Reliability of CT,DECT,and MRI for the diagnosis of hemorrhagic transformation after thrombectomy

Abstract

Introduction:

Computed Tomography (CT) is the main modality used for the diagnosis and classification of hemorrhagic transformation (HT) after thrombectomy, however its reliability has shown limitations. Dual-energy CT (DECT) and magnetic resonance imaging (MRI) have been suggested to enhance the reliability of HT detection and classification, but direct three-way comparison of these modalities is lacking. To measure and compare the reliability of CT, DECT and MRI for the diagnosis, classification, and therapeutic consequences of HT after thrombectomy.

Patients and methods:

Between June 2017 and September 2019, 66 of 324 patients included in the BP-TARGET trial underwent CT, DECT and MRI scans within 36 h after thrombectomy. Seven readers, including three neurologists, two diagnostic, and two interventional neuroradiologists independently reviewed the images. They were asked for each patient and each imaging modality to score the presence of a hemorrhagic transformation (of any type), the type of hemorrhagic transformation according to the European Cooperative Acute Stroke Study (ECASS), and whether they would start the patient on antiplatelet based on the imaging finding. The readers repeated the same readings 1 month later. Interrater and intrarater agreement were measured using Kappa statistics.

Results:

There were frequent discrepancies between CT, DECT and MRI scans evaluations. The use of MRI led to an increased rate of HT diagnosis compared to CT and DECT scans. Interrater agreement for ECASS classification was only fair-to-moderate for all three imaging modalities but improved to a substantial level after dichotomization into 0/HI1/HI2 versus PH1/PH2. The interrater agreement for the decision to start antiplatelet therapy was substantial only with CT (κ = 0.636 [0.577–0.694]) and remained moderate with MRI and DECT.

Conclusion:

In our study, the imaging modality influenced the diagnosis and classification of HT, the management of antiplatelet therapy, and the interrater and intrarater agreement. These findings may guide the choice of imaging modality in research or clinical settings.

Graphical abstract

Keywords

Stroke haemorrhagic transformation agreement reliability CT scan MRI dual energy CT antiplatelet therapy

Introduction

Hemorrhagic transformation (HT) is a common occurrence after acute ischemic stroke (AIS), affecting up to 40% of patients.¹ The diagnosis and classification of HT is an important element of stroke care as certain types of HT have been associated with a worse outcome and increased mortality.² It is also an important step in deciding when to initiate secondary preventive treatment with antithrombotic therapy. The 2019 American guidelines³ recommend a follow-up CT or MRI scans at 24 h after thrombolysis. Brain imaging is similarly often performed 24 h after thrombectomy.^4–9 One of the main classifications widely used is the European Cooperative Acute Stroke Study (ECASS) classification¹⁰ which includes five categories: no hemorrhagic transformation, hemorrhagic infarction (HI) type 1, hemorrhagic infarction type 2, parenchymal hematoma (PH) type 1, and parenchymal hematoma type 2.

Unenhanced head computed tomography (CT) is the main imaging modality used in this context. Previous studies have reported a lack of agreement between physicians for the diagnosis and classification of HT on brain CT^11,12 including after thrombectomy.^13,14 One of the main reasons given is that arterial injection of iodine contrast during thrombectomy may cause contrast staining mimicking HT on follow-up CT.¹⁴ This lack of agreement could lead to large variations in the initiation of secondary preventive treatment with antithrombotic therapy.¹⁵ The use of dual-energy CT (DECT)^16,17 or magnetic resonance imaging (MRI)^11,12,18 have been suggested to enhance the reliability of HT detection and classification, but direct three-way comparison of these modalities is lacking. Using a subset of patients enrolled in the “Safety and efficacy of intensive blood pressure lowering after successful endovascular therapy in acute ischaemic stroke” (BP-TARGET) trial¹ who all underwent DECT, CT and MRI between 24 and 36 h after thrombectomy, we aimed first to determine whether DECT and MRI change the diagnosis and classification of HT after thrombectomy compared to unenhanced head CT alone. In a second part, we aimed to evaluate the reliability of each imaging modality for the diagnosis and classification of HT as well as the clinical decision to initiate antiplatelet therapy.

Patients and methods

Ethics

This retrospective study was approved by the research ethics board at our institution; the need for informed consent was waived. All data were anonymized and de-identified. The corresponding author had access to all data. Anonymized data are available on reasonable request to the corresponding author. This study was prepared in accordance with the Guidelines for Reporting Reliability and Agreement Studies.¹⁹

Patient selection

We analyzed imaging data from the “Safety and efficacy of intensive blood pressure lowering after successful endovascular therapy in acute ischaemic stroke” (BP-TARGET) trial.¹ This study included patients who underwent successful endovascular treatment (defined by a mTICI score of 2b or more) for acute stroke with anterior large vessel occlusion (i.e. internal carotid and/or proximal middle cerebral artery -M1- occlusions) between June 2017 and September 2019. The primary outcome was the rate of HT on follow-up CT scan 24–36 h after reperfusion.¹ The study protocol required an follow-up imaging within 36 h after recanalization.²⁰ In one center (which enrolled 164 patients), 66 patients underwent additional DECT and MRI scans between 24 and 36 h after thrombectomy and were included in this study. All study participants consented to the BP-TARGET protocol and provided informed consent to be examined with DECT and MRI for study purposes.

The number of cases in the present study (n = 66) was sufficient to ensure relatively small confidence intervals according to tables provided by Donner et al.²¹

CT, DECT, and MRI

DECT scans were performed with the CT750HD (General Electric Health Company, Milwaukee, WI) using fast kVp-switching technique with a routine protocol that resulted in a radiation dose of 47.81 mGy (CT dose index volume).

The scans were automatically postprocessed and transferred to the picture archive communication system. This includes conventional CT images at 80 keV, together with Iodine overlay maps (ION) and virtual non contrast (VNC) images, which were calculated using a previously reported algorithm.¹⁷ Conventional CT images obtained with DECT technology are considered to have an equal or improved image quality compared to single energy CT.²²

MRI was performed at 3T (Ingenia Elition; Philips Healthcare; Best; Netherlands) including diffusion-weighted-imaging (DWI), T2 Fluid-Attenuation-Inversion-Recovery (FLAIR), susceptibility-weighted-imaging (SWI), T2 Fast Field Echo (FFE) and T1 Turbo spin echo (TSE). Acquisition parameters are detailed in Supplemental Table 1.

Raters

All raters were clinicians who actively manage patients with AIS and who routinely analyze post-thrombectomy DECT scans, CT scans and MRI scans for the assessment of HT using ECASS classification. Raters were asked to specify their training background (interventional or diagnostic neuroradiology, neurology) and the number of years treating strokes. The seven readers were from the same institution, including four neuroradiologists (2 interventional (with 11 and 4 years of experience), 2 diagnostic neuroradiologists (with 10 and 2 years of experience) and three stroke neurologists (with 12, 4, and 2 years of experience respectively).

Image analysis

All images (CT, DECT and MRI) were first anonymized and uploaded in the Picture Archiving and Communication System (PACS) in a random order. For each of the 66 patients the following details were provided: patients’ gender, age, the initial and 24-h National Institute of Health Stroke Scale (NIHSS), the previous use or not of intravenous tissue plasminogen activator.

First part – consensus reading of CT, DECT, and MRI

Two interventional neuroradiologists, each with over 10 years of experience in neuroradiology, conducted a blind, and independent review of all images in order to determine the presence or absence of HT for each patient and each modality (CT, DECT. and MRI). In case of positive findings, the images were graded based on the ECASS classification¹⁰ (Supplemental Table 2). Any discrepancies were resolved through consensus in a subsequent joint reading session involving both reviewers.

Second Part – Agreement study

A training video was used for basic instructions. The video reviewed the ECASS classification¹⁰ and provided examples as well as common errors in interpretation. Subsequently, the raters completed eight practice cases. During the readings, raters were provided with a one-page summary table of ECASS classification (Supplemental Table 2).

Regarding the reading session, CT scans were first shown. The same 66 cases were shown again with the same information but in different order with DECT and then MRI. In total, raters evaluated 198 cases from the same pool of 66 patients, presented in triplicate. Importantly, raters were unaware that these 198 cases corresponded to the same 66 patients viewed in three different instances. For each patient and each imaging modality (CT, DECT, and MRI), raters were asked to answer the following questions: – Is there an HT? (yes/no) – What is the type of HT according to the ECASS classification? (0, HI1, HI2, PH1, PH2) – Having reviewed this exam, would you initiate antiplatelet therapy for this patient? (yes/no). Raters re-evaluated the cases at least 1 month after completing their initial readings. To minimize recall bias, the case order was randomly permuted, and raters were blinded to the results of their initial assessments.

Statistical analysis

Continuous values were reported as medians with interquartile ranges and categorical values as frequencies. First, the proportion of HT (presence/absence, grade according to the ECASS classification) among each imaging modality (CT, DECT, and MRI) were compared using a one-way ANOVA.

Interrater agreement for the presence of HT, ECASS classification and antiplatelet therapy decision were assessed using Fleiss’ kappa statistics,²³ while intrarater agreement was assessed with Cohen’s unweighted kappa statistics.²⁴ We pre-specified analyses of the interrater and intrarater reliability for the trichotomized classification 0 versus HI1/HI2 versus PH1/PH2, and for the dichotomized classification PH1/PH2 versus other classes. Kappa were interpreted according to Landis and Koch.²⁵ Kappa values range from −1 (perfect disagreement) to +1 (perfect agreement), 0 indicating no agreement among the raters other than what would be expected by chance alone. The following thresholds were used for Kappa interpretation: 0 < Kappa ⩽ 0.2 defines slight agreement, 0.2 < K ⩽ 0.4 is fair agreement, 0.4 < K ⩽ 0.6 is moderate agreement, 0.6 < K ⩽ 0.8 is substantial agreement, and 0.8 < K ⩽ 1 is excellent agreement.

Analyses were performed with 95% bias-corrected confidence intervals obtained by 1000 bootstrap resampling. All analyses were conducted using the R software version 3.3.2 2 (R Foundation for Statistical Computing, Vienna, Austria) at a significance level of 5%.

Results

Patient demographic data

From June 2017 to September 2019, a total of 324 patients were randomized in the BP TARGET trial. Among them, 66 patients performed CT, DECT and MRI scan 24–36 h after reperfusion and were included in this study (Supplemental Figure 1). Patients and treatment characteristics and main outcomes are reported in Supplemental Table 3 for the overall study sample.

The median age was 72 years (range 34–95). The median time from reperfusion to CT and DECT imaging was 26 h (11–47 h; interquartile range 24–29 h). The median time from reperfusion to MRI imaging was 27 h (range 11–46 h; interquartile range 25–29 h). The median time between CT/DECT and MRI imaging was 28 min (range 2–1164 min; interquartile range 16–47 min).

Detection and classification of HT with CT, DECT, and MRI – Consensus reading

According to the consensus reading, there were significant differences in the rates of detected HT (of any type) based on the imaging modality (p < 0.001). A hemorrhagic transformation (any types) was detected in 27/66 cases (40.9%) with DECT, 34/66 cases (51.5%) with CT and 51/66 cases (77.3%) with MRI (Table 1). Significant discrepancies were also observed specifically for HI-type hemorrhagic transformations (p = 0002): 34/66 (56.1%) with MRI versus 19/66 (28.8%) with CT and 16/66 (24.2%) with DECT. There was no statistically significant difference observed for parenchymal hematomas (PH; p = 0.44).

Table 1.

Proportion of patients with HT (any types and according to each ECASS classification subtypes) among each imaging modality according to the consensus reading.

Type of hemorrhage	CT	DECT	MRI	p Value
No HT	32/66 (48.5%)	39/66 (59.1%)	15/66 (22.7%)	<0.001
HT (any types)	34/66 (51.5%)	27/66 (40.9%)	51/66 (77.3%)	<0.001
Hemorrhagic transformation (HI)	19/66 (28.8%)	16/66 (24.2%)	34/66 (51.5%)	0.002
HI-1	11/66 (16.7%)	11/66 (16.7%)	18/66 (27.3%)
HI-2	8/66 (12.1%)	5/66 (7.6%)	16/66 (24.2%)
Parenchymal hematoma (PH)	15/66 (22.7%)	11/66 (16.7%)	17/66 (30.4%)	0.44
PH-1	9/66 (13.6%)	6/66 (9.1%)	5/66 (7.6%)
PH-2	6/66 (9.1%)	5/66 (7.6%)	12/66 (18.2%)

Date are number (% of total sample).

HT grading according to the consensus reading for each case and each imaging modality is graphically displayed in Figure 1. In 21/66 cases (31.8%), the three imaging techniques agreed on the HT grade (no HT in 13 cases, HI-1 in two, HI-2 in 1 and PH-2 in 5). In 33/66 cases (50.0%), CT and DECT (but not MRI) agreed on the HT grade. In all these cases, an upward shift to more severe bleeding categories was observed with the use of MRI. In 19 of 33 patients, MR imaging converted a negative CT/DECT score for HT into a positive one (HI-1 in 13, HI-2 in 5 and PH-1 in 1 patient respectively). In the remaining patients, MRI converted HI-1 grade on CT/DECT into a HI-2 grade (6 patients), HI-2 grade into a PH-1 grade (2 patient) or PH-2 grade (1 patients) and PH-1 grade into a PH-2 grade (5 patients). In the remaining 12 cases (20%), there were discrepancies in the HT grade between CT and DECT. In 7 out of these 12 cases, DECT changed a positive CT diagnosis for HT to a negative one, and in the remaining 5 cases, it lowered the HT grade determined by CT. Illustrative case of inconsistencies in HT diagnosis across the three imaging modalities is presented in Figure 2 and Supplemental Figure 2.

Figure 1.

Hemorrhagic transformation grading using the ECASS classification and according to the consensus reading for each case and each imaging modality.

Figure 2.

Illustrative case of inconsistencies in HT diagnosis across the three imaging modalities according to the consensus reading. NCCT scan diagnosed a PH-1 (panel a and b), while a HI-1 was detected on the MRI scan (panel E and F), and DECT (virtual non contrast scan) showed no signs of HT (panel c and d) (case #65 in the Figure 1).

Reliability of the detection of HT and decisions to start antiplatelet therapy among seven raters

The reading session of the seven raters gave similar results compared to the consensus reading with increased HT diagnosis and grading with MRI than DECT and CT. The mean number (min-max) of HT was 45.6% (33.9%–69.7%) on DECT, 56.5% (50%–72.7%) on CT and 75.1% (69.7%–78.8%) on MRI. Decisions to initiate antiplatelets were more frequent with DECT than MRI (p = 0.039) but not with CT (p = 0.29). The mean (min-max) number of decisions to initiate antiplatelets based on the imaging findings was 68.2% (45.4%–81.8%) on DECT, 56.7% (40.9%–68.2%) on CT, and 50.2% (22.7%–65.1%) on MRI (Supplemental Figure 3). Detailed descriptive results can be found in Supplemental Table 4.

Interrater agreement

The interrater agreement for the detection of HT (of any type) was moderate on DECT (K = 0.503 [0.448–0.558]), and substantial on CT (K = 0.724 [0.666–0.782]) and MRI (K = 0.727 [0.612–0.842]). There were 46/66 cases (69.7%) with at least 6/7 raters agreeing on the presence/absence of HT on DECT, 56/66 cases (84.8%) on CT, and 59/66 (89.4%) on MRI (Figure 3, detailed results in Table 2 and Supplemental Table 5).

Figure 3.

Interrater agreement. (a) Graphical display of the interrater agreement for the diagnosis of HT, various classifications of HT, and decisions to start antiplatelet therapy on DECT, CT and MRI (Fleiss’ Kappa). For each measurement, a Kappa value > 0.6 is considered to be substantial agreement. (b) Proportion of cases with agreement among at least 6/7 raters for each of the aforementioned categories and for each imaging modality.

Table 2.

Interrater agreement (Kappa) and proportion of cases with agreement among ⩾6/7 raters for the presence of HT, various HT classifications and the decision to start antiplatelet therapy on DECT, CT, and MRI.

Imaging modality	Interrater agreement	Proportion of cases with agreement among ⩾6/7 raters
Hemorrhagic Transformation (any type)
DECT	0.503 [0.448–0.558]	46/66 (69.7%)
CT	0.724 [0.666–0.782]	56/66 (84.8%)
MRI	0.727 [0.612–0.842]	59/66 (89.4%)
ECASS Classification (0 vs HI1 vs HI2 vs PH1 vs PH2)
DECT	0.379 [0.312–0.447]	31/66 (47.0%)
CT	0.493 [0.447–0.540]	37/66 (56.1%)
MRI	0.445 [0.416–0.473]	25/66 (37.9%)
Trichotomized ECASS Classification (0 vs HI1/HI2 vs PH1/PH2)
DECT	0.394 [0.322–0.465]	34/66 (51.5%)
CT	0.593 [0.545–0.640]	46/66 (69.7%)
MRI	0.590 [0.550–0.630]	41/66 (62.1%)
Dichotomized ECASS Classification (PH1/PH2 vs others)
DECT	0.625 [0.455–0.794]	54/66 (81.8%)
CT	0.767 [0.660–0.874]	58/66 (87.9%)
MRI	0.636 [0.566–0.707]	48/66 (72.7%)
Decision to start Antiplatelet therapy
DECT	0.511 [0.423–0.600]	46/66 (69.7%)
CT	0.636 [0.577–0.694]	50/66 (75.7%)
MRI	0.555 [0.503–0.608]	24/66 (36.4%)

Substantial values (kappa > 0.6) are bolded.

The interrater agreement for the global ECASS classification (0 vs HI1 vs HI2 vs PH1 vs PH2) was fair for DECT (K = 0.379 [0.312–0.447]) and moderate for CT (K = 0.493 [0.447–0.540]) and MRI (K = 0.445 [0.416–0.473]). Simplification of the classification into a trichotomized system (0 vs HI1/HI2 vs PH1/PH2) improved the interrater agreement on all imaging modalities but it nevertheless remained fair on DECT (K = 0.394 [0.322–0.465]) and moderate on CT (K = 0.593 [0.545–0.640]) and MRI (K = 0.590 [0.550–0.630]). After dichotomizing the classification into PH1/PH2 versus all other classes, the interrater agreement improved to a substantial level on DECT (K = 0.625 [0.455–0.794]), CT (K = 0.767 [0.660–0.874]) and MRI (K = 0.636 [0.566–0.707]). The interrater agreement for the decision to start antiplatelet therapy was substantial only on CT (K = 0.636 [0.577–0.694]) and moderate on DECT and MRI. Similar results were observed when raters were categorized based on their experience and specialty (Supplemental Table 6). None of the measurements performed for interrater agreement reached the level of “excellent agreement” (K > 0.8).

Intrarater agreement

The mean intrarater agreement for the diagnosis of HT (of any type) was moderate on DECT (K = 0.544), substantial on CT (K = 0.746), and excellent on MRI (K = 0.811). A total of 3/7 raters (42.9%) had a least a substantial intrarater agreement on DECT, 6/7 raters (85.7%) on CT, and all raters (7/7) on MRI. Raters changed their judgment on the presence/absence of a HT between readings in 20.8% of cases on DECT, 12.1% on CT, and 7.4% on MRI (Figure 4, detailed results in Supplemental Table 7).

Figure 4.

Intrarater agreement. (a) Graphical display of the mean intrarater agreement (bars) and individual intrarater agreement values (red diamonds) for the diagnosis of HT, various classifications of HT, and decisions to start antiplatelet therapy on DECT, CT and MRI (Kappa). For each measurement, a Kappa value > 0.6 is considered to be substantial agreement. (b) Proportion of cases with a change of judgments between readings for each of the aforementioned categories and for each imaging modality.

The mean intrarater agreement for the global ECASS classification (0 vs HI1 vs HI2 vs PH1 vs PH2) was fair on DECT (with none of the raters having at least a substantial intrarater agreement), moderate on CT, and substantial on MRI. Simplifying the classification into a trichotomized system (0 vs HI1/HI2 vs PH1/PH2) improved the intrarater agreement to a moderate level on DECT and a substantial level on CT, while it remained substantial on MRI. Dichotomizing the classification into PH1/PH2 versus all other classes improved the intrarater agreement to a substantial level on all imaging modalities. The mean intrarater agreement for the decision to start antiplatelet therapy was substantial on all imaging modalities (3/7 raters on DECT, 6/7 raters on CT, and 4/7 raters on MRI).

Discussion

This study suggests that there are discrepancies between CT, DECT and MRI for the assessment of HT after thrombectomy. As previously reported,^11,12,18 MRI was associated with an increased rate of HT diagnosis and a trend to categorize bleeding as more severe than with other modalities. DECT on the other hand was found to yield the lowest rate of HT mostly by reclassifying hemorrhagic transformations diagnosed by CT and MRI as contrast staining only. This variability in the diagnosis and classification of HT lead to wide discrepancies in the decisions to initiate antiplatelet therapy. Notably, an inverse correlation was observed between the rates of HT diagnosis and the decisions to initiate antiplatelet therapy. For instance, the use of MRI led to a 65% increase in HT diagnoses but also a 61% decrease in the initiation of antiplatelets compared with DECT.

The imaging modalities also significantly affected the agreement among raters. The interrater agreement for the assessment of the presence/absence of HT (of any type) was only substantial with CT and MRI and remained moderate with DECT. Perfect agreement among raters was obtained in almost 70% of cases with MRI and CT, while it was obtained in only 40% of cases with DECT. Levels of interrater agreement for the detection of HT (of any type) also varied greatly between specialties for each of the 3 imaging modalities (Supplemental Table 6).

Our study also suggests that the ECASS classification system, which consists of 5 categories for the diagnosis of HT, may be too complex, as suggested in a previous study.¹³ Regardless of the imaging used, interrater agreement was well below the “substantial” level, even between experts of the same specialty, with the same experience. Perfect agreement was obtained in only a quarter of cases. The ECASS classification was historically created (and refined multiple times) to allow for a detailed and nuanced assessment of the type of hemorrhagic transformation after a stroke (mostly in the setup of patients treated with thrombolytics) in order to correlate the type of HT with patients’ prognosis. It has since then been widely adopted for thrombectomy trials as well to serve as a safety endpoint. However recent studies have suggested that it is mostly the parenchymal-hematoma HT types that lead to patient worsening. The dichotomization of ECASS into PH1/PH2 versus other therefore seems to be clinically relevant.² In our study, this dichotomization (PH1/PH2 vs other) increased the agreement in line with previous work.¹³

Regarding the decision to initiate antiplatelets, substantial agreement was achieved only when using CT. There could be several explanations for the inferior reliability of DECT and MRI. For DECT, this could be explained by the lower reliability of DECT for the diagnosis of HT. The frequent disagreements among raters for the detection of HT probably led to further disagreements about the initiation of antiplatelet therapy. For MRI, the phenomenon could be explained by the higher prevalence of HT. Raters tend to agree on the initiation of antiplatelet therapy in the absence of HT, but they might disagree more frequently about the amount of HT that would preclude the initiation of antiplatelets. The decision to initiate antiplatelets is inherently related to the interpretation of brain imaging, and is also more clinically meaningful for the patient than the simple detection and classification of HT. Our findings suggest that a more clinically-oriented classification of hemorrhagic transformation might be helpful in the future for patient care.

One key finding in our study is the limited reliability of DECT for diagnosing and categorizing HT, even though the raters had been utilizing DECT for an extended period in our institution. This contradicts previous studies that claimed “DECT may be particularly helpful in patients who have recently undergone intra-arterial stroke therapy.”²⁶ Review of literature revealed that DECT reliability has never been thoroughly assessed.^16,26–28 Only one study has compared the reliability of DECT and CT. This study included only 2 evaluators and did not assess intrarater agreement.¹⁷ Previous studies suggested better accuracy of DECT than CT,²⁹ but this was not tested in our study. Previous studies used follow-up CT as the gold standard, assuming that if there is a complete washout of hyper-attenuation on follow-up CT only if it is due to contrast,^16,26–28 but this assumption has not been proven. Our study highlights the importance of further research, to investigate the value of DECT in this context.

Our study is different from previous publications. While previous studies have compared either CT to DECT^{16,17,26–28} or CT to MRI,^11,12,18 this is the first study to directly compare all three imaging modalities performed within a short time frame. Unlike earlier research, which relied on consensus readings from two or more raters,^{11,16–18,26–28} our study specifically examined the interrater agreement for each imaging modality. This was achieved by involving raters from diverse backgrounds and asking them to independently determine the HT grade for each case. All raters agreed to participate in the intrarater agreement study, which is an important component of reliability studies that is rarely explored in imaging research.³⁰ None of the prior studies have investigated how HT grading impacts decisions on antiplatelet therapy initiation. Finally, the patients included in this study also differed significantly from those in previous studies. This study included patients treated with thrombectomy and successful reperfusion for acute stroke with anterior large vessel occlusion whereas previous studies have mostly involved patients treated with intravenous thrombolysis.^{11,12,16,22,26–28}

Before discussing the implications of our findings for clinical practice, it is important to consider the limitations of our study. The reading conditions in our study were not the same as those in day-to-day clinical practice, and answering a questionnaire is not the same as caring for patients in real life. Additionally, in clinical practice, decisions are often made by a team of specialists, whereas in our study, specialists were interviewed separately. No specific guidance or protocols were provided to the raters on when they should or should not recommend initiating antiplatelet therapy. We did not evaluate the reliability of the Heidelberg bleeding classification³¹ which is another classification commonly used for the diagnosis and classification of HT in recent trial.^1,32 It is plausible that the reliability of the Heidelberg classification might be lower as it includes more categories than ECASS classification (such as subarachnoid, intra-ventricular and subdural hematoma). We did not evaluate interrater agreement according to the presence or not of symptomatic intracranial hemorrhage due to the lack of a clear, optimal definition. Depending on the chosen definition, the interrater agreement results could vary considerably. A time gap existed between CT, DECT and MRI scans. Therefore, the possibility remains that, within this interval, new HT emerged or deteriorated. No accuracy study has been conducted in this research due to the inability to establish a clear reference standard, either by simultaneously evaluating all three imaging modalities (as there is no means to determine which modality provides the “true diagnosis”) or by relying on follow-up imaging (owing to the dynamic nature of HT). Our study involved patients and physicians from a single system of care, and the confidence intervals of the kappa values displayed overlap. Our results should therefore be extrapolated with caution and we encourage our colleagues to replicate this experiment.

Our study is the only one that compares the three diagnostic modalities for HT diagnosis and classification. It’s essential for clinicians to consider the imaging method used, as it can significantly impact the rates of HT diagnosis and antiplatelet regimen prescription. Based on our findings, CT seems to have the highest interrater and intrarater agreement for both HT assessment and antiplatelet therapy prescription. Using DECT instead might lower the interrater and intrarater agreement in addition to lead to lower rates of HT diagnoses and higher rates of antiplatelet prescription. On the contrary, MRI might lead to an increase in the rate of HT diagnosis and a decrease of antiplatelet initiation. Our results have implications for future research, as many studies use HT rates and subtypes as secondary^4–9 or even primary outcomes.¹

Conclusion

The detection and classification of HT was influenced by the imaging modality in our study (DECT, CT, or MRI). The overall ECASS classification showed a limited reliability which improved to a substantial level after dichotomization into Parenchymal hematoma (type 1 or 2) versus other classes, with all imaging modalities. Substantial agreement for both the assessment of hemorrhagic transformation and the decision to start antiplatelets was achieved only with CT.

Supplemental Material

sj-docx-1-eso-10.1177_23969873251331484 – Supplemental material for Reliability of CT, DECT, and MRI for the diagnosis of hemorrhagic transformation after thrombectomy

Supplemental material, sj-docx-1-eso-10.1177_23969873251331484 for Reliability of CT, DECT, and MRI for the diagnosis of hemorrhagic transformation after thrombectomy by William Boisseau, Augustin Lecler, Stanislas Smajda, Pierre Seners, Quentin Holay, Lucy Bernardaud, Oriana Tarabay, Julien savatovsky, Michel Piotin, Mikael Mazighi and Robert Fahed in European Stroke Journal

Footnotes

Acknowledgements

There are no acknowledgments for this study.

Author’s note

Guidelines. This study was conducting using the GRRAS guidelines.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Informed consent

The need for informed consent was waived.

Ethical approval

This retrospective study was approved by the research ethics board at our institution.

Guarantor

Contributorship

WB and RF: Conceptualization, Writing, original draft and data curation. WB, AL, SS, PS, QH, LB, OT, RF: Data curation. MP, JS, MM: Supervision; writing – review/editing. RF: formal analysis. All authors reviewed/edited and approved the final manuscript.

ORCID iDs

William Boisseau

Robert Fahed

Data availability

Anonymized patient data are available for use in independent scientific research upon reasonable request (boisseau.william@gmail.com). Data will be provided following review and approval of a research proposal (including a statistical analysis plan) and completion of a data sharing agreement.

Supplemental material

Supplemental material for this article is available online.

References

Mazighi

Richard

Lapergue

, et al. Safety and efficacy of intensive blood pressure lowering after successful endovascular therapy in acute ischaemic stroke (BP-TARGET): a multicentre, open-label, randomised controlled trial. Lancet Neurol 2021; 20: 265–274.

Ospel

Qiu

Menon

, et al. Radiologic patterns of intracranial hemorrhage and clinical outcome after endovascular treatment in acute ischemic stroke: results from the ESCAPE-NA1 Trial. Radiology 2021; 300: 402–409.

Powers

Rabinstein

Ackerson

, et al. Guidelines for the early management of patients with acute ischemic stroke: 2019 update to the 2018 guidelines for the early management of acute ischemic stroke: a guideline for healthcare professionals from the American Heart Association/American Stroke Association. Stroke 2019; 50: e370.

Berkhemer

OA.

A randomized trial of intraarterial treatment for acute ischemic stroke. N Engl J Med 2015; 372: 394–420.

Bracard

Ducrocq

Mas

, et al. Mechanical thrombectomy after intravenous alteplase versus alteplase alone after stroke (THRACE): a randomised controlled trial. Lancet Neurol 2016; 15: 1138–1147.

Campbell

BCV

, et al. Endovascular therapy for ischemic stroke with perfusion-imaging selection. N Engl J Med 2015; 372: 1009–1018.

Goyal

, et al. Randomized assessment of rapid endovascular treatment of ischemic stroke. N Engl J Med 2015; 372: 1019–1030.

Jovin

Chamorro

Cobo

, et al. Thrombectomy within 8 hours after symptom onset in ischemic stroke. N Engl J Med 2015; 372: 2296–2306.

Saver

Goyal

Bonafe

, et al. Stent-retriever thrombectomy after intravenous T-PA vs. t-PA alone in stroke. N Engl J Med 2015; 372: 2285–2295.

10.

Hacke

Kaste

Fieschi

, et al. Intravenous thrombolysis with recombinant tissue plasminogen activator for acute hemispheric stroke. The European Cooperative Acute Stroke Study (ECASS). JAMA 1995; 274: 1017–1025.

11.

Arnould

M-C

Grandin

Peeters

, et al. Comparison of CT and three MR sequences for detecting and categorizing early (48 hours) hemorrhagic transformation in hyperacute ischemic stroke. AJNR Am J Neuroradiol 2004; 25: 939–944.

12.

Renou

Sibon

Tourdias

, et al. Reliability of the ECASS radiological classification of postthrombolysis brain haemorrhage: a comparison of CT and three MRI sequences. Cerebrovasc Dis 2010; 29: 597–604.

13.

Guenego

Lecler

Raymond

, et al.; s. Hemorrhagic transformation after stroke: inter- and intrarater agreement. Eur J Neurol 2019; 26: 476–482.

14.

Lun

Walker

Guenego

, et al. Is this contrast? Is this blood? An agreement study on post-thrombectomy computed tomography scans. Front Neurol 2020; 11: 593098.

15.

Guenego

Mine

Bonnet

, et al. Thrombectomy for distal medium vessel occlusion with a new generation of stentretriever (Tigertriever 13). Interv Neuroradiol 2022; 28: 444–454.

16.

Gupta

Phan

Leidecker

, et al. Evaluation of dual-energy CT for differentiating intracerebral hemorrhage from iodinated contrast material staining. Radiology 2010; 257: 205–211.

17.

Almqvist

Holmin

Mazya

MV.

Dual energy CT after stroke thrombectomy alters assessment of hemorrhagic complications. Neurology 2019; 93: e1068–e1075.

18.

Hsia

Latour

Somani

, et al. Comparison of hemorrhage on CT versus MRI after thrombectomy: the HECATE study. Stroke: Vascular and Interventional Neurology 2024; 4: e001441.

19.

Kottner

Audigé

Brorson

, et al. Guidelines for Reporting Reliability and Agreement Studies (GRRAS) were proposed. J Clin Epidemiol 2011; 64: 96–106.

20.

Mazighi

Labreuche

Richard

, et al. Blood Pressure Target in acute stroke to reduce HemorrhaGe after endovascular therapy: the Randomized BP TARGET Study Protocol. Front Neurol 2020; 11: 480.

21.

Donner

Rotondi

MA A

. Sample size requirements for interval estimation of the kappa statistic for Interobserver Agreement Studies with a binary outcome and multiple raters. Int J Biostat 2010; 6: Article 31.

22.

Pomerantz

Kamalian

Zhang

, et al. Virtual monochromatic reconstruction of dual-energy unenhanced head CT at 65–75 keV maximizes image quality compared with conventional polychromatic CT. Radiology 2013; 266: 318–325.

23.

Fleiss

JL.

Measuring nominal scale agreement among many raters. Psychol Bull 1971; 76: 378–382.

24.

Cohen

A coefficient of agreement for nominal scales. Educ Psychol Meas 1960; 20: 37–46.

25.

Landis

Koch

GG.

The measurement of observer agreement for categorical data. Biometrics 1977; 33: 159–174.

26.

Phan

Yoo

Hirsch

, et al. Differentiation of hemorrhage from iodinated contrast in different intracranial compartments using dual-energy head CT. AJNR Am J Neuroradiol 2012; 33: 1088–1094.

27.

Tijssen

Hofman

Stadler

, et al. The role of dual energy CT in differentiating between brain haemorrhage and contrast medium after mechanical revascularisation in acute ischaemic stroke. Eur Radiol 2014; 24: 834–840.

28.

Zaouak

Sadeghi

Sarbu

, et al. Differentiation between cerebral hemorrhage and contrast extravasation using dual energy computed tomography after intra-arterial neuro interventional procedures. J Belg Soc Radiol 2020; 104: 70.

29.

Chen

Zhang

Quan

, et al. Diagnostic accuracy of dual-energy computed tomography to differentiate intracerebral hemorrhage from contrast extravasation after endovascular thrombectomy for acute ischemic stroke: systematic review and meta-analysis. Eur Radiol 2022; 32: 432–441.

30.

Farzin

Gentric

Pham

, et al. Agreement studies in radiology research. Diagn Interven Imaging 2017; 98: 227–233.

31.

von Kummer

Broderick

Campbell

, et al. The Heidelberg Bleeding Classification: classification of bleeding events after ischemic stroke and reperfusion therapy. Stroke 2015; 46: 2981–2986.

32.

Van Der Steen

, et al. Safety and efficacy of aspirin, unfractionated heparin, both, or neither during endovascular stroke treatment (MR CLEAN-MED): an open-label, multicentre, randomised controlled trial. Lancet 2022; 399: 1059–1069.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

5.22 MB