RPS 711 - Stroke and neurovascular imaging: inside and outside of the blood vessels

Author Block: A. Toth, Y. (. Cho, E. Wilson, J. Crow, E. Bass, J. Joyce, M. G. Matheus, S. Tipnis, M. V. Spampinato; Charleston, SC/US
Purpose: Our goal was to identify the optimal combination of dedicated neurovascular reconstruction kernels and quantum iterative reconstruction (QIR) levels for ultra-high-resolution (UHR) photon-counting detector (PCD)-CT angiography of the head and neck.
Methods or Background: 18 patients with intracranial aneurysms were prospectively included in this study. CT angiograms were obtained in UHR mode using a clinical dual-source PCD-CT scanner. Images were reconstructed with a slice thickness of 0.2 mm, utilizing six strength levels of a dedicated neurovascular kernel (Hv48/Hv56/Hv64/Hv72/Hv80/Hv89) and four levels of QIR (1-4). We assessed image noise, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and vessel sharpness for all reconstructions.
Results or Findings: With higher kernel sharpness and lower QIR, image noise continuously increased. The best performing reconstructions in terms of CNR were Hv48 and Hv72 in combination with QIR-4. Vessel sharpness improved with higher kernel levels, reaching a plateau with the Hv64 and Hv72 reconstructions, as observed in the small intracranial arteries (maximum ΔHU values of 260.59 and 255.11, respectively). Based on the results of the quantitative analysis, the kernels identified as the top performers and selected for further evaluation in the qualitative analysis were Hv56, Hv64, and Hv72.
Conclusion: Vessel sharpness increased with higher kernels levels, reaching a plateau at Hv64 and Hv72. Overall, Hv56, Hv64 and Hv72 were recognized as the best performing kernels based on the quantitative results. In the following steps, qualitative image quality evaluation will be conducted by three readers using a 5-point Likert scale. This evaluation will focus on the best-performing kernels in combination with the available QIR levels to assess overall image quality and diagnostic performance.
Limitations: The limitations of the study are the relatively small patient cohort and the fact that quantitative measurements were performed by a single observer.
Funding for this study: Funding was provided by Siemens Healthineers (research grant).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by the Institutional Review Board of the Medical University of South Carolina (Pro00123327)).

Author Block: Y. Wang, H. Zhu, J. Wen, S. Ma, S. Yang; Beijing/CN
Purpose: The study aimed to evaluate the effectiveness of non-contrast dual-layer spectral computed tomography (DLCT) for detecting early ischemic changes in patients with acute ischemic stroke (AIS).
Methods or Background: NCCT is a common imaging technique for suspected AIS patients. This study involved 27 AIS patients who underwent both DLCT and MRI within 12 hours of symptom onset. A retrospective analysis was conducted on the imaging data, focusing on quantitative measurements from regions identified as acute infarction on diffusion-weighted MRI. Various parameters, including conventional CT values, virtual monoenergetic (monoE) CT values, and electron density relative to water (EDW) were compared between ischemic and normal brain. Statistical analyses, including the Mann-Whitney U test and ROC curve analysis, were performed to assess the diagnostic performance of these parameters. Z test was performed to compare the ROC curves of different parameters.
Results or Findings: This study analyzed 59 lesions in 27 patients to evaluate the diagnostic performance of various spectral parameters. The study found that EDW had the highest area under the curve (AUC) of 0.957, with an sensitivity of 95%, specificity of 85%. MonoE at 100 keV achieved an AUC of 0.955, with high sensitivity (95%) and negative predictive value (94%). In contrast, monoE at 40 keV showed the lowest performance, with an AUC of 0.701 and sensitivity of 66%. Conventional CT images had an AUC of 0.887, comparable to monoE at 70 keV (0.910). Statistically significant differences were noted between the AUCs of EDW/100 keV and conventional CT, while no significant difference was found between EDW and 100 keV.
Conclusion: Our findings indicated that EDW and monoE CT images obtained from DLCT can improve the detection of AIS compared to conventional non-contrast CT imaging.
Limitations: The sample size was small.
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This retrospective study was approved by the local Ethics Committee of Civil Aviation General Hospital (2024-L-K-122).

Author Block: X. Hao, J. Tian, Z. Yao; Shanghai/CN
Purpose: The purpose of this study was to investigate the dynamic changes of the perivascular space diffusion index (ALPS) and its correlation with aquaporin 4 (AQP4) polarization following cerebral ischemia in rats, using advanced diffusion tensor imaging (DTI) technique.
Methods or Background: Rats were divided into a normal group (n=5) and an ischemic group (n=25). The ischemic group underwent transient middle cerebral artery occlusion (tMCAO) and was further subdivided into five subgroups (n=5 each) based on the time post-ischemia (1, 3, 7, 14, and 28 days). Rats underwent MRI scans, including DTI, T2-weighted imaging (T2WI), and susceptibility-weighted imaging (SWI). Subsequently, immunofluorescence staining for AQP4 and glial fibrillary acidic protein (GFAP) was performed. The ALPS index was analyzed based on T2W, SWI, and fractional anisotropy (FA) obtained by DTI post-processing. And regions of interest were selected on the ipsilateral periventricular area, the ipsilateral corpus callosum/cingulate area and their mirror areas of the contralateral side. The AQP4 polarization was analyzed by GFAP/AQP4 in the non-glial scar area around the infarction in cortex and striatum.
Results or Findings: ALPS indexes were markedly reduced in ischemic rats, particularly on the affected side, with a notable drop on day 1, then rising at days 14 and 28. AQP4 polarization mirrored this trend, falling initially, then rising significantly by days 14 and 28. The ALPS index closely aligns with AQP4 index fluctuations.
Conclusion: The DTI-based ALPS index mirrors changes in AQP4 polarization after stroke, dropping sharply in the hyperacute phase and recovering in the early chronic phase, which could be a useful biomarker for glymphatic pathway function following stroke.
Limitations: This experiment requires high consistency of the lesions.
Funding for this study: National Natural Science Foundation of China (No. 81801660) and the grant of National Natural Science Foundation of China (No. 82272061)
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: Institutional Animal Care and Use Committee of Fudan University

Author Block: A. Bhogal¹, S. Uniken Venema¹, P. T. Deckers¹, K. Van De Ven², M. Versluis², K. Braun¹, B. Van Der Zwan¹, J. Siero¹; ¹Utrecht/NL, ²Best/NL
Purpose: Angiography is critical for visualizing cerebral blood flow in intracranial steno-occlusive diseases. Current 4D magnetic resonance angiography (MRA) techniques primarily focus on macrovascular structures, yet few have quantified hemodynamic timing. This study introduces a novel model to estimate macrovascular arterial transit time (mATT) derived from arterial spin labeling (ASL)-based 4D-MRA. We provide examples of our method that visualize mATT differences throughout the brain of patients with intracranial steno-occlusive disease (moyamoya), as well as changes in mATT resulting from the cerebrovascular reactivity (CVR) response to acetazolamide (ACZ).
Methods or Background: The study population consisted of twelve patients with intracranial steno-occlusive disease, with a clinical indication to undergo hemodynamic imaging with an ACZ challenge to measure CVR. CVR is measured using multi-PLD ASL-MRI, acquired pre- and post-ACZ, and includes a four-dimensional dynamic MRA sequence using an ASL-scheme. The scan indications varied, but mostly involved cerebrovascular hemodynamic evaluation before or after neurosurgical intervention.
Results or Findings: We provide examples of our method that visualize mATT differences throughout the brain of patients with intracranial steno-occlusive disease (moyamoya), as well as changes in mATT in response to an ACZ injection. Furthermore, we present a method that projects sparse arterial signals into a 3D native brain-region atlas space and correlates regional mATT with other hemodynamic parameters of interest, such as tissue transit time and CVR.
Conclusion: Our approach offers a non-invasive, quantitative assessment of macrovascular dynamics, which enhances the understanding of large-vessel and tissue-level hemodynamics and augment monitoring of treatment outcomes in steno-occlusive disease patients. This can directly be used in stroke trial stratifications and peri-procedural treatment monitoring. Furthermore, it sets the stage for more in-depth investigations of the macrovascular contribution to brain hemodynamics.
Limitations: Pilot data in a small subgroup of patients (n=12).
Funding for this study: This work was supported by the W.M. De Hoop Foundation, the Janivo Foundation and Friends of UMC Utrecht & Wilhelmina Children’s Hospital, and an NWO VIDI grant awarded to A.A.B. (VI.Vidi.223.085). JCWS is supported by the Brain Center Young Talent Fellowship 2019 of the University Medical Center Utrecht, The Netherlands.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: The Medical Ethics Review Committee of the University Medical Centre Utrecht declared that the Medical Research Involving Human Subjects Act (WMO) did not apply to the present research since all study measures were part of routine clinical practice. All patients or their legal representative (i.e., parent or guardian) provided written informed consent to use their data. Healthy subjects were acquired under a sequence development ethical protocol, which was approved by the Medical Ethics Review Committee of the University Medical Centre Utrecht. Informed consent was given by each healthy subject.

Author Block: S. Uniken Venema, P. Deckers, J. W. Dankbaar, B. Van Der Worp, J. Hendrikse, B. Van Der Zwan, K. Braun, A. Bhogal, J. Siero; Utrecht/NL
Purpose: The primary objective is to describe a clinically feasible advanced neuroimaging protocol developed at an academic medical center that uses multi-delay arterial spin labeling (ASL) and blood oxygen level dependent (BOLD)-MRI with acetazolamide. This protocol is designed to assess cerebrovascular reactivity (CVR) in patients with intracranial steno-occlusive disease (e.g. moyamoya), while avoiding the limitations of PET scans.
Methods or Background: Image acquisition on 3-Tesla MRI involves acetazolamide-augmented multi-delay ASL and dynamic BOLD, in addition to structural sequences. Image processing is done using customized MATLAB-based toolboxes. ASL-CVR is calculated by subtracting pre-acetazolamide cerebral blood flow (CBF) from post-acetazolamide CBF and additional hemodynamic maps, such as arterial transit time (ATT), are generated. Imaging interpretation includes assessment of scan quality and success of the hemodynamic challenge.
Results or Findings: Since 2018, approximately 100 patients were scanned using this protocol. Multi-delay ASL enables a more accurate assessment of CBF and CVR compared to single-delay ASL in patients with prolonged ATT owing to their stenosis, and enables quantifying ATT simultaneously – a useful marker in itself. While CVR assessment is primarily done using ASL-derived CVR maps, BOLD-CVR provided useful complementary information in some cases. A typical patient with an intracranial stenosis experiencing ischemic symptoms will demonstrate lower baseline CBF, lower CVR and prolonged ATT in the affected hemisphere.
Conclusion: Our multi-delay ASL-based protocol demonstrates clinical feasibility and utility, allowing detailed cerebral hemodynamic evaluations of individual patients that is useful for clinical decision-making. This work serves as a practical guide for clinicians and MRI experts seeking to implement these advanced imaging methods in their institutions.
Limitations: Potential pitfalls in imaging acquisition and interpretation, including motion artefacts, inadequate labeling, the effects of anesthesia on CVR and the uncertainties of acetazolamide-augmented BOLD, must be carefully considered.
Funding for this study: This work was supported by the W.M. De Hoop Foundation, the Janivo Foundation and Friends of UMC Utrecht & Wilhelmina Children’s Hospital.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: NedMec (study number 21-406)

Author Block: H. Briody, J. Henry, R. Bruen, P. Mchugh, P. Rohan, M. Javadpour, P. Nicholson; Dublin/IE
Purpose: Favorable cortical venous outflow (VO) is linked to better outcomes in acute ischemic stroke. It’s role in aneurysmal subarachnoid hemorrhage (aSAH) remains unclear. This study investigates the association between VO profiles and functional outcomes in aSAH.
Methods or Background: Patients with aSAH referred to a tertiary neurosurgery center between 2016 and 2023 were included if presentation computed tomographic angiography (CTA) demonstrated satisfactory venous system opacification. VO was assessed using the cortical vein opacification score (COVES). The primary outcome was poor functional outcome (Glasgow Outcome Scale [GOS] 1-3) at 90 days. Associations between COVES and outcomes were assessed using univariable and multivariable (adjusted for World Federation of Neurosurgical Societies [WFNS] grade) binomial regression.
Results or Findings: Of 675 patients with aSAH and available CTA, 204 (30%) met inclusion criteria. The median age was 54.2 years (range 12-85). 182 (89%) had favorable VO. No significant association was found between unfavorable VO and poor 90-day functional outcome (RR 0.78, 95% CI 0.38-1.58, p=0.48), even after adjusting for WFNS grade (RR 1.04, 95% CI 0.38-2.83, p=0.94). Unfavorable VO was not associated with poor outcome at discharge, need for cerebrospinal fluid diversion, or shunt dependence.
Conclusion: The study failed to demonstrate a link between venous outflow (as measured by COVES) and outcomes in aSAH. This is an important negative finding. It suggests that, unlike in ischemic stroke, venous outflow might not be a major determinant of outcome in aSAH. Prospective studies are needed to definitively assess the role of VO in aSAH.
Limitations: This retrospective, single-center study may be subject to selection bias due to the inclusion criteria requiring adequate jugular bulb opacification on CTA. The impact of delayed cerebral ischemia, a major determinant of outcome in aSAH, was not specifically assessed.
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Institutional review board

Author Block: D. Botta, A. Cusin, L. Sveikata, K-O. Loevblad, F. T. Kurz; Geneva/CH
Purpose: The aim of this study is to evaluate the detection of cerebral microbleeds (CMBs) and cortical microinfarcts using 7 Tesla (7T) MRI in patients with cerebral amyloid angiopathy (CAA) and to correlate these findings with cognitive performance as measured by the Montreal Cognitive Assessment (MoCA).
Methods or Background: 19 patients with probable or possible CAA were scanned using both 7T and 3T MRI. Imaging protocols at 7T included susceptibility-weighted imaging (SWI) for CMB detection with an in-plane resolution of 0.15x0.15mm, and 3D FLAIR and T1 MP2RAGE sequences for cortical microinfarcts detection. Cognitive function was assessed using the MoCA score. Statistical analyses were conducted to assess correlations between lesion burden and MoCA scores.
Results or Findings: 7T MRI detected 379 CMBs compared to 179 at 3T (average: 18.37±34.18 CMBs at 7T vs 9.53±15.03 at 3T). Additionally, 7T identified 27 cortical microinfarcts across 5 patients, while none were detected at 3T. MoCA scores ranged from 9 to 30 with a mean of 23.0±4.82. Weak correlations were found between the number of CMBs and MoCA scores (7T: r =-0.28; 3T: r =-0.24).
Conclusion: 7T MRI at high-resolution is superior to 3T for detecting CMBs and cortical microinfarcts in CAA patients. However, the weak correlation between lesion burden and cognitive decline suggests that other factors may also contribute to cognitive impairment in these patients.
Limitations: Not applicable.
Funding for this study: No funding was provided for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Written informed consent was obtained from all participants, the study was approved by the institutional review board of Geneva University Hospitals.

Author Block: H. Chen¹, S. Xu², G. Zhang², J. Wang¹, X. Yin¹; ¹Baoding/CN, ²Shanghai/CN
Purpose: Intracranial aneurysms and infundibula with similar morphology and anatomical location are difficult to distinguish using low-dose cerebral CT angiography (CTA). This study evaluated the diagnostic performance of cerebral low-dose CTA with artificial intelligence iterative reconstruction (AIIR) for differentiating intracranial aneurysms and infundibula.
Methods or Background: Sixty-four patients (38 male, mean age 62.2 ± 12.5 years) with suspected intracranial aneurysms were prospectively enrolled. Each patient underwent routine-dose (RD) and low-dose (LD) cerebral CTA. The RD protocol used 100kVp, ref. 180mAs, and hybrid iterative reconstruction (HIR), whereas the LD protocol used 100kVp, ref. 30mAs, and AIIR. Two radiologists, blinded to scan/reconstruction parameters, independently detected aneurysms and infundibula on low-dose images. The diagnostic reports of RD CTA served as references. Diagnostic performance in differentiating aneurysms and infundibula was assessed using receiver operating characteristic (ROC) analysis, calculating sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, and area under the curve (AUC) with 95% confidence interval (CI).
Results or Findings: A total of 64 lesions were identified, including 44 aneurysms and 20 infundibula. Reader 1 detected 62 out of 64 lesions (96.9%) on low-dose images, while Reader 2 detected 61 out of 64 (95.3%). Two infundibula with sizes of 1.5 mm and 2.4 mm were missed by both readers, whereas one aneurysm of 1.8 mm was missed by Reader 2. In differentiating aneurysms and infundibula, the sensitivity, specificity, PPV, NPV, diagnostic accuracy, and AUC for Reader 1 were 97.72%, 100%, 100%, 94.74%, 98.39%, and 0.989 (95% CI: 0.963–1.015), while for Reader 2 they were 97.67%, 100%, 100%, 94.74%, 98.36%, and 0.988 (95% CI: 0.962–1.015).
Conclusion: The AIIR shows the potential in reducing the radiation dose of the cerebral CTA when diagnosing intracranial aneurysms and infundibula.
Limitations: Not applicable.
Funding for this study: the Key Research and Development Program of Hebei Province (grant number 202330604010017)
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was approved by the local institutional review board.

Author Block: S. Nataf¹, O. Curtinot², T. Maghfour², G. Boulouis², A. Aignatoaie¹, C. Ozsancak¹, M. Pasi², C. Cohen¹; ¹Orleans/FR, ²Tours/FR
Purpose: Intracerebral hemorrhage (ICH) is associated with poor outcome. Identifying patients with higher risk of disability is a key feature of optimal care. Recently, temporal muscle thickness (TMT) has been shown to predict ability after ischemic stroke. We explored the relationship between temporal muscle trophicity and the functional ability of non-traumatic ICH patients.
Methods or Background: Patients with acute non-traumatic ICH (2021-2022) from two university centers were retrospectively included. Imaging ICH parameters were retrieved (e.g. volume, location). Temporal trophicity was assessed through thickness (TMT), area (TMA) and density (TMD) on baseline brain computed tomography (CT). Good prognosis at 3-6 months was defined as a modified Rankin Scale (mRS)<3. Association between TMT, TMA and TMD and 3-6months mRS was evaluated using student T-test.
Results or Findings: Among 453 ICH patients, 216 with both CT and 3-6months mRS were analyzed (49% women, 70±15.9y). Mean hemorrhage volume was 47.3±46mm³, mainly lobar (46%) and located in basal ganglia (41%). Mean TMT was 7.32±2.70mm in the good prognosis group, vs 6.37±2.13mm in the poor prognosis group (p=0.005). Mean TMA was 408±203mm² in the good prognosis group, vs 343±161mm² in the poor prognosis group (p=0.01). Prognosis was not significatively associated with TMD (p=0.051). 3-6months mRS was associated with hemorrhage volume (p=0.01).
Conclusion: To our knowledge, this is the first study to analyze and demonstrate that temporal muscle trophicity serves as a prognostic factor in intracerebral hemorrhage (ICH). Temporal muscle thickness (TMT) is a simple and practical indicator for assessing overall health and guiding patients toward targeted rehabilitation.
Limitations: Since many patients underwent only MRI, further research is needed to evaluate the association between temporal muscle trophicity and prognosis using this imaging modality.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was approved by a Comité de protection des personnes (CPP).

Author Block: A. Akhavi Milani; Tabriz/IR
Purpose: To evaluate the performance of 3D variable flip angle turbo spin echo black-blood MRI (BB-MRI) in diagnosing cerebral venous thrombosis (CVT). The secondary objectives included, comparing BB-MRI with conventional MRI, MRV, MPRAGE, and SWI in diagnosing CVT, and assessing the utility of BB-MRI in estimating thrombus age.
Methods or Background: This study was registered in PROSPERO [ID: blinded]. The PubMed/MEDLINE, Web of Science, Scopus, and Embase databases were systematically searched and studies were selected based on predefined eligibility criteria. The risk of bias was assessed using the QUADAS-2 tool. Meta-analysis was performed to calculate pooled sensitivity, specificity, and AUC.
Results or Findings: Nine studies were included in the review. Two of them insufficiently reported the quantitative data; therefore, seven studies involving 176 CVT patients and 217 controls, encompassing 610 thrombosed and 3,279 normal cerebral venous segments were included in the meta-analysis. The studies demonstrated a high bias risk in the patient selection and reference standard domains. The pooled sensitivity and specificity of BB-MRI were 96% [95% CI: 92%–98%] and 96% [95% CI: 93%–98%] on a patient-based level, and 92% [95% CI: 87%–95%] and 98% [95% CI: 92%–99%] on a venous segment-based level, respectively. The AUC was 0.98 for patient-based data and 0.96 for venous segment-based data. For secondary objectives, a narrative summary indicated that BB-MRI outperforms conventional MRI, MRV, and MPRAGE. It also outperforms SWI in assessing cerebral cortical veins. Moreover, BB-MRI can prove beneficial in thrombus age estimation.
Conclusion: BB-MRI demonstrates significant potential in diagnosing CVT. Further comparative studies are required to specify its role in clinical decision-making for CVT.
Limitations: The small number of the retrieved studies.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: This was a systematic review study for which ethics committee approval was waived.

Author Block: A. Hillal¹, T. Ullberg², J. Wassélius¹; ¹Lund/SE, ²Malmö/SE
Purpose: Early identification of the underlying cause of intracerebral hemorrhage (ICH) is important for treatment and prognosis. This study aims to investigate the association of hematoma volume and other clinical parameters on the distribution of Cerebral Amyloid Angiopathy (CAA) probability according to the simplified Edinburgh CT criteria in a large, unselected intracerebral hemorrhage (ICH) population.
Methods or Background: Patients with spontaneous ICH residing in Skane county registered with clinical data in the Swedish Stroke Register 2016–2020 were included. Radiological parameters were evaluated using baseline non-contrast CT (NCCT) for categorization according to the simplified Edinburgh CT criteria by the presence of subarachnoid hemorrhage (SAH) and fingerlike-projections (FLP). Multivariable logistic regression analysis was used to determine factors associated with an increased (intermediate/high) CAA probability.
Results or Findings: Of 666 patients with lobar ICH, 190 (29%) had high CAA probability, 92 (14%) intermediate, and 384 (58%) low CAA probability. Patients with increased CAA probability more often presented with decreased level of consciousness, larger hematoma volumes, and had higher 90-day mortality. Female sex, age, and increasing baseline hematoma volume (Odds Ratio up to 30) were associated with increased odds of having an increased CAA probability.
Conclusion: We identified a strong association between baseline hematoma volume and an increased probability of CAA in lobar ICH patients on NCCT, indicating that large hematoma volumes per se may contribute to the occurrence of FLP and SAH, and act as a confounder for the Simplified Edinburgh CT Criteria. Validation against MRI is warranted.
Limitations: The lack of MRI studies to allow for the correlation between CT imaging characteristics and the gold imaging standard MRI Boston criteria
Funding for this study: ALF grants to Teresa Ullberg and Johan Wasselius, the Crafoord Foundation to JW, VINNOVA to Johan Wasselius, and by SUS Stiftelser & Fonder to Johan Wasselius.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was approved, and individual informed consent was waived by the Swedish Ethical Review Authority (reference number 2020-06800).

Author Block: C. M. Offersen, J. Johansen, A. H. Brandt, T. C. Truelsen, A. Pai, S. Darkner, M. B. Bachmann Nielsen, J. F. Carlsen; Copenhagen/DK
Purpose: Moderate inter-rater variability of Diffusion-Weighted Imaging (DWI) – Fluid-Attenuated Inversion Recovery (FLAIR) mismatch in wake-up stroke raises concerns about the potential exclusion of eligible patients for treatment with thrombolysis. A novel FLAIR algorithm has shown promising potential to perform this mismatch assessment but has only been evaluated on a small dataset. In the present study, we aimed to evaluate an updated version of the FLAIR algorithm for predicting the DWI-FLAIR mismatch in a large cohort of wake-up stroke patients.
Methods or Background: We conducted a single-centre, retrospective study. A consecutive cohort of patients suspected of wake-up stroke, who underwent MRI between 2019 and 2021 was included. Two radiologists and one resident, blinded to clinical data, manually assessed DWI-FLAIR mismatch according to the current clinically used binary categorisation. Cohens Kappa was calculated for the inter-rater agreement. The FLAIR algorithm depends on a DWI segmentation. We used a commercial DWI segmentation model and then tested the ability of the FLAIR algorithm on the identified ischemic lesions to predict manual mismatch. This was analysed with logistic regression test.
Results or Findings: The DWI model identified 495 patients with possible ischemic lesions. Manual radiological assessments found 365 of those patients to have actual ischemic lesions. Inter-rater agreement for binary DWI-FLAIR mismatch assessment was moderate (κ = 0.461 ± 0.028SD). We found a high accuracy (82.3 ± 3.3SD). Low sensitivity (59.1 ± 9.1SD). Specificity (91.5 ± 4.2SD) and AUC (0.845 ± 0.03SD) were high in the automatic mismatch assessment.
Conclusion: The FLAIR algorithm predicted DWI-FLAIR mismatch status with a high AUC, which suggests the algorithm could provide a more standardized decision on mismatch, and reduce the inter-rater variability through an objective assessment to assist the radiologist.
Limitations: This was a single-centre, retrospective study.
Funding for this study: Innovation Fund Denmark
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Danish National Center for Ethics