial, carotid and renal imaging. Because this review used a structured but single-reviewer, non-registered search, its conclusions should be confirmed by a protocol-registered systematic review with direct searches of bibliographic databases.

REFERENCES

1. Johannesen SK, Viken JN, Vergmann AS, Grauslund J. Optical coherence tomography angiography and microvascular changes in diabetic retinopathy: a systematic review. Acta Ophthalmol. 2019;97(1):7-14. doi: 10.1111/aos.13859.
2. Xu F, Li Z, Gao Y, Yang X, Huang Z, Li Z, et al. Retinal microvascular signs in pre- and early-stage diabetic retinopathy detected using wide-field swept-source optical coherence tomographic angiography. J Clin Med. 2022;11(15):4332. doi: 10.3390/jcm11154332.
3. Simonett JM, Scarinci F, Picconi F, Giorno P, De Geronimo D, Di Renzo A, et al. Early microvascular retinal changes in optical coherence tomography angiography in patients with type 1 diabetes mellitus. Acta Ophthalmol. 2017;95(8):e751-5. doi: 10.1111/aos.13404.
4. Abràmoff MD, Lavin PT, Birch M, Shah N, Folk JC. Pivotal trial of an autonomous AI-based diagnostic system for detection of diabetic retinopathy in primary care offices. NPJ Digit Med. 2018;1:39. doi: 10.1038/s41746-018-0040-6.
5. Budoff MJ, Raggi P, Beller GA, Berman DS, Druz RS, Malik S, et al. Noninvasive cardiovascular risk assessment of the asymptomatic diabetic patient: the Imaging Council of the American College of Cardiology. JACC Cardiovasc Imaging. 2016;9(2):176-92. doi: 10.1016/j.jcmg.2015.11.011.
6. Young LH, Wackers FJ, Chyun DA, Davey JA, Barrett EJ, Taillefer R, et al. Cardiac outcomes after screening for asymptomatic coronary artery disease in patients with type 2 diabetes: the DIAD study: a randomized controlled trial. JAMA. 2009;301(15):1547-55. doi: 10.1001/jama.2009.476.
7. Muhlestein JB, Lappé DL, Lima JAC, Rosen BD, May HT, Knight S, et al. Effect of screening for coronary artery disease using CT angiography on mortality and cardiac events in high-risk patients with diabetes: the FACTOR-64 randomized clinical trial. JAMA. 2014;312(21):2234-43. doi: 10.1001/jama.2014.15825.
8. Larghat AM, Swoboda PP, Biglands JD, Kearney MT, Greenwood JP, Plein S. The microvascular effects of insulin resistance and diabetes on cardiac structure, function, and perfusion: a cardiovascular magnetic resonance study. Eur Heart J Cardiovasc Imaging. 2014;15(12):1368-76. doi: 10.1093/ehjci/jeu142.
9. Llewellyn A, Kraft J, Holton C, Harden M, Simmonds M. Imaging for detection of osteomyelitis in people with diabetic foot ulcers: a systematic review and meta-analysis. Eur J Radiol. 2020;131:109215. doi: 10.1016/j.ejrad.2020.109215.
10. Malahfji M, Mahmarian JJ. Imaging to stratify coronary artery disease risk in asymptomatic patients with diabetes. Methodist Debakey Cardiovasc J. 2018;14(4):266-72. doi: 10.14797/mdcj-14-4-266.
11. Boned-Murillo A, Albertos-Arranz H, Diaz-Barreda MD, Orduna-Hospital E, Sánchez-Cano A, Ferreras A, et al. Optical coherence tomography angiography in diabetic patients: a systematic review. Biomedicines. 2022;10(1):88. doi: 10.3390/biomedicines10010088.
12. Yu Y, Cui S, He Y, Zhang J, Lu N, Yang Y, et al. Early diabetic retinopathy evaluation with OCTA: a study on vascular branching and fragmentation. Invest Ophthalmol Vis Sci. 2024;65(14):21. doi: 10.1167/iovs.65.14.21.
13. Vaz-Pereira S, Morais-Sarmento T, Esteves Marques R. Optical coherence tomography features of neovascularization in proliferative diabetic retinopathy: a systematic review. Int J Retina Vitreous. 2020;6:26. doi: 10.1186/s40942-020-00230-3.
14. Mavrogeni SI, Bacopoulou F, Markousis-Mavrogenis G, Giannakopoulou A, Kariki O, Vartela V, et al. Cardiovascular magnetic resonance as pathophysiologic tool in diabetes mellitus. Front Endocrinol (Lausanne). 2021;12:672302. doi: 10.3389/fendo.2021.672302.
15. Zhang Z, Chen Y, Zhou X, Liu S, Yu J. The value of functional magnetic resonance imaging in the evaluation of diabetic kidney disease: a systematic review and meta-analysis. Front Endocrinol (Lausanne). 2023;14:1226830. doi: 10.3389/fendo.2023.1226830.
16. Raggi P, Shaw LJ, Berman DS, Callister TQ. Prognostic value of coronary artery calcium screening in subjects with and without diabetes. J Am Coll Cardiol. 2004;43(9):1663-9. doi: 10.1016/j.jacc.2003.09.068.
17. Anand DV, Lim E, Hopkins D, Corder R, Shaw LJ, Sharp P, et al. Risk stratification in uncomplicated type 2 diabetes: prospective evaluation of the combined use of coronary artery calcium imaging and selective myocardial perfusion scintigraphy. Eur Heart J. 2006;27(6):713-21. doi: 10.1093/eurheartj/ehi808.
18. Wackers FJ, Young LH, Inzucchi SE, Chyun DA, Davey JA, Barrett EJ, et al. Detection of silent myocardial ischemia in asymptomatic diabetic subjects: the DIAD study. Diabetes Care. 2004;27(8):1954-61. doi: 10.2337/diacare.27.8.1954.
19. Zellweger MJ, Maraun M, Osterhues HH, Keller U, Müller-Brand J, Jeger R, et al. Progression to overt or silent CAD in asymptomatic patients with diabetes mellitus at high coronary risk: main findings of the prospective multicenter BARDOT trial with a pilot randomized treatment substudy. JACC Cardiovasc Imaging. 2014;7(10):1001-10. doi: 10.1016/j.jcmg.2014.07.010.
20. Murthy VL, Naya M, Foster CR, Gaber M, Hainer J, Klein J, et al. Association between coronary vascular dysfunction and cardiac mortality in patients with and without diabetes mellitus. Circulation. 2012;126(15):1858-68. doi: 10.1161/CIRCULATIONAHA.112.120402.
21. Kwong RY, Sattar H, Wu H, Vorobiof G, Gandla V, Steel K, et al. Incidence and prognostic implication of unrecognized myocardial scar characterized by cardiac magnetic resonance in diabetic patients without clinical evidence of myocardial infarction. Circulation. 2008;118(10):1011-20. doi: 10.1161/CIRCULATIONAHA.107.727826.
22. Turkbey EB, Backlund JYC, Genuth S, Jain A, Miao C, Cleary PA, et al. Myocardial structure, function, and scar in patients with type 1 diabetes mellitus. Circulation. 2011;124(16):1737-46. doi: 10.1161/CIRCULATIONAHA.111.022327.
23. Ng ACT, Auger D, Delgado V, van Elderen SGC, Bertini M, Siebelink HM, et al. Association between diffuse myocardial fibrosis by cardiac magnetic resonance contrast-enhanced T1 mapping and subclinical myocardial dysfunction in diabetic patients: a pilot study. Circ Cardiovasc Imaging. 2012;5(1):51-9. doi: 10.1161/CIRCIMAGING.111.965608.
24. Rijzewijk LJ, van der Meer RW, Smit JWA, Diamant M, Bax JJ, Hammer S, et al. Myocardial steatosis is an independent predictor of diastolic dysfunction in type 2 diabetes mellitus. J Am Coll Cardiol. 2008;52(22):1793-9. doi: 10.1016/j.jacc.2008.07.062.
25. Zhang Q, Gong D, Huang M, Zhu Z, Yang W, Ma G. Recent advances and applications of optical coherence tomography angiography in diabetic retinopathy. Front Endocrinol (Lausanne). 2025;16:1438739. doi: 10.3389/fendo.2025.1438739.
26. Lorenz MW, Price JF, Robertson C, Bots ML, Polak JF, Poppert H, et al. Carotid intima-media thickness progression and risk of vascular events in people with diabetes: results from the PROG-IMT collaboration. Diabetes Care. 2015;38(10):1921-9. doi: 10.2337/dc14-2732.
27. Collins R, Cranny G, Burch J, Aguiar-Ibáñez R, Craig D, Wright K, et al. A systematic review of duplex ultrasound, magnetic resonance angiography and computed tomography angiography for the diagnosis and assessment of symptomatic, lower limb peripheral arterial disease. Health Technol Assess. 2007;11(20):iii-iv, xi-xiii, 1-184. doi: 10.3310/hta11200.
28. Ghirardini F, Martini R. Current opinion on diagnosis of peripheral artery disease in diabetic patients. Medicina (Kaunas). 2024;60(7):1179. doi: 10.3390/medicina60071179.
29. Fernández-Espinosa G, Boned-Murillo A, Orduna-Hospital E, Díaz-Barreda MD, Sánchez-Cano A, Bielsa-Alonso S, et al. Retinal vascularization abnormalities studied by optical coherence tomography angiography (OCTA) in type 2 diabetic patients with moderate diabetic retinopathy. Diagnostics (Basel). 2022;12(2):379. doi: 10.3390/diagnostics12020379.
30. Wang H, Liu X, Hu X, Xin H, Bao H, Yang S. Retinal and choroidal microvascular characterization and density changes in different stages of diabetic retinopathy eyes. Front Med (Lausanne). 2023;10:1186098. doi: 10.3389/fmed.2023.1186098.

APPENDICE

TABLE 1. Imaging modalities by vascular territory and the main findings in diabetes.

Territory	Modality	Main finding	Evidence level
Retina	OCTA	Capillary dropout, FAZ enlargement, microaneurysms; changes may precede visible retinopathy [1,2,11]	Systematic reviews; mostly cross-sectional primary studies
Retina	Fundus photography with autonomous AI	Validated for detection of more than mild retinopathy and macular oedema in primary care [4]	Prospective pivotal study
Coronary	CAC score	Strong prognostic stratification; about 40% have a score below 10 [5,16]	Consensus document; cohort studies
Coronary	Myocardial perfusion imaging	Silent ischaemia in about 22% of selected asymptomatic patients; routine screening did not reduce events [6,18]	Randomised trial
Coronary	Coronary CT angiography	High prevalence of plaque; routine screening did not significantly reduce events [7]	Randomised trial
Myocardium	CMR	Increased LV mass and torsion, reduced perfusion reserve without obstructive CAD [8]	Small cohort studies
Carotid	Ultrasound (CIMT)	Higher IMT associated with events; progression not a valid surrogate [26]	Individual-participant meta-analysis
Lower limb	Duplex, CTA, MRA	Contrast-enhanced MRA most accurate overall in general PAD populations [27]	Systematic review (not diabetes-specific)
Kidney	Functional MRI	ASL, BOLD, DWI and related sequences under study [15]	Systematic review and meta-analysis
Foot	MRI	Sensitivity 96.4%, specificity 83.8% for osteomyelitis [9]	Systematic review and meta-analysis

Source: Prepared by the authors.