INTERVENTIONAL NEURORADIOLOGY

Clinical significance of post-interventional cerebralhyperdensities after endovascular mechanical thrombectomyin acute ischaemic stroke

Omid Nikoubashman & Arno Reich & Mirco Gindullis & Katharina Frohnhofen &

Rastislav Pjontek & Marc-Alexander Brockmann & Jörg B. Schulz &

Martin Wiesmann

Received: 2 July 2013 /Accepted: 4 November 2013 /Published online: 5 December 2013# Springer-Verlag Berlin Heidelberg 2013

AbstractIntroduction This study aims to investigate the clinical signif-icance of post-interventional cerebral hyperdensities (PCHD)after endovascular mechanical thrombectomy in acute ischae-mic stroke.Methods Data of 102 consecutive patients who received post-interventional CT scans within 4.5 h after mechanicalthrombectomy were analysed retrospectively.Results Sixty-two of 102 patients (60.8 %) had PCHD ontheir post-interventional CT scans. The most common site ofPCHD was the basal ganglia. PCHD were persisting in 13 of62 patients (21.0 %), and transient in the remaining 49 patients(79.0 %) within 24 h. Four patients with PCHD and fourpatients without PCHD suffered from parenchymal haemor-rhage. Neither ASA nor Clopidogrel, Tirofiban or rtPA wererisk factors for PCHD. Final infarction size was congruentwith or bigger than areas of PCHD in 93.3 % of cases in ourseries.Conclusion PCHD was not a risk factor for parenchymalhaemorrhage in our series. The occurrence of PCHD wasstrongly related to the prior presence of infarction. PCHDwas also a strong predictor for final infarction size.

Keywords Stroke . Post-interventional CT . Hyperdensity .

Haemorrhage .Mechanical recanalisation

Introduction

Post-interventional cerebral hyperdensities (PCHD), alsoknown as post-interventional contrast extravasation or con-trast staining, are common after neurointerventional stroketreatment. Nevertheless, little is known about their natureand clinical implications. Only a few studies deal with PCHDafter intra-arterial lysis [1–6] or mechanical recanalisation[7–9] (Table 1).

We therefore investigated the prevalence of PCHD in alarge cohort of patients who underwent endovascular stroketreatment by mechanical recanalisation, analysing also epide-miologic factors including pharmacological treatment. PCHDmay be caused by oozing of contrast agent into the parenchy-ma or by actual haemorrhage. We sought to differentiate thisby analysing follow-up CT studies. We analysed whetherpost-interventional hyperdensities increase the risk for grosshaemorrhage, and whether PCHD indicate cerebral infarction.To the best of our knowledge, the correlation between PCHDand infarction has not been addressed in the literature.

Patients and methods

Patients

All patients admitted to our hospital for an acute ischaemicstroke in the anterior circulation who underwent endovascularstroke treatment via mechanical recanalisation, and received apost-interventional CT examination within 4.5 h after treat-ment, were included in this study.

Overall, 104 consecutive patients fulfilled these criteriabetween January 2010 and October 2012. Two patients hadsigns of intracerebral haemorrhage during the intervention andwere excluded. Thus, we included 102 patients in this study.

Electronic supplementary material The online version of this article(doi:10.1007/s00234-013-1303-1) contains supplementary material,which is available to authorized users.

O. Nikoubashman :K. Frohnhofen : R. Pjontek :M.<A. Brockmann :M. Wiesmann (*)Klinik für Diagnostische und Interventionelle Neuroradiologie,Uniklinik Aachen, Pauwelsstr. 30, 52074 Aachen, Germanye-mail: mwiesmann@ukaachen.de

O. Nikoubashman :A. Reich :M. Gindullis : J. B. SchulzKlinik für Neurologie, Uniklinik Aachen, Aachen, Germany

Neuroradiology (2014) 56:41–50DOI 10.1007/s00234-013-1303-1

Stroke management

Details on the stroke management at our institution, and thetreatment regiments which the patients received, are availableOnline as supplementary material.

Methods

All data were evaluated retrospectively. Neurological statuswas assessed using the National Institute of Health StrokeScale (NIHSS) and the Modified Rankin Scale (mRS).

CTscans were performed on a 16-slice CTscanner (SiemensSOMATOM®Definition AS; Straton MX P, Erlangen, Germa-ny). Radiological examinations were analysed in randomisedorder on standard PACS workstations by two experiencedneuroradiologists who were blinded to all clinical data. After-wards, a consensus reading was performed to obtain a referencestandard for statistical analyses.

Areas of infarction and hyperdense areas were assessedusing a standardised binary ASPECTS evaluation form thatdivides the brain hemispheres into 12 distinct areas [10].Minimum, maximum and mean Hounsfield Units (HU) ofhyperdense areas in post-interventional and follow-up CTexaminations were calculated based on visually defined Regionsof interest (ROI). Statistical analyses were focused onmaximumHU (HUmax).

Pre- and post-interventional vessel status was assessed inCT angiography and digital substraction angiography usingthe thrombolysis in cerebral infarction (TICI) scale with anadditional category “2c” for subtotal recanalisation (>2/3 ofthe vessel territory) [11].

Statistical analyses

We applied Chi2 tests, Fisher's exact tests and t tests whereadequate. Logistic regression analyses were performed forcomplex correlations. Results with a p value under 0.05 wereconsidered significant. All statistical analyses were performedusing SPSS® 20 software (IBM®, San Jose, California, USA).

Definitions

Post-interventional cerebral hyperdensities Non-solid visual-ly distinctive parenchymal hyperdense areas diagnosed within4.5 h after recanalisation, with a diameter of at least 0.1 cm2,and without a space occupying effect, and with an increaseddensity of at least 5 HU (HUmax) compared to the unaffectedcontralateral.

Transient PCHD PCHD that became iso- or hypodense to theunaffected contralateral hemisphere within 24 h.

Persisting PCHD PCHD that remained hyperdense during atleast 24 h.

Haemorrhage (according to the ECASS 1 and 2 definition):

– PH: Parenchymal haemorrhage (coagulum) with masseffect

– PH1: Less that 30% of the infarcted area with mild space-occupying effect

– PH2: Greater than 30 % of the infarcted area with signif-icant space-occupying effect

– HI: Haemorrhagic infarction: petechial infarction withoutspace-occupying effect; non-solid and without space-occupying effect, respecting borders of grey and whitematter [12].

– HI1: Small petechiae– HI2: More confluent petechiae

Early HI occurring within 36 h after recanalisation wasdistinguished from delayed HI occurring 36 h afterrecanalisation [13].

Infarction Infarction diagnosis was based on MRI via diffu-sion weighted imaging including ADC maps in combinationwith T2-FLAIR sequences whenever possible. Infarction di-agnosis via CT was complicated when final infarction size

Table 1 Overview of articles dealing with post-interventional hyperdensities

Article Year Method No. PCHD Prevalence Haemorrhage risk

Rouchaud et al. [9] 2013 Mechanical recanalisation 63 38 60 % Increased risk

Parrilla et al. [8] 2012 Mechanical recanalisation 48 15 31 % No increased risk

Kim et al. [7] 2012 i.a. thrombolysis and thrombus fragmentation 68 38 56 % Increased risk

Jang et al. [1] 2006 i.a. thrombolysis 94 31 33 % Increased risk

Yoon et al. [6] 2004 i.a. thrombolysis 62 21 34 % Increased risk

Nakano et al. [3] 2001 i.a. thrombolysis 77 37 48 % Increased risk

Yokogami et al. [5] 1996 i.a. thrombolysis 35 18 51 % Increased risk

Wildenhain et al. [4] 1994 i.a. thrombolysis 10 6 60 % Increased risk

Komiyama et al. [2] 1993 i.a. thrombolysis 2 2 Case reports

No. number of patients, PCHD post-interventional cerebral hyperdensities

42 Neuroradiology (2014) 56:41–50

could not be assessed in the presence of persistinghyperdensities. We supposed that a complete infarct demarca-tion can be expected in the following settings: (1) patientswithout PCHD—CT scan performed at least 12 h afterrecanalisation; (2) patients with PCHD—CT scan performedat least 12 h after recanalisation if all former hyperdense areasare hypodense. Since it is unknown yet to which extent PCHDcorrespond to parenchymal infarction, final infarction sizewas not assessed if PCHD remained isodense to normalparenchyma in follow-up CT scans performed within 24 hafter recanalisation. Final infarct size was also not assessed,when PCHD remained hyperdense in the last follow-up CT.

Results

Patients

One hundred two patients were analysed (49 women, 53 men,n. s.). Mean age of patients was 70.3 years (SD 17.1 years).Men were significantly younger than women (67.8 versus74.3 years, p =0.036).

Interventional treatment

Details on interventional treatments are presented in Table 2.

Clinical findings

Initial median mRS and NIHSS scores of all patients were 5.0and 14.5, respectively. The mRS score at discharge and themRS score at 3 months were 4.5 and 4.0, respectively.

There was no significant difference between the mRSscores of patients with and without PCHD (administration:p =0.520, discharge: p =0.463, 3 months: p =0.382, two-tailed Student's t test).

Procedural parameters

Mean time between onset of symptoms until recanalisation ordiscontinuation of therapy was 325.2 min (median, 303.0;ranging from 120 to 813 min; SD, 134.2). Mean duration ofneuroradiological procedure was 131.0 min (median, 115.0;ranging from 30 to 351 min; SD, 74.7). Mean time betweenrecanalisation or discontinuation of therapy until post-interventional CT scan was 76.6 min (median, 70.0; rangingfrom 27 to 250 min; SD, 34.2).

An average of 2.5 recanalisation attempts per patient wereperformed (median, 2.0; ranging from 1 to 8; SD, 1.7).Recanalisation was successful after one single pass in 39 cases(38.2 %). Stenting of the internal carotid artery was necessaryin 23 cases (22.5 %). One stent was placed in the M1 segmentof the middle cerebral artery.

Radiological findings

PCHD

Sixty-two of 102 patients (60.8 %) had PCHD on their post-interventional CTscans. However, we found PCHD in only 22of 72 patients (31 %) when we considered only the latest CTperformed within 24 h after recanalisation.

All visually distinctive hyperdense areas had an increaseddensity of at least 5 HUmax compared to the unaffectedcontralateral hemisphere and were therefore included in thisstudy.

Overall, there were 744 ASPECTS areas in 62 patients.PCHD was present in 206 of 744 ASPECTS areas (27.7 %).Localisations of PCHD were the following (according toASPECTS): lentiform nucleus (in n =50 of 62 cases,80.6 %), M2 area (n =30, 48.4 %), insula (n =30, 48.4 %),caudate nucleus (n =29, 46.8 %), M5 area (n =19, 30.6 %),M1 area (n =11, 17.7 %), M4 area (n =10, 16.1 %), M3 area(n =9, 14.5 %), M6 area (n =8, 12.9 %), A area (n =6, 9.7 %),internal capsule (n =4, 6.5 %). One patient with a foetalposterior cerebral artery had PCHD in the P area.

PCHD were persisting in 13 of 62 patients (21.0 %), andtransient in the remaining 49 patients (79.0 %) within 24 h.Mean HUmax of all hyperdense areas was 83.7 HU (median,77.0; ranging from 58 to 152 HU; SD, 22.5).

Mean HUmax of areas corresponding to transienthyperdensities was 79.2 HU (median, 74; ranging from 61to 152 HU; SD, 19.8) and 87.8 HU for areas corresponding topersisting hyperdensities (median, 80; ranging from 72 to 142;SD,19.6). This difference was not statistically significant (p =0.217, Student's t test). Furthermore, mean HUmax ofhyperdense areas of patients with parenchymal haemorrhagewas not significantly different from those of patients withoutparenchymal haemorrhage (p =0.476, Student's t test). Char-acteristics of patients with and without PCHD are provided inTable 3.

Haemorrhage

Depending on availability of follow-up studies, the occurrenceof haemorrhage could be assessed in 88 of 102 patients. Thiscomprises 48 patients with PCHD and 40 patients withoutPCHD.

Parenchymal haemorrhage Four of 88 patients (4%) sufferedfrom parenchymal haemorrhage with a space occupying effect(ECASS PH2). All four haemorrhagic events occurred within1 to 4 days after recanalisation. One of these four patients hadPCHD (Fig. 1). There was symptomatic mass haemorrhage intwo of these four cases—including one patient with PCHD.Haemorrhage in this latter patient occurred 1 day afterrecanalisation, involved the hyperdense area and was lethal.

Neuroradiology (2014) 56:41–50 43

Four additional patients (4 %) had smaller asymptomaticparenchymal haemorrhage (ECASS PH1) within 4 days afterrecanalisation. Three of these four patients had PCHD (Fig. 2).PH1 haemorrhage involved hyperdense areas in all 3 patientswith PCHD. An overview of all eight patients with PH isprovided in Table 4.

Haemorrhagic infarction Early haemorrhagic infarction oc-curring within 36 h after recanalisation was present in 2 of 88patients (2 %). Overall, there was HI in 22 of 88 cases(25.0 %). Complete recanalisation (≥TICI 2b) was significant-ly associated with the occurrence of HI (p =0.047, two-sidedFisher's exact test).

Infarction

Follow-up imaging suitable for infarction diagnosis was avail-able for 84 patients. This comprises 45 patients with PCHDand 39 patients without PCHD.

Follow-up imaging of these 84 patients was performedafter a mean period of 4.2 days after recanalisation (median,2.2; ranging from 7.0 h to 28.9 days; SD, 4.9 days). Infarctsize was assessed via CT in 58 cases (69.0 %) and MRI in 26cases (31.0 %).

Predisposing factors for the occurrence of PCHD

Basic statistics indicated that female sex, amount of contrastagent, presence of infarction on intitial CTscan and number ofrecanalisation attempts may be predisposing factors for theoccurrence of PCHD (Table 3).

However, logistic regression analyses revealed that onlythe presence of infarction in the initial CT scan constituted anindependent predisposing factor for the occurrence of PCHD(p =0.016) whereas the influence of sex (p =0.205), theamount of contrast agent (p =0.287) and the number ofrecanalisation attempts (p =0.261) did not provesignificant.

Table 2 Overview of procedural parameters. Number of patients indicated if less than all

Vessels Left/right 57:45 (p =0.276, binomial test)MCA 64

Terminal ICAwith bifurcation 23

Proximal ICA into MCA 8

Proximal ICA and MCA 1

Proximal ICA 4

ACA 1

ACA and MCA 1

Initial degreeof occlusion

TICI 0 97 (95.1 %)

TICI 1 2 (2 %)

TICI 2a 2 (2 %)

TICI 2ba 1 (1 %)

Final degree ofocclusionb

TICI 0 7 (6.9 %)

TICI 1 8 (7.8 %)

TICI 2a 3 (2.9 %)

TICI 2b 11 (10.8 %)

TICI 2c 6 (5.9 %)

TICI 3 67 (65.7 %)

rtPA therapy Patients with rtPA regardless ofapplication method

74 (72.5 %)

Patients with intravenously applied rtPAbridging therapy

65 (63.72 %)

Patients with intraarterially applied rtPA 23 (22.5 %)

Both intravenous and intraarterial rtPA 14 (13.7 %)

rtPA dose Mean cumulative rtPA dose regardless ofapplication method

39.3 mg (median, 35.0; ranging from 8 to 98 mg; SD, 20.1).

Mean dose of intraarterially applied rtPA 21.78 mg (median, 15.0; ranging from 5 to 50 mg; SD, 14.5)

Mean dose of intravenously applied rtPA 37.0 mg (median, 30; ranging from 10 to 98 mg; SD, 20.5).

ICA internal carotid artery, ACA anterior cerbral artery, MCA middle cerebral artery, rtPA recombinant tissue-type plasminogen activatora Occlusion of the pericallosal artery with severe perfusion deficits in CT perfusion imagingbAll TICI scores improved after treatment except for seven patients with initial and final TICI 0 occlusions (93.1 %)

44 Neuroradiology (2014) 56:41–50

Effects of pharmacologic treatment

There was no significant correlation between PCHD, pPCHD,PH or HI and treatment with acetylsalicylic acid (ASA),Clopidogrel, recombinant tissue-type plasminogen activator(rtPA) or Tirofiban (Tables 5 and 6).

Basic statistics indicated that there seemed to be a statisticaltrend towards a correlation between Tirofiban and persistingPCHD (p =0.054, one-tailed Fischer's exact test). However,regression analysis (forward, stepwise) revealed that the in-fluence of Tirofiban on persisting PCHD was not significant(p =0.421) when statistical analyses were corrected for theactual stenting procedure as a possible time-consuming cofac-tor (p =0.012) (Nagelkerke's Pseudo R2=0.127).

Correlations between PCHD and haemorrhage

PCHD were not correlated with the occurrence of parenchy-mal haemorrhage or haemorrhagic infarction (Table 4).

Correlations between PCHD and early signs of infarctionin initial imaging

Of 1,224 ASPECTS areas (90.4 %) in the initial scans of all102 patients, 1,106 were normal while infarction was visiblein 118 initial ASPECTS areas (9.6 %).

In the 102 post-interventional scans, PCHD was found in70 of 118 initially infarcted ASPECTS areas (59.3 %) and in136 of 1,106 initially normal areas (12.3 %). Thus, sensitivityand specificity of initially infarcted areas for PCHDwere 34.0and 95.3 %, respectively. The positive predictive value for

Table 3 Overview of patientswith and without post-interven-tional cerebral hyperdensities(PCHD)

Number of patients analysed areindicated in parentheses. p valuesaccording to Pearson's Chi2 test,Fisher's exact test and Student'st test

PH parenchymal haemorrhageaccording to ECASS criteria(see above), HI haemorrhagicinfarction according to ECASScriteria (see above), GP 2b/3aGlycoprotein 2b/3a

* significant p-values

PCHD No PCHD p value

n (n) 62 (60.8 %) 40 (39.2 %) 0.037*

Female/male (102) 35:27 14:26 0.034*

Age (102) 71.31 years 70.24 years 0.737

Median NIHSS score at admission 14.5 15.0 0.428

Median mRS score at admission 5.0 5.0 0.520

Median mRS score at discharge 4.5 4.5 0.463

Median mRS score at 3 months 4.0 4.0 0.382

Symptom onset—recanalisation (69) 339.51 min 301.58 min 0.258

Intervention duration (102) 137.81 min 115.45 min 0.182

Recanalisation—post-interventional CT (102) 72.85 min 75.25 min 0.716

No. of recanalisation attempts (102) 2.8 2.0 0.019*

Infaction in initial CT (102) 39/62 14/40 0.006*

Final TICI score >2a (102) 53/62 31/40 0.420

Amount of contrast agent (102) 238.71 ml 206.25 ml 0.032*

i.v. rtPA (102) 39/62 26/40 0.830

i.a. rtPA (102) 15/62 8/40 0.621

GP 2b/3a inhibition (102) 19/62 11/40 0.734

Infarction >1/3 MCA territory (84) 15/45 13/39 1.000

PH (88) 4/48 4/40 0.536

PH 1 3/4 1/4

PH 2 1/4 3/4

Symptomatic 1/1 1/3

HI (88) 12/48 10/40 1.000

Fig. 1 CT scans of a 85-year-old patient (case 5, Table 4) showing largeareas of PCHD in the post-interventional CT scan (a), and mass haem-orrhage (ECASS PH2) in the follow-up CT scan on the same day (b).Even though this example may suggest otherwise, a large extent ofhyperdensities did not prove to correlate significantly with the occurrenceof haemorrhage (p =0.853, Student's t test)

Neuroradiology (2014) 56:41–50 45

hypodensities to predict PCHD was 59.3 %. The correlationbetween initially infarcted areas and the occurrence of PCHDwas significant (p <0.001; Chi2 test).

Correlations between PCHD and final infarction in followup imaging

Of 1,008 ASPECTS areas (65.2 %) in the follow-upscans of 84 patients in whom final infarction size couldbe assessed, 657 were normal while infarction waspresent in 351 follow-up ASPECTS areas (34.8 %).Infarction was found in 127 of 134 areas with PCHD(94.8 %) and in 224 of 874 areas without PCHD(25.6 %). Thus, sensitivity and specificity of PCHDfor final infarction were 36.2 and 98.9 %, respectively.The positive predictive value was 94.8 %. The correla-tion between areas with PCHD and the occurrence offinal infarction was significant (p <0.001; Chi2 test).

Final infarct size in 45 patients with PCHD was congruentwith areas of PCHD in 16 patients (35.6 %) (Fig. 3) and final

infarct size surpassed areas of PCHD in 26 of 45 patients(57.8 %) (Fig. 4).

Infarct size surpassed areas of PCHD in cases with a finalTICI score <2b (p =0.039, Pearson's Chi2 test). Apart fromthat, there were no statistically significant differences betweenpatients with congruent infarctions and patients in whominfarction surpassed areas of PCHD.

Final infarct size was smaller than areas of PCHD in threepatients (6.7 %) who had PCHD and no corresponding infarc-tion in a total of seven ASPECTS areas. Infarction diagnosiswas based on CT in two of these three patients andMRI in onepatient.

Discussion

Prevalence of PCHD

PCHD have been reported in both after intra-arterial throm-bolytic therapy and mechanical thrombectomy in 30 to 60 %

Fig. 2 Initial CTscan (a), post-interventional CTscan (b) and follow-upT2-FLAIR (c) and T2* (d) MRI in a 64-year-old woman. The initiallyhypodense lentiform nucleus (a : arrow) is markedly hyperdense in the

post-interventional CT scan (b : arrow). The MRI scan performed 3 dayslater reveals a small parenchymal haemorrhage within the lentiformnucleus (ECASS PH1) (c and d : arrow)

Table 4 Table showing the eight patients with parenchymal haemorrhage

PCHD pPCHD PH delay sympt HI <36 h HI rtPA rtPA dose Antiplatelet

1 – – PH 2 1 – – – i. a. 50 Yes

2 – – PH 1 3 – – HI 1 i. v. 90 Yes

3 – – PH 2 4 – – – – – –

4 – – PH 2 3 Yes n.a. – – – –

5 Yes – PH 2 1 Yes – – i. a. 25 Yes

6 Yes – PH 1 0 – – HI 1 i. a. 53 –

7 Yes – PH 1 0 – – n.a. i. v. 30 –

8 Yes – PH 1 4 – n.a. n.a. i. a. 50 Yes

rtPA dose in milligram. Antiplatelet: intensified antiplatelet treatment with Tirofiban or double antiplatelet inhibition with ASA and Clopidogrel

PCHD post-interventional cerebral hyperdensity, pPCHD persisting PCHD, PH parenchymal haemorrhage according to ECASS criteria (see above),Delay diagnosis of haemorrhage in days after recanalisation, sympt symptomatic, HI haemorrhagic infaction according to ECASS criteria (see above),i. a. intra-arterial, i. v. intravenous

46 Neuroradiology (2014) 56:41–50

of cases [1–6]. Parilla et al. observed a prevalence of 30 %which is only half as high as the prevalence of 60 % in theseries of Rouchaud et al. and the prevalence of 62 % in ourseries [8, 9]. This wide range is most likely due to methodo-logical differences. Parilla et al. analysed CTs performedwithin 24 h and found a prevalence of 30 %, whereasRouchaud et al. analysed flat-detector CT scans obtainedimmediately after the procedure and found a prevalence of60 % [8, 9]. This corresponds well with our results. We founda prevalence of 61%within 4.5 h after recanalisation, but only31 % when we considered only the latest CT within 24 h.Thus, assessing the occurrence of PCHD via flat-detector CTmay lead to a considerable prevalence, while PCHD may besystematically underdiagnosed if CT scans are not performedearly after the procedure.

Pathophysiology of PCHD

Hyperdensity of brain tissue after stroke treatment can becaused by the extravasation of either blood or contrast agents.Normally, the blood–brain barrier (BBB) is hardly permeablefor iodinated contrast agents or red blood cells (RBC) [14].Ischaemic damage and, to some extent, local toxicity of rtPAcompromise the integrity of the BBB [15–19]. Consequently,in affected areas contrast agent can extravasate into the paren-chyma. This explains the occurrence of PCHD in strokepatients if they have received iodinated contrast agents during

diagnostic work-up. Selective intra-arterial administration ofcontrast agents during neurointerventional treatment leads tohigher concentrations of contrast agent in the infarcted areaand thereby further promotes the occurrence of PCHD.

Disintegration of the BBB in infarcted brain tissue can alsolead to gradual extravasation of RBC into the parenchyma.This process is termed physiologic petechial bleeding orhaemorrhagic transformation. It usually occurs in the subacutephase (>48 h) of cerebral infarction [20]. PCHD andhaemorrhagic transformation share the same features as theyare mostly limited to grey matter, and are not space-occupying. As PCHD are observed immediately afterneurointerventional treatment, it appears unlikely that extrav-asation of blood contributes significantly to the increaseddensity of PCHD. Nevertheless, in principle, conventionalCTcannot differentiate whether the hyperdensity in suspectedPCHD is caused by contrast agent or extravasated blood [21].

Some authors distinguish between transient and persistinghyperdensities and suggest that transient PCHD correspond toextravasated contrast agent that is washed out in the course oftime, whereas persisting PCHD might correspond to blood inthe context of petechial bleeding [3, 4, 6]. However, it isconceivable that transient PCHD may also correspond tosubtle haemorrhage, when only small and rapidly reabsorbedamounts of blood pass the damaged BBB. Furthermore, it ispossible that persisting PCHD correspond to contrast agentthat is trapped in infarcted parenchyma and cannot be washedout. Our analysis and data in the literature lack systematiccomparisons of post-interventional CT scans and MRI scansthat could help distinguishing iodinated contrast agent fromblood.

In the end, it is important to note that PCHD, regardless oftheir true nature, do not resemble parenchymal haemorrhagebut physiologic haemorrhagic transformation which is associ-ated with a benign natural history [13, 22].

Predictors of PCHD

In accordance with the assumed pathomechanism of PCHD asoutline above, the occurrence of PCHDwas correlated mainlywith the presence of infarction in the initial CT scan both inour data and in the literature [5].

Even though it appears reasonable to assume that higheramounts of contrast agent, an increased duration betweenonset of symptoms and recanalisation, or the application ofrtPA increase the likelyhood of PCHD, neither our results nordata in the literature do support this [8, 9].

Prognostic value of PCHD to predict haemorrhage

The overall haemorrhage rate of 7.8 % for parenchymalhaemorrhage in our series was comparable to haemorrhagerates provided in the literature [23–29].

Table 5 Relationship between peri- and post-procedural rtPA andTirofiban treatment and post-interventional cerebral hyperdensities(PCHD), pPCHD (persisting PCHD), PH (parenchymal haemorrhage)and HI (haemorrhagic infarction) in all patients

Intra-arterial rtPA Intravenous rtPA Tirofiban

PCHD (n =102) p =0.404 p =0.500 p =0.456

pPCHD (n =88) p =0.569 p =0.585 p =0.054*

PH (n =88) p =0.539 p =0.398 p =0.197

HI(n =88) p =0.303 p =0.069 p =0.425

One-tailed Fisher exact tests

*p =0.421 (not significant when corrected for the actual stenting proce-dure as a possible time-consuming cofactor (p =0.012) (Nagelkerke'sPseudo R2 =0.127))

Table 6 Relationship between peri- and post-procedural antiplatelettherapy and persisting post-interventional cerebral hyperdensities(pPCHD) as well as parenchymal haemorrhage (PH) in 48 patients withPCHD. (14 of 62 patients with PCHD were not included due to a lack ofadequate post-interventional follow up imaging.) One-tailed Fisher exacttests

ASA Clopidogrel Tirofiban

pPCHD p =0.269 p=0.278 p =0.070

PH p =0.121 p=0.096 p =0.407

Neuroradiology (2014) 56:41–50 47

Our data and the data provided by Parilla et al. do notsuggest there is an increased haemorrhage rate in mechanical-ly treated patients with PCHD, regardless whether transient orpersisting (Tables 3 and 4) [8].

Some authors have reported that persisting PCHD may beassociated with an increased risk for parenchymal haemor-rhage and clinical worsening [4, 6]. However, results providedby Nakano et al. as well as our data do not support thishypothesis [3]. None of the eight cases of parenchymal haem-orrhage was associated with persisting PCHD.

These results do conflict with analyses dealing with pa-tients who were treated via intra-arterial thrombolytic therapy:all authors observed an increased haemorrhage risk for pa-tients with PCHD [1–6].

In this context, results provided by Rouchaud et al., whoanalysed mechanically treated patients, stand out: the authorsreport 11 PH2 (28.9 %) and 21 PH1 (55.3 %) bleedings in 38patients with PCHD on the one hand, and no PH2 and 2 PH1(8.0 %) bleedings in 25 patients without PCHD. Their results

are highly significant, but it is noteworthy that the indicatedoverall haemorrhage rate of 53.0 % in their series is excep-tionally high [8, 28, 29].

The overall number of analysed patients may still be toosmall for a definitive statement. Nevertheless, present datamay imply that the natural history of PCHD in the contextof intra-arterial thrombolysis and PCHD in the context ofmechanical recanalisation differ. Present data may imply thathigh local concentrations of rtPA that are applied duringintra-arterial thrombolytic therapy bear an additional riskfor haemorrhage [8].

Correlation between PCHD and early signs of infarctionin initial imaging

One of the main goals of this study was to elucidate therelationship between PCHD and infarction. Yokogami et al.showed that PCHD are likely to appear in infarcted parenchy-ma. There was also a significant correlation between the

48 Neuroradiology (2014) 56:41–50

Fig. 3 Initial CT scan (a), post-interventional CT scan (b), and follow-up T2-FLAIR MRI (c) in a 56-year-old man show matching areas of PCHD (b :arrow) and ischaemia (c: arrow) in the right caudate and lentiform nucleus

Fig. 4 Post-interventional CTscan (a) and follow-up diffusion-weightedMRI (b) in a 43-year-old man. Infarcted areas (b) exceed areas with PCHD (a)in this case

presence of infarction in the initial CT scan and occurrence ofPCHD in our series. More specifically, the location of initialinfarction predicted the location of PCHD: Prior infarction inthe basal ganglia (lentiform nucleus and caudate nucleus) andthe surrounding cortex (insula andM1 area) were significantlyassociated with the occurrence of PCHD in these areas. Thisclose relationship between infarction and PCHD is alsoexpressed in an overall positive predictive value of 59.3 %for prior infarction to predict PCHD.

Correlation between PCHD and final infarction in follow-upimaging

The overall positive predictive value for PCHD to predict finalinfarction was 94.8 % in our series. Even though this figure isvery high, one would expect a higher positive predictive valuegiven the hypothesis that disruption of the BBB, thus infarc-tion, is postulated in the presence of PCHD. Final infarctdemarcation was lacking in three patients, more precisely in7 of 134 areas with PCHD. Sulcal contrast agent mimickingparenchymal PCHD due to volume effects might be a possibleexplanation for the discrepancy between PCHD and infarctdemarcation in two of these three patients. However, there wasa clear discrepancy between PCHD and final infarction in onepatient with a follow-up MRI examination (Fig. 5). Thisfinding was unexpected given the hypothesis that disruptionof the BBB is needed for the development of PCHD. Apossible explanation could be a transiently increased perme-ability of the BBB for the contrast agent Iopamidol due totransient ischaemia [30].

In summary, an area with PCHD ismost certainly infarcted.However, our data suggest that in rare cases PCHD may alsooccur in the context of transient ischemia.

Final infarction size was congruent with, or bigger than,areas of PCHD in all but three cases (93.3 %) in our series (seeabove). More precisely, infarction size surpassed areas ofPCHD in 57.8 % of patients (Fig. 4). The majority of PCHD

cases involved the basal ganglia—first and foremost thelentiform nucleus (80.6 %). Consequently, additional infarc-tion was mainly found in additional cortical areas. Further-more, our results imply that infarction size is likely tosurpass areas of PCHD if final recanalisation is unsat-isfactory (<TICI 2b).

In summary, final infarction size can be estimated on thebasis of PCHD even in the absence of classical early CT signsof infarction [31]. Infarction size is at least as large as areaswith PCHD in the majority of cases. More precisely, finalinfarction size is likely to surpass areas of PCHD and to affectmore than the basal ganglia if final recanalisation isincomplete.

Conclusions

We found that PCHD does not correlate with parenchymalhaemorrhage. Furthermore, PCHD is likely to correspond todysfunction of the blood–brain barrier and therefore mayserve as a valuable tool for predicting final infarction size.

Limitations

The retrospective approach is a major limitation of this anal-ysis. Nonetheless, there is need of a better understanding ofPCHD.We therefore think that even retrospective data may behelpful in daily clinical practice.

In addition, our analysis comprises the largest number ofpatients published to date and may serve as a landmark untilthere are more comprehensive analyses.

Finally, the binary nature of the ASPECTS system does notdo justice to partial changes in a given area, even though itallows comparability between different scans and is thereforea helpful tool in neuroradiological analysis.

Conflict of interest We declare that we have no conflict of interest.

References

1. Jang YM, Lee DH, Kim HS, Ryu CW, Lee JH, Choi CG, Kim SJ,Suh DC (2006) The fate of high-density lesions on the non-contrastCTobtained immediately after intra-arterial thrombolysis in ischemicstroke patients. Korean J Radiol 7(4):221–228

2. Komiyama M, Nishijima Y, Nishio A, Khosla VK (1993)Extravasation of contrast medium from the lenticulostriate arteryfollowing local intracarotid fibrinolysis. Surg Neurol 39(4):315–319

3. Nakano S, Iseda T, Kawano H, Yoneyama T, Ikeda T, Wakisaka S(2001) Parenchymal hyperdensity on computed tomography afterintra-arterial reperfusion therapy for acute middle cerebral arteryocclusion: incidence and clinical significance. Stroke 32(9):2042–2048

Fig. 5 Post-interventional CTscan (a) and follow-up diffusion-weightedMRI (b) in a 77-year-old woman. There is PCHD (a : arrow) withoutcorresponding infarction (b : arrow) in the right insula and inferior frontalgyrus

Neuroradiology (2014) 56:41–50 49

4. Wildenhain SL, Jungreis CA, Barr J, Mathis J, Wechsler L, HortonJA (1994) CT after intracranial intraarterial thrombolysis for acutestroke. AJNR Am J Neuroradiol 15(3):487–492

5. Yokogami K, Nakano S, Ohta H, Goya T, Wakisaka S (1996)Prediction of haemorrhagic complications after thrombolytic therapyfor middle cerebral artery occlusion: value of pre- and post-therapeutic computed tomographic findings and angiographic occlu-sive site. Neurosurgery 39(6):1102–1107

6. Yoon W, Seo JJ, Kim JK, Cho KH, Park JG, Kang HK (2004)Contrast enhancement and contrast extravasation on computed to-mography after intra-arterial thrombolysis in patients with acuteischemic stroke. Stroke 35(4):876–881

7. Kim JT, Heo SH, Cho BH, Choi SM, Lee SH, Park MS, Yoon W,Cho KH (2012) Hyperdensity on non-contrast CT immediately afterintra-arterial revascularisation. J Neurol 259(5):936–943

8. Parrilla G, García-Villalba B, Espinosa de Rueda M, Zamarro J,Carrión E, Hernández-Fernández F, Martín J, Hernández-Clares R,Morales A, Moreno A (2012) Haemorrhage/contrast staining areasafter mechanical intra-arterial thrombectomy in acute ischemicstroke: imaging findings and clinical significance. AJNR Am JNeuroradiol 33(9):1791–1796

9. Rouchaud A, Pistocchi, S, Blanc R, Engrand N, Bartolini B and PiotinM (2013) Predictive value of flat-panel CT for haemorrhagic trans-formations in patients with acute stroke treated with thrombectomy. JNeurointerv Surg. doi: 10.1136/neurintsurg-2012-010644

10. Pexman JH, Barber PA, Hill MD, Sevick RJ, Demchuk AM, HudonME, HuWY, Buchan AM (2001) Use of the Alberta Stroke ProgramEarly CT Score (ASPECTS) for assessing CT scans in patients withacute stroke. AJNR Am J Neuroradiol 22(8):1534–1542

11. LeeKY, Han SW, KimSH, NamHS,Ahn SW, KimDJ, Seo SH, KimDI, Heo JH (2007) Early recanalisation after intravenous administra-tion of recombinant tissue plasminogen activator as assessed by pre-and post-thrombolytic angiography in acute ischemic stroke patients.Stroke 38(1):192–193

12. Trouillas P, von Kummer R (2006) Classification and pathogenesis ofcerebral haemorrhages after thrombolysis in ischemic stroke. Stroke,Cerebrovascular Unit 37(2):556–561

13. Thomalla G, Sobesky J, Köhrmann M, Fiebach JB, Fiehler J, ZaroWeber O, Kruetzelmann A, Kucinski T, Rosenkranz M, Röther J,Schellinger PD (2007) Two tales: haemorrhagic transformation butnot parenchymal haemorrhage after thrombolysis is related to sever-ity and duration of ischemia: MRI study of acute stroke patientstreated with intravenous tissue plasminogen activator within 6 hours.Stroke 38(2):313–318

14. BanksWA (2009) Characteristics of compounds that cross the blood–brain barrier. BMC 9(1):S3

15. Lukic-Panin V, Deguchi K, Yamashita T, Shang J, Zhang X, Tian F,Liu N, Kawai H, Matsuura T, Abe K (2010) Free radical scavengeredaravone administration protects against tissue plasminogen activa-tor induced oxidative stress and blood brain barrier damage. CurrNeurovasc Res 7(4):319–329

16. Niego B, Freeman R, Puschmann TB, Turnley AM, Medcalf RL(2012) t-PA-specific modulation of a human blood–brain barriermodel involves plasmin-mediated activation of the Rho kinase path-way in astrocytes. Blood 119(20):4752–4761

17. Okajima K, Abe H, Binder BR (1995) Endothelial cell injury inducedby plasmin in vitro. J Lab Clin Med 126(4):377–384

18. Rudd MA, Johnstone MT, Rabbani LE, George D, Ware JA,Loscalzo J (1991) Thrombolytic therapy causes an increase in

vascular permeability that is reversed by 1-deamino-8-D-vasopres-sin. Circulation 84(6):2568–2573

19. Meyer JS (1958) Importance of ischemic damage to small vessels inexperimental cerebral infarction. J Neuropathol Exp Neurol 17(4):571–585

20. Osborn AG, Salzman KL, Barkovich AJ (2009) Diagnostic imaging:brain, 2nd edition. Lippincott Williams & Wilkins, Philadelphia

21. Brockmann C, Scharf J, Nölte IS, Seiz M, Groden C,Brockmann MA (2010) Dual-energy CT after peri-interventionalsubarachnoid haemorrhage: a feasibility study. Clin Neuroradiol20(4):231–235

22. Molina CA, Alvarez-Sabín J, Montaner J, Abilleira S, Arenillas JF,Coscojuela P, Romero F, Codina A (2002) Thrombolysis-relatedhaemorrhagic infarction: a marker of early reperfusion, reducedinfarct size, and improved outcome in patients with proximal middlecerebral artery occlusion. Stroke 33(6):1551–1556

23. Fiorelli M, Bastianello S, von Kummer R, del Zoppo GJ, Larrue V,Lesaffre E, Ringleb AP, Lorenzano S, Manelfe C, Bozzao L (1999)Haemorrhagic transformation within 36 hours of a cerebral infarct:relationships with early clinical deterioration and 3-month outcomein the European Cooperative Acute Stroke Study I (ECASS I) cohort.Stroke 30(11):2280–2284

24. HackeW, Donnan G, Fieschi C, Kaste M, von Kummer R, BroderickJP, Brott T, Frankel M, Grotta JC, Haley EC Jr, Kwiatkowski T,Levine SR, Lewandowski C, Lu M, Lyden P, Marler JR, Patel S,Tilley BC, Albers G, Bluhmki E, Wilhelm M, Hamilton S,ATLANTIS TI, null, ECASS TI and null, NINDS r-PASG I (2004)Association of outcome with early stroke treatment: pooled analysisof ATLANTIS, ECASS, and NINDS rt-PA stroke trials. Lancet363(9411):768–774

25. Larrue V, von Kummer RR,Müller A, Bluhmki E (2001) Risk factorsfor severe haemorrhagic transformation in ischemic stroke patientstreated with recombinant tissue plasminogen activator: a secondaryanalysis of the European-Australasian Acute Stroke Study (ECASSII). Stroke 32(2):438–441

26. (1997) Intracerebral haemorrhage after intravenous t-PA therapy forischemic stroke. The NINDS t-PA Stroke Study Group. Stroke28(11): 2109–2118

27. (1996) Thrombolytic therapy with streptokinase in acute ischemicstroke. The Multicenter Acute Stroke Trial—Europe Study Group. NEngl J Med 335(3): 145–150

28. Soize S, Barbe C, Kadziolka K, Estrade L, Serre I and Pierot, L(2013) Predictive factors of outcome and haemorrhage after acuteischemic stroke treated by mechanical thrombectomy with a stent-retriever. Neuroradiology 55(8):977−987

29. Möhlenbruch M, Seifert M, Okulla T, Wüllner U, Hadizadeh DR,Nelles M, Greschus S, Wilhelm K, Schild HH, Klockgether T,Urbach H (2012) Mechanical thrombectomy compared to local-intraarterial thrombolysis in carotid T and middle cerebral arteryocclusions: a single center experience. Clin Neuroradiol 22(2):141–147

30. Abrahám CS, Harada N, Deli MA, Niwa M (2002) Transient fore-brain ischemia increases the blood–brain barrier permeability foralbumin in stroke-prone spontaneously hypertensive rats. Cell MolNeurobiol 22(4):455–462

31. von Kummer R, Meyding-Lamadé U, Forsting M, Rosin L, Rieke K,Hacke W, Sartor K (1994) Sensitivity and prognostic value of earlyCT in occlusion of the middle cerebral artery trunk. AJNR Am JNeuroradiol 15(1):9–15, discussion 16–8

50 Neuroradiology (2014) 56:41–50