Iodine-based dual-energy CT predicts early neurological decline from cerebral edema after large hemispheric infarction

Background & Purpose: Ischemia affecting two thirds of the MCA territory predicts development of malignant cerebral edema. However, early infarcts are hard to diagnose on conventional head CT. We hypothesize that high-energy (190keV) virtual monochromatic images (VMI) from dual-energy CT (DECT) imaging enables earlier detection of secondary injury from malignant cerebral edema (MCE). Methods: Consecutive LHI patients with NIHSS ≥ 15 and DECT within 10 hours of reperfusion from May 2020 to March 2022 were included. We excluded patients with parenchymal hematoma-type 2 transformation. Retrospective analysis of clinical and novel variables included VMI Alberta Stroke Program Early CT Score (ASPECTS), total iodine content, and VMI infarct volume. Primary outcome was early neurological decline (END). Secondary outcomes included hemorrhagic transformation, decompressive craniectomy (DC), and medical treatment of MCE. Fisher’s exact test and Wilcoxon test were used for univariate analysis. Logistic regression was used to develop prediction models for categorical outcomes. Results: Eighty-four LHI patients with a median age of 67.5 [IQR 57,78] years and NIHSS 22 [IQR 18,25] were included. Twenty-nine patients had END. VMI ASPECTS, total iodine content, and VMI infarct volume were associated with END. VMI ASPECTS, VMI infarct volume, and total iodine content were predictors of END after adjusting for age, sex, initial NIHSS, and tPA administration, with a AUROC of 0.691 [0.572,0.810], 0.877 [0.800, 0.954], and 0.845 [0.750, 0.940]. By including all three predictors, the model achieved AUROC of 0.903 [0.84,0.97] and was cross validated by leave one out method with AUROC of 0.827. Conclusion: DECT with high-energy VMI and iodine quantification is superior to conventional CT ASPECTS and is a novel predictor for early neurological decline due to malignant cerebral edema after large hemispheric infarction.


Introduction
Malignant cerebral edema (MCE) after large hemispheric infarction (LHI) leads to high levels of morbidity and mortality.Previous studies have investigated multiple variables linked to the development of malignant cerebral edema (MCE), especially early neurological decline (END) and life-threatening cerebral herniation after large hemispheric infarction.2][3][4] Initial computed tomography perfusion (CTP) ratio of cerebral blood volume (CBV) to cerebral spinal uid (CSF) was shown to have a sensitivity and speci city greater than 95% for predicting a malignant infarct. 5However, in this small study, the primary outcome of malignant infarct included a decreased level of consciousness de ned only as an increase in NIHSS ≥ 1.A prospective study of MRI diffusion-weighted imaging (DWI) infarct volume > 82 mL had a sensitivity of 98%, but low speci city for predicting malignant infarction. 6Using automated segmentation and a neural-based algorithm, it was shown that the initial CSF volume to tissue ratio and the progressive decline in CSF volume were predictors of cerebral edema formation. 3,4Nonetheless, routine identi cation of LHI patients at risk of clinical worsening and even death due to progressive malignant cerebral edema remains problematic.
The pathophysiology of cerebral edema after ischemic stroke is a complex, multifaceted cascade of events that includes cytotoxic, ionic, and vasogenic edema, all of which contribute to an increase in brain swelling.][9][10][11] As edema progresses, oncotic cell death occurs and endothelial cell dysfunction results in extravasation of blood components, leading to hemorrhagic transformation (HT) and potentiation of further insult. 7,10,11Therefore, identifying individuals at risk of END due to MCE is a primary goal in initial triage and resuscitation of stroke patients with LHI.3][14][15] As the indications for acute mechanical thrombectomy for AIS with established large infarct continue to expand, the precedence of determining whom may worsen clinically, bene t from initial care at a comprehensive stroke center, or warrant acute neurosurgical intervention continues to elevate. 16,17Unfortunately, during the rst 48 hours of LHI resuscitation, risk strati cation generally depends on poorly-predictive data such as the neurological exam, midline shift, and clinical judgement.Advanced imaging techniques may address this practice gap by improving the rapid triage of these severely ill AIS patients.
The clinical utility of DECT in stroke has increased over the past decade, especially in the determination of an end stage of cerebral edema after AIS-hemorrhagic transformation.Through enhanced reconstructions from DECT, high-energy (190 keV) VMI allows the suppression of iodine attenuation, which has a similar pixel appearance as acute blood products on CT brain imaging. 18This CT technology also reduces beam hardening artifact, improves contrast resolution, and decreases image noise. 19everal studies have demonstrated high sensitivity for discriminating hemorrhagic transformation from iodine extravasation after mechanical thrombectomy (MT) utilizing dual-energy technology. 18,20,21VMI also has the ability to more accurately estimate the ischemic core compared to conventional non-contrast CT (NCCT) imaging. 22The likely explanation for this enhanced performance is VMI's ability to highlight brain water content (BWC) and hypoattenuation, which on a fundamental level radiographically represents the early cascade of cytotoxic, ionic, and vasogenic edema.
Despite the improved visual delineation of cerebral edema and ability to distinguish between hemorrhagic transformation and iodinated contrast extravasation after AIS and MT, limited investigations exist quantifying iodine contrast extravasation on DECT as an imaging biomarker of blood brain barrier (BBB) disruption.To ful ll an unmet need for advanced techniques to rapidly quantify cerebral edema in LHI patients, routinely available imaging biomarkers must be established and may help identify those most appropriate for intensive neuromonitoring and early treatment in a comprehensive stroke center (CSC) and neurocritical care unit (NCCU).Like the concept of MRI hyperintense acute reperfusion marker (HARM) noted on FLAIR imaging due to gadolinium-based contrast agents crossing an early, disrupted BBB after AIS, we suspect that the leakage of iodinated contrast represents a CT-based equivalent, representative of cerebral edema development. 23As rapid MRI imaging procurement is not always clinically feasible for triage purposes, we postulate DECT's ease of obtainment and performance may be the key to its widespread deployment.We hypothesize improved prediction of END due to MCE after LHI using high energy (190 keV) monochromatic images and quanti cation of iodinated contrast extravasation from DECT data.

Study Design and Participants
We completed a retrospective, observational study of a prospective, single institutional database (NCT0418947) of acute anterior circulation, large hemispheric infarction (LHI) patients who underwent intra-arterial (IA) mechanical thrombectomy (MT) and study-site standard post-reperfusion DECT within 10 hours of reperfusion.This study was approved by the University of Maryland School of Medicine Institutional Review Board and complies with ethical standards.For this type of retrospective study, informed consent is not required.LHI was de ned as NIHSS ≥ 15 at the time of presentation of initial stroke syndrome.Exclusion criteria included patients who did not undergo standard post-IA therapy DECT imaging, signi cant post-MT clinical improvement de ned as a 24-hour NIHSS < 12, and signi cant hemorrhagic transformation classi ed as parenchymal hematoma type 2 (PH-2).We followed the Strengthening the reporting of observational studies in epidemiology (STROBE) checklist in the design of this study. 24

Data Collection
6][27][28] Stress glucose index was determined by rst calculating estimated average glucose from hemoglobin A1c (28.7 * A1c -46.7), then calculating ratio of initial glucose measurement to estimated average glucose. 25,26 nitions & Novel Variables Iodine density: software automated quanti cation in a ROI (region of interest) of brain parenchymal iodine extravasation measured in mg/mL on iodine images.

VMI infarct volume: automated volumetric quanti cation of infarct volume (mL) on post-MT VMI CT.
Total iodine content: total iodine extravasation (mg) post-MT quanti ed on iodine images in infarcted cerebral territory.

Imaging Technical Analysis
We completed a retrospective analysis of both initial acute non-contrast, conventional CT (NCCT) and post-MT VMIs.DECT novel variables included quanti ed VMI CT ASPECTS and iodine extravasation variables including maximum iodine density in infarcted hemisphere, max iodine densities in corresponding ASPECTS locations, and calculated total iodine content in infarcted cerebral hemisphere.
The study site utilized both Siemens and Philips DECT scanner technology for standard of care in management of AIS patients.

Volume of Infarction
Using the Philips Intellispace Portal CT viewer application, we conducted infarct volumetric calculations of the high energy (190 keV) VMIs.Utilizing grey-level mapping, the 190 keV sequence was placed in a window width of W30:L30 or standard stroke viewing. 29A volume explorer application and "smart segmentation" tool allowed the use of a 3-D smart ROI to delineate the edges of infarcted tissue of the affected hemisphere.This border was adjusted along the infarct edge on each single axial slice to create a rendered lesion mapping and an automated infarct volume in milliliters (mL).To investigate the sensitivity of VMI infarct volume quanti cation in comparison to standard of care MRI DWI, two board certi ed neuroradiologists quanti ed nal DWI infarct volume (mL) for rst 21 patients with MRI imaging for comparison.Utilizing the manual lesion segmentation tool in Carestream Vue, PACS (Carestream Health Inc., Rochester, NY), the ROI delineating the infarct were manually segmented on individual slices from the DWI (b1000) sequence and the nal infarct volume (mL) was obtained through automated summation.

Iodine Extravasation Quanti cations
For post image processing and iodine quanti cation, we implemented the modi ed brain hemorrhage application on the post-processing workstation (syngo.via,version VB10B; Siemens Healthcare, Forchheim, Germany) or the spectral CT viewer software (Philips Intellispace Portal; Philips Healthcare, Best, the Netherlands).For each DECT scan, the maximum iodine density in the affected hemisphere was determined using a free-hand ROI tool with automated calculation of iodine density (mg/mL).The nal maximum iodine density was calculated by an averaged value in three planes: axial, coronal, and sagittal in the affected hemisphere.Further iodine density extrapolation was conducted by isolating each region of the ASPECTS score (caudate, insular, internal capsule, lentiform, M1, M2, M3, M4, M5, and M6) and again utilizing a ROI tool, an iodine density averaged in 3 planes was determined (mg/mL) in each corresponding location.
To estimate the novel DECT variable of total affected hemisphere iodine content (mg), we completed a 2step process.First, the mean of the iodine densities (mg/mL) of the ten ASPECTS regions was calculated.Secondly, the product of the VMI infarct volumes (mL) and the mean iodine densities of the affected hemisphere (mg/mL) was determined, and this value was de ned as the affected hemisphere total iodine content (mg).

Outcomes
The primary outcome was de ned as the development of END, using a composite outcome variable of both clinical worsening (increase NIHSS ≥ 4 or decrease in GCS > 2) or malignant radiographical edema (midline shift ≥ 5mm at the level of the septum pellucidum).Secondary outcomes included hemorrhagic transformation, implementing the Heidelberg bleeding classi cation (HI-1: scattered small petechiae, no mass effect, HI-2: con uent petechiae, no mass effect, or PH-1: hematoma within infarcted tissue, occupying < 30%, no mass effect), need for decompressive craniectomy (DC), and medical treatment of MCE with hyperosmolar therapy, sedation, or hyperventilation. 30

Statistical Analysis
Statistical analyses were performed by SAS version 9.4 (Cary, NC, USA).Fisher's exact test and Wilcoxon Rank Sum test were used for univariate analysis.For multivariate analysis, both known clinical predictors of MCE de ned by previous literature and variables of signi cance during the univariate analysis were tted into the model.The summary statistics of the area under the receiver operating characteristic curve (AUROC) were calculated to measure prediction performance and the nal model was cross validated by leave one out procedure.Linear regression modeling was used to relate MRI DWI infarct volumes to VMI infarct volumes to determine sensitivity of the VMI infarct volume calculation.

Data Availability
Anonymized data not published within this article will be made available by request from any quali ed investigator.

Results
From May 2020 until March 2022, eighty-four LHI patients with median age of 67.5 [IQR 57,78] years and median initial NIHSS 22 [IQR 18,25] met inclusion criteria (Table 1).Forty-four (52.4%) patients were female, and majority of patients were either White (37 patients, 44%) or Black (45 patients, 53.6%).The two groups were well matched in relation to medical comorbidities other than those with HTN, who were less likely to have END (p=0.037)No differences were noted in volume of iodinated contrast administration (p=0.183),patients with congestive heart failure (CHF) (p=0.166),baseline serum creatinine (p=0.157) or GFR (p=0.509) in relation to contrast clearance.Linear regression of VMI infarct volume vs. MRI DWI infarct volume for rst 21 patients who underwent MRI imaging demonstrated a coe cient of determination (R 2 ) of 0.991, p<0.001 (Figure 1).No statistically signi cant difference was noted in 90-day post-discharge mRS (modi ed Rankin Score) (p=0.433) or when dichotomized to favorable (mRS ≤ 3) or unfavorable (mRS 4-5) outcome (p=0.594).

Early Neurological Decline
A total of 29 patients (34.5%) met the primary endpoint of END, with 8 of the 29 (27.6%)patients having NIHSS increase ≥ 4, 21 (72.4%)patients developing MLS ≥ 5mm, and 22 (75.9%)patients having GCS decline > 2. Mean time from last known normal to meeting the primary endpoint was 41 hours and 49 minutes, with the END occurring on average on hospital day 2.03.In a univariate analysis, NIHSS at 24hrs., GCS at 24 and 48 hrs., post-MT ASPECTS on NCCT and VMI, delta ASPECTS between NCCT and VMI, and VMI infarct volume were all associated with END, with p-values noted in Table 1.The median VMI infarct was 159.9 mL in the END group vs. 62.4 mL in those not in the END group, p<0.001.Novel iodine related variables associated with END that were statistically signi cant include iodine densities (mg/mL) in the insular, internal capsule, M1, M4, M5, M6 locations (Table 2).Maximum iodine density for the affected hemisphere trended towards but did not reach signi cance (p=0.069).The median total iodine content for the END group was 62.3 mg vs. 21 mg, p<0.001 (Figure 2).In a linear regression comparing total iodine content and infarct volume, a positive correlation was found, R 2 = 0.674, p <0.001 (Figure 3).

Secondary Outcomes
In an analysis of the three secondary outcomes, 34 patients (40.1%) developed HT, 11 patients (13.1%) received medical treatment for MCE, and 7 patients (8.3%) underwent DC.Only one demographic variable was associated with HT-stress glucose ratio (p=0.034).Medical treatment for MCE was associated with race (p=0.032),younger age (p=0.003), and higher baseline creatinine (p=0.012).DC was associated with younger age (p=0.0002),higher baseline creatinine (p=0.0007), and lower GCS at 24 and 48 hours (p=0.02 and 0.014).With relation to novel variables, lower VMI ASPECTS was associated with all 3 secondary outcomes.Higher total iodine content was associated with HT (59.57mg in HT group vs. 35.53mg in non-HT group, p<0.05) and need for DC (65.3 mg in DC group vs. 37.9 mg in non-DC group, p=0.005).Increased VMI infarct volume was associated with medical treatment for MCE (163.6 mL medical treatment group vs. 99.1 mL non-treatment group, p=0.012) and DC (179 mL in DC group vs. 101.1 mL in non-DC group, p=0.013), but not HT.Iodine density in the deep locations of the ASPECTS including caudate, insular, internal capsule, and lentiform all were signi cantly higher in the HT group (p<0.05),but only the ID in the cortical locations of M1 (p=0.04) and M5 (p=0.013) were associated with DC.

Discussion
We investigated several novel DECT biomarkers using VMIs and iodine related variables to aid in the prediction of END after LHI.We used a composite score of END, including clinical decline and radiographical worsening, to capture the cascade of effects of malignant cerebral edema.In this at-risk population of LHI patients who underwent MT and post-reperfusion DECT imaging, we demonstrated that iodine contrast extravasation quanti cation, speci cally total iodine content, and VMI infarct volume both were strong predictors of END.DECT in comparison to conventional CT, therefore better re ects total brain water content, BBB disruption, and the secondary injury that occurs through cytotoxic, ionic, and vasogenic edema.
The volume of ischemic injury can be rapidly measured after reperfusion without the need for MRI, as all VMIs from DECT data were obtained within 10 hours after MT.In comparison to gold-standard MRI DWI infarct volumes, VMIs delineate early ischemic changes equally well and outperforms conventional NCCT. 31An additional bene t of DECT is the rapid quanti cation of iodinated contrast.As infarct volume increased in this study, so did the quantity of iodinated contrast extravasation.This likely represents the degree of BBB disruption with a large contribution from reperfusion injury after MT secondary to hyperemia, continued cerebral metabolic depression and worsened paracellular permeability. 32anti cation of iodine contrast extravasation on DECT, speci cally as a marker of BBB disruption, showed a strong association with secondary outcomes and appears to be a useful radiographic biomarker.Several studies have demonstrated the utility of DECT to differentiate acute blood products and iodinated contrast extravasation post-MT.Early differentiation is key, given most cases of postthrombolytics or post-MT reperfusion hemorrhagic transformation occurs in the rst 24 hours after intervention. 33,34In our cohort, larger infarct burden demonstrated by ASPECTS on both conventional NCCT and VMIs was associated with HT, with iodine leakage in the deep structures also associated with HT.Commonly, HT after AIS can be asymptomatic; however, it has been shown to worsen 3 month outcomes measured by mRS, suggesting another potential pharmacological target aimed at avoidance of endothelial dysfunction, oxidative stress, leukocyte in ltration, and vascular activation. 10,11,32Medical treatment for MCE was infrequent in this cohort, despite clinical and radiographical evidence of MCE.Medical treatment, unlike HT, was more associated with the total infarct volume of affected tissue as compared to the leakage of contrast only to deep structures.Surgical decompression occurred in less than 10% of our cohort, but similarly to medical treatment, occurred in younger patients with worsened baseline renal function both likely due to clinical decision making and limitations in ICP monitoring.Quanti cation of both tissue injury by infarct volume and BBB disruption through total iodine content were associated with the need for DC.All patients who underwent DC did so in the rst 4 days after onset of stroke symptoms.Early identi cation of LHI patients who require DC and appropriate triage to a comprehensive stroke center is paramount.
The retrospective review design of this study is the primary limitation.Despite excluding END due to fever, sedation, or seizure, confounding by unmeasured variables remains possible.Routine use of DECT imaging on initial evaluation, pre-MT, did not occur until midway through this study, therefore we were unable to compare initial VMIs to post-MT imaging, a future direction of this work.Contrast clearance post-MT may be impacted by heart failure or renal disfunction.The administration of contrast during MT is protocolized with no difference between the patients who developed END and those who did not.Also, our cohort is small, although our goal was proof of clinical utility and prospective evaluation is necessary moving forward.Future investigations must be directed towards utilization of the high energy (190 keV) VMI derived from DECT as the initial imaging modality in acute stroke investigation which is now standard of care at our center.Given its success at determining the infarct core compared to MRI, we anticipate its greatest value lies in its use as a clinical triage tool when evaluating risk of early neurological decline 22 Automated quanti cation of iodinated contrast extravasation and infarct volumetrics will aid in the rapid utility of the DECT in AIS triage.The ease of obtainment, readily available technology, and shortened time outside of the ICU make DECT imaging a promising acute neuroimaging modality after LHI. 35

Conclusion
VMIs and iodine quanti cation from DECT is superior to conventional CT ASPECTS for demonstration of cerebral infarction burden and is a novel predictor for early neurological decline due to malignant cerebral edema after large hemispheric infarction.manuscript creation and critically editing for important intellectual content, Guarav Jindal: imaging analysis, manuscript creation and critically editing for important intellectual content, Uttam Bodanapally: study design and creation, imaging analysis, manuscript creation and critically editing for important intellectual content, J Marc Simard: manuscript creation and critically editing for important intellectual content, Neeraj Badjatia: manuscript creation and critically editing for important intellectual content, Gunjan Parikh: senior author, study design, conceived and designed analysis, imaging processing, manuscript creation and critically revising for important intellectual content.
. All authors have agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.