Article Text

Download PDFPDF

Extended report
Neuropsychiatric lupus or not? Cerebral hypoperfusion by perfusion-weighted MRI in normal-appearing white matter in primary neuropsychiatric lupus erythematosus
  1. Efrosini Papadaki1,2,
  2. Antonis Fanouriakis3,4,
  3. Eleftherios Kavroulakis1,
  4. Dimitra Karageorgou1,
  5. Prodromos Sidiropoulos3,
  6. George Bertsias3,5,
  7. Panagiotis Simos2,6,
  8. Dimitrios T Boumpas4,5,7,8,9
  1. 1 Department of Radiology, School of Medicine, University of Crete, University Hospital of Heraklion, Heraklion, Greece
  2. 2 Institute of Computer Science, Foundation of Research and Technology, Heraklion, Greece
  3. 3 Department of Rheumatology, Clinical Immunology and Allergy, School of Medicine, University of Crete, University Hospital of Heraklion, Heraklion, Greece
  4. 4 Department of Internal Medicine, Attikon University Hospital, Medical School, National and Kapodestrian University of Athens, Athens, Greece
  5. 5 Institute of Molecular Biology and Biotechnology, Foundation of Research and Technology-Hellas, Heraklion, Greece
  6. 6 Department of Psychiatry, School of Medicine, University of Crete, University Hospital of Heraklion, Heraklion, Greece
  7. 7 Laboratory of Autoimmunity and Inflammation, Biomedical Research Foundation of the Academy of Athens, Athens, Greece
  8. 8 Joint Academic Rheumatology Program, Medical School, National and Kapodestrian University of Athens, Athens, Greece
  9. 9 Medical School, University of Cyprus, Nicosia, Cyprus
  1. Correspondence to Ass. Professor Efrosini Papadaki, Department of Radiology, School of Medicine, University of Crete, University Hospital of Heraklion, Heraklion 71306, Greece; fpapada{at}otenet.gr

Abstract

Objectives Cerebral perfusion abnormalities have been reported in systemic lupus erythematosus (SLE) but their value in distinguishing lupus from non-lupus-related neuropsychiatric events remains elusive. We examined whether dynamic susceptibility contrast-enhanced perfusion MRI (DSC-MRI), a minimally invasive and widely available method of cerebral perfusion assessment, may assist neuropsychiatric SLE (NPSLE) diagnosis.

Methods In total, 76patients with SLE (37 primary NPSLE, 16 secondary NPSLE, 23 non-NPSLE) and 31 healthy controls underwent conventional MRI (cMRI) and DSC-MRI. Attribution of NPSLE to lupus or not was based on multidisciplinary assessment including cMRI results and response to treatment. Cerebral blood volume and flow were estimated in 18 normal-appearing white and deep grey matter areas.

Results The most common manifestations were mood disorder, cognitive disorder and headache. Patients with primary NPSLE had lower cerebral blood flow and volume in several normal-appearing white matter areas compared with controls (P<0.0001) and lower cerebral blood flow in the semioval centre bilaterally, compared with non-NPSLE and patients with secondary NPSLE (P<0.001). A cut-off for cerebral blood flow of 0.77 in the left semioval centre discriminated primary NPSLE from non-NPSLE/secondary NPSLE with 80% sensitivity and 67%–69% specificity. Blood flow values in the left semioval centre showed substantially higher sensitivity than cMRI (81% vs 19%–24%) for diagnosing primary NPSLE with the combination of the two modalities yielding 94%–100% specificity in discriminating primary from secondary NPSLE.

Conclusion Primary NPSLE is characterised by significant hypoperfusion in cerebral white matter that appears normal on cMRI. The combination of DSC-MRI-measured blood flow in the brain semioval centre with conventional MRI may improve NPSLE diagnosis.

  • magnetic resonance imaging
  • systemic lupus erythematosus
  • disease activity

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Introduction

Systemic lupus erythematosus (SLE) affects the peripheral and central nervous system in up to 30% of cases (neuropsychiatric SLE (NPSLE)). NPSLE can present with focal neurological deficits, often related to antiphospholipid antibodies (aPL),1 or diffuse manifestations that vary from overt (eg, seizures, psychosis) to subtle presentations such as headache, mood disturbances, anxiety disorders and cognitive dysfunction.2–4 Only in 40% symptoms can be attributed to underlying SLE (primary NPSLE). In the remaining cases, they may be secondary to treatment, other systemic causes or concomitant psychiatric disorders (secondary NPSLE).2–6

Despite advances, NPSLE still poses diagnostic and therapeutic challenges. To facilitate patient care, we have developed recommendations for the management of patients with SLE presenting with neuropsychiatric manifestations.5 Within the European League Against Rheumatism (EULAR) recommendations, brain MRI is the imaging technique of choice.2 3 5 6 On conventional imaging, the most frequent finding is focal, small, white matter hyperintensities, commonly related to small-vessel disease.7–12 However, these abnormalities are non-specific and may be absent in >50% of patients with NSPLE .7 9 12–15 To this end, advanced MRI techniques promise greater sensitivity in detecting microstructural and metabolic changes in normal-appearing white matter and deep grey matter structures.16–21

Dynamic susceptibility contrast-enhanced T2*-weighted perfusion MRI (DSC-MRI) provides a minimally invasive assessment of brain haemodynamic status through quantitative estimates of regional cerebral blood volume (CBV, a marker of blood volume per cerebral tissue volume) and cerebral blood flow (CBF, a marker of instantaneous capillary flow). In contrast to radionuclide techniques, DSC-MRI does not expose patients to radiation (only to gadolinium administration), affords higher spatial resolution, inherent co-registration to anatomical images and simultaneous measurement of CBV and CBF.22 Although DSC-MRI has become available in clinical practice, very few studies have used it in SLE, with contradictory results.16 22–25

Direct associations between brain perfusion findings and neuropsychiatric symptoms, disease activity and severity indices, and attribution to SLE have been scarce. We performed DSC-MRI studies to assess (a) haemodynamic changes in the normal-appearing white and grey matter of patients with NPSLE, (b) associations between brain perfusion and disease activity and (c) the utility of perfusion indices in the diagnosis of primary NSPLE and the discrimination between primary and secondary NPSLE.

Methods

Participants

Patients from the Department of Rheumatology, Clinical Immunology and Allergy of the University Hospital of Heraklion, Greece, who met the revised American College of Rheumatology (ACR)26 or  Systemic Lupus International Collaborating Clinics (SLICC)27 classification criteria for SLE, were included. Patients with thromboembolic cardiovascular disease or other primary central nervous system (CNS) diseases (elicited through history or evident on MRI) were excluded. A healthy control (HC) group of 31 volunteers who had no history of neurological or other diseases and no evident MRI abnormalities were enrolled to generate perfusion MRI reference standards. All participants signed informed consent (see also online supplementary material).

Case definitions and attribution of neuropsychiatric manifestations to SLE

Neuropsychiatric manifestations were defined according to the 1999 ACR nomenclature and case definitions.28 Diagnostic work-up was according to the EULAR recommendations5; diagnosis of cognitive dysfunction was made either with a formal neuropsychological assessment (performed by an expert neuropsychologist (PS))29 and/or the Montreal Cognitive Assessment.30 Attribution of NP events was based on physician judgement, following multidisciplinary approach6 and considering patient age, risk factors for primary NPSLE (aPL), prior major neuropsychiatric manifestation, generalised disease activity, findings of conventional MRI imaging and other diagnostic procedures. Additionally, the ACR ‘exclusion’ and ‘association’ factors (ie, their absence favours attribution to SLE), as well as ‘SLE-favouring factors’ of the Italian Study Group on NPSLE attribution model,31 were also considered. Attribution was ascertained after prospective follow-up of patients and has been additionally retrospectively cross-checked with existing attribution models31 32 (online supplementary table 1). Physicians were blind to the DSC-MRI results.

MRI

All MRI scans were performed on a 1.5 T MR scanner (Vision/Sonata, Siemens, Erlangen). Lesion burden of small (<1 cm) white matter hyperintensities on T2/FLAIR sequences was classified according to their absolute number: grade 0=none, grade 1 (mild)=1–5, grade 2 (moderate)=6–10 or grade 4 (severe) >10.24 The T2* DSC-MRI was performed with 2D single-shot multislice Gradient Echo Planar Imaging sequence (time repetition (TR)/time echo (TE)/flip angle (FA): 1500 ms/40 ms/30o, band width (BW): 2442 Hz/pixel, echo spacing: 0.47 ms, echo planar imaging factor 64). Relative CBV, CBF and mean transit time values of normal-appearing white matter and grey matter areas were calculated in 18 regions of interest (ROI) of the brain, including representative areas in the periventricular region, semioval centre and subcortical white matter (frontal, parietal, temporal, occipital lobes) and grey matter in the thalami, putamen and caudate nuclei. The aforementioned values were normalised for each patient, divided by the respective average values of the middle cerebellar peduncles.

Statistical analysis

One-way analysis of variance (ANOVA) was used to assess differences in perfusion measurements between HC and patient subgroups, separately for each of the 18 ROI (Bonferroni-adjusted α=0.003). In separate analyses, disease duration, hypertension and smoking were used as covariates. Associations with clinical variables were assessed through Pearson correlation coefficients (zero-order and partial, controlling for age, disease duration and hypertension). Receiver operator characteristic (ROC) analysis was used to determine the sensitivity and specificity of each perfusion measure alone to differentiate HC and patient subgroups. We estimated the sensitivity of each measure and corresponding cut-off value in distinguishing between the clinical subgroups in order to maintain a minimum specificity of 80%. Likewise, we computed the specificity of each measure to obtain a minimum sensitivity of 80%. All analyses were performed with SPSS V.20.

Results

Clinical and demographic data

From the 76 patients with SLE included in the study, 37 were diagnosed as primary NSPLE, 16 as secondary NPSLE and 23 as non-NPSLE (online supplementary table S1). Most common manifestations in the primary NPSLE group were mood disorder (32.40%), cognitive disorder (13.5%) and headache (10.8%); in secondary NPSLE, mood disorders and anxiety disorders (25.0% each) (table 1). The three patient groups were comparable in terms of age, gender, treatment regimens, smoking and disease duration. Patients with primary NPSLE had higher total Systemic Lupus Erythematosus Disease Activity Index (SLEDAI) (P<0.0001) and SLEDAI scores excluding any neurological component (non-neurological SLEDAI, P=0.02). The HC group was comparable in terms of age with each of the three patient subgroups, except for a higher percentage of men (29% vs 10.8% in primary NPSLE, 0% in non-NPSLE and 12.5% in secondary NPSLE, respectively) Smoking did not differ between HC (19.4%) and the patient subgroups.

Table 1

Demographic, clinical and conventional MRI characteristics of SLE patients with and without neuropsychiatric involvement

Higher lesion burden by conventional MRI in primary NPSLE

MRI was normal in 15/37 (40.5%) primary NPSLE, 12/16 (75.0%) secondary SLE, 15/23 (65.2%) non-NPSLE patients and 28/31 (90.3%) HC. Within patients with abnormal MRI, the three subgroups did not differ significantly on the grade of white matter hyperintensities. Nevertheless, grade 3 lesions were slightly more prevalent in primary NPSLE (21.6%) compared with secondary NPSLE (12.5%) and non-NPSLE (13%) patients (table 1).

Primary NPSLE displays widespread hypoperfusion in normal-appearing white matter

Table 2 shows DSC-MRI results in several regions where significant between-group differences were found on perfusion measures. Analysis of covariance results remained essentially unchanged after inclusion of hypertension, disease duration and smoking as covariates (online supplementary result). Post hoc comparisons revealed significantly (P<0.003) reduced normalised CBF in patients with primary NPSLE compared with HC in the temporal, frontal, occipital, periventricular and semioval centre (the mass of white matter present in each hemisphere below the cerebral cortex) normal-appearing white matter bilaterally (figure 1). Reduced normalised CBV in primary NPSLE versus HC was found in frontal, periventricular and semioval centre normal-appearing white matter, bilaterally, as well as in all normal-appearing grey matter regions (online supplementary figure S1). Differences between HC and non-NPSLE or secondary NPSLE patients were not statistically significant in any ROI. Moreover, primary NPSLE displayed reduced CBF compared with non-NPSLE patients in left occipital and bilateral temporal and semioval centre normal-appearing white matter. The former group also displayed reduced CBV in the left semioval centre and thalamus. Together, these data suggest that primary NPSLE is characterised by widespread hypoperfusion across various cerebral regions.

Table 2

Main effects of group on normalised CBF and CBV in normal-appearing white matter and normal-appearing deep grey matter

Figure 1

Average normalised cerebral blood flow (CBF) values in normal-appearing white matter regions of interest (ROIs) for healthy controls (HC), non-neuropsychiatric systemic lupus erythematosus (NPSLE), secondary NPSLE (S-NPSLE) and primary NPSLE patients (P-NPSLE). The locations of ROIs where perfusion measurements were made on scans from a representative HC participant are shown by white circles. Significant pairwise group differences (at Bonferroni-adjusted P<0.003) are indicated by brackets. Bars indicate SE. L, left hemisphere; PVL, periventricular white matter; R, right hemisphere; SOC, semioval centre.

Normal-appearing white matter hypoperfusion is greater in primary compared with secondary NPSLE

Attribution of neuropsychiatric, particularly non-focal, events to SLE is challenging in clinical practice. We examined the utility of perfusion MRI in discriminating primary from secondary NPSLE by performing pairwise comparisons in the ROIs where significant between-group effects were detected (table 2) and after controlling for MRI lesion load (grade 1–3). Patients with primary NPSLE demonstrated significantly reduced normalised CBF in the left and right semioval centres (P=0.001; figure 1) and CBV in the left thalamus (P=0.0001; online supplementary figure S1), compared with patients with secondary NPSLE. Similar, yet non-significant, trends were observed in several other normal-appearing white matter regions and grey matter structures. These findings suggest that patients with primary NPSLE may exhibit hypoperfusion at least in selected areas of the brain compared with patients with neuropsychiatric manifestations not attributable to SLE.

Hypoperfusion in the brain semioval centre bilaterally correlates with SLE activity

The clinical relevance of hypoperfusion in the left and right semioval centres was further examined through correlational analyses against disease characteristics in the entire SLE cohort. Weak negative correlations were found between disease duration and normalised CBF in the left (r=−0.282, P=0.04) and right semioval centres (r=−0.321, P=0.02). Reduced perfusion in these normal-appearing white matter ROIs correlated with increased disease activity (SLEDAI score: r=−0.409, P=0.003, and r=−0.399, P=0.003, respectively). These associations remained significant after controlling for age, disease duration and hypertension (left semioval centre: r=−0.386, P=0.007; right semioval centre: r=−0.355, P=0.009).

Clinical validity of perfusion measures in the diagnosis of primary NPSLE

To define the diagnostic utility of perfusion MRI, we performed ROC analyses to calculate sensitivity and specificity for discrimination between patient groups. Table 3 shows results of these analyses only if either specificity and/or sensitivity exceeded 50%. This criterion was met by three CBF indices, namely average normalised CBF across all normal-appearing white matter ROIs, and normalised CBF in the left and right semioval centres. Among single ROI indices, best results were obtained with CBF from the left semioval centre. Specifically, discrimination of primary NPSLE from HC was optimal using a normalised CBF cut-off of 0.77, to achieve 80% sensitivity and 78% specificity (figure 2, upper panel). Normalised CBF values from the left semioval centre (using the same cut-off of 0.77) showed 80% sensitivity and 67% specificity to discriminate primary NPSLE from non-NPSLE (figure 2, middle panel). Finally, normalised CBF values from the left semioval centre discriminated between primary and secondary NPSLE with 80% sensitivity and 69% specificity (figure 2, lower panel). Together, these findings indicate that perfusion measurements in the left semioval centre can differentiate patients with primary NPSLE from other groups of HC or patients with SLE (non-NPSLE or secondary NPSLE).

Figure 2

Receiver  operator  characteristic  curve illustrating the relationship between sensitivity and 1-specificity in discriminating healthy participants from patients with primary neuropsychiatric systemic lupus erythematosus (NPSLE) (top panel), patients with SLE without neuropsychiatric manifestations from patients with primary NPSLE (middle panel) and patients with secondary neuropsychiatric manifestations from patients with primary NPSLE (lower panel), using normalised cerebral blood flow in left semioval centre normal-appearing white matter.

Table 3

Receiver operating characteristic analyses results, estimating sensitivity and specificity of perfusion measures in differentiating healthy participants (n=31), systemic lupus erythematosus (SLE) control (n=39), patients with primary neuropsychiatric SLE (NPSLE) (n=37) and secondary NPSLE (n=16)

Combined perfusion and conventional MRI data allow for more accurate diagnosis of primary NPSLE with a sensitivity and specificity >80%

Since conventional MRI is the recommended neuroimaging modality for NPSLE diagnosis, we sought to define the comparative and combined diagnostic utility of perfusion MRI and conventional MRI for identifying primary from secondary and non-NPSLE patients (table 4). CBF measures alone (especially in the left semioval centre) conferred substantial improvement in the ability to detect primary NPSLE cases over conventional MRI (81% vs 19%–24% for grade 1–3 brain lesions) with a modest reduction in specificity (68% vs 89%–100%). CBF in the right semioval centre provided similar specificity with left semioval centre CBF, although with moderately reduced sensitivity. Sensitivity estimates based on average normal-appearing white matter CBF were clearly inferior to those based on CBF in either semioval centre.

Table 4

Sensitivity and specificity of neuroimaging data in differentiating between systemic lupus erythematosus (SLE) controls (non-neuropsychiatric SLE (NPSLE)/secondary NPLSE) and patients with primary NPSLE

Discussion

This is the first DSC-MRI study in SLE to evaluate perfusion in normal-appearing white matter (ie, excluding white matter hyperintensities from the analyses) including also patients with secondary NPSLE as a separate clinical group. Our findings confirmed the presence of diffuse normal-appearing white matter hypoperfusion in primary NPSLE. Hypoperfusion, more pronounced in the semioval centre, could differentiate between patients with primary NPSLE and those with either no NP symptoms or with NP symptoms not attributed to the disease. Adopting a normalised left semioval centre CBF cut-off value of 0.77 increased the sensitivity for differentiating between primary and secondary NPSLE by a factor of 4 (81% vs 19%–24%) compared with conventional MRI findings alone, with only a small reduction in specificity (68% vs 75%).

White matter hyperintensities are the most common MRI finding in diffuse NPSLE (40%–80%) but their specificity is low.10 In accordance with previous reports,7 9 10 12 MRI was normal in 55% of all SLE and 40% of primary NPSLE patients. Notably, histologically verified, small-vessel injury may be undetectable even with postmortem high-field MRI of 7T.8 In contrast, DSC-MRI detected brain hypoperfusion in 73% of patients with primary NPSLE, 27% of non-NPSLE patients and 33% of patients with secondary NPSLE who had normal conventional MRI (online supplementary table S2). Previous perfusion MRI studies have yielded conflicting results,16 22–25 possibly due to study variations in MR scanners and postprocessing techniques. Our finding of diffuse hypoperfusion in NPSLE concords with studies using PET and SPECT revealed moderate-to-severe perfusion defects in multiple brain regions.33–38

Hypoperfusion in primary NPSLE was more pronounced in the semioval centre, a region associated with the lowest normalised CBF and CBV values, compared with other normal-appearing white matter areas. This may be due to increased susceptibility of this territory to haemodynamic impairment. Specifically, the distal region of semioval centre is perfused by the internal carotid artery and is located in the border zone between the territories of supply of the superficial perforators of the middle cerebral artery and the anterior cerebral artery (internal watershed area),39 thus being prone to cerebral small-vessel disease, which is prominent in SLE.40 41 Hypoperfusion was noted despite the fact that numerically more patients with primary NPSLE were using antiplatelet agents at the time of evaluation. Also, these patients had longer disease duration compared with patients with secondary NPSLE (mean 5.3 vs 2.8 years, respectively); although one could argue that this could have impacted brain perfusion, observational studies suggest that most SLE-attributed neuropsychiatric manifestations occur early in the disease course and do not arise as a result of long-standing disease. To this end, differences in perfusion estimates in our study were not affected by disease duration (online supplementary result).

Autopsy findings have revealed widespread vascular lesions in NPSLE. Vasculopathy or hypercoagulopathy, resulting in diffuse chronic ischaemia and neuronal loss, are putative pathogenic processes.8 42 43 Our finding of widespread reduction of CBF and CBV in normal-appearing white matter of patients with NPSLE is consistent with these histopathological changes. True cerebral vasculitis is infrequent (6%–9%) and has been linked to inflammatory-like lesions on MRI, defined as medium or large, ill-defined, enhancing T2 hyperintensities.10 Such lesions were detected only in 5% of our patients with primary NPSLE. Inflammatory activity has been related to increased perfusion due to inflammation-mediated vasodilation.44 Our findings of diffuse hypoperfusion, along with the rarity of inflammatory-like lesions on conventional MRI, suggest that in most cases primary NPSLE is characterised by chronic widespread decreased perfusion and ischaemia due to microvasculopathy and subsequent neuronal loss in the absence of high-grade inflammation.

In agreement with previous SPECT studies, we observed a trend for reduced perfusion in non-NPSLE patients compared with HC.33 35 37 38 Histopathological studies have showed non-specific focal vasculopathy in SLE without NP manifestations, while diffuse vasculopathy and microthrombi occurred more often in NPSLE. It has been proposed that ischaemic injury may occur in all patients with SLE, but primary NP manifestations emerge only after exceeding a certain threshold of ischaemic injury.8 By analogy, the normalised CBF cut-off value of 0.77 could correspond to the magnitude of hypoperfusion caused by ischaemic injury, which is necessary to induce NP manifestations in patients with SLE.

Basal ganglia pathology in NPSLE is far less common (12%–30%) than in the white matter.37 45 In the present study, significant basal ganglia hypoperfusion in patients with NPSLE was evidenced only on CBV, although a potential bias in DSC-MR-based perfusion estimates due to increased iron deposition in the putamen, enhancing the susceptibility effect, cannot be ruled out.46

Attribution of neuropsychiatric manifestations to SLE largely relies on expert physician judgement; neither clinical parameters nor conventional MRI have sufficient individual diagnostic accuracy. Moreover, recently proposed attribution models show suboptimal performance in non-focal manifestations. We employed multidisciplinary assessment tailored to each case and NP manifestation for attribution,6 47 and similar approaches have been used by others.48 We have also recently cross-checked our judgement with existing attribution models.6 49 In this process, tools that could aid the diagnostic process are eagerly awaited. While the cause of preferential hypoperfusion in the left semioval centre in primary NPSLE is not yet definitive, it provides proof of concept that attribution of neuropsychiatric manifestations to SLE can be assisted by parameters more objective than sole physician assessment. Notably, however, these results apply mainly to diffuse NPSLE as most of our patients had such manifestations and may not be directly extrapolated to all NPSLE syndromes.

In addition to sample size, there are some limitations in our study. First, we had no data regarding internal carotid arteries atherosclerosis, which might affect cerebral perfusion. Although Doppler carotid artery examination was not performed in the context of our study, substantial left–right asymmetries in brain perfusion (suggestive of significant unilateral carotid stenosis) were not evident in our population (online supplementary material). Second, the HC group, which was recruited to obtain perfusion MRI reference standards, was not formally matched to patients with SLE and included more men. Nonetheless, auxiliary analyses showed (a) no significant gender bias on perfusion measurements and (b) between-groups differences in perfusion were unchanged when men were excluded (online supplementary results). Additional limitations include (a) the need for data normalisation using measurements from a brain structure assumed not to be affected by the disease (here, the middle cerebellar peduncle as the least frequently involved area in diffuse NPSLE50) and (b) the unreliability of cortical perfusion estimates, which were not obtained in the present study to avoid susceptibility effects from small vessels in the adjacent subarachnoid space.16 23–25

In conclusion, DSC-MRI was able to detect brain haemodynamic abnormalities, even in the absence of macroscopic anatomic lesions, in patients with primary NPSLE. The present findings highlight the potential utility of perfusion indices in clinical practice for the diagnosis of NPSLE, in addition to conventional neuroimaging techniques recommended by the EULAR, with limited or no additional expense for the patient. Larger longitudinal studies, including comparisons before and after therapy, are needed to validate these data.

Acknowledgments

AF’s contribution to this work has been supported in part by a grant from the Hellenic Society of Rheumatology.

References

Footnotes

  • AF and EK contributed equally.

  • Handling editor Tore K Kvien

  • Contributors EP designed the study, performed and interpreted the MRI exams, analysed the MR perfusion data, wrote, edited and critically reviewed the manuscript. AF performed clinical examination of patients, edited the manuscript. EK collected image data and performed image analysis. DK performed image analysis. PS performed clinical examination of patients. GB performed clinical examination of patients, edited and critically reviewed the manuscript. PS performed neuropsychological assessment and statistical analysis, edited and critically reviewed the manuscript. DTB designed the study, performed clinical examination of patients, edited and critically reviewed the manuscript. All authors contributed substantially to discussion of content.

  • Competing interests None declared.

  • Patient consent Obtained.

  • Ethics approval University Hospital of Heraklion Ethical Committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.