Implementation of multimodal computed tomography in a telestroke network: Five‐year experience

Abstract Aims Penumbral selection is best‐evidence practice for thrombectomy in the 6‐24 hour window. Moreover, it helps to identify the best responders to thrombolysis. Multimodal computed tomography (mCT) at the primary centre—including noncontrast CT, CT perfusion, and CT angiography—may enhance reperfusion therapy decision‐making. We developed a network with five spoke primary stroke sites and assessed safety, feasibility, and influence of mCT in rural hospitals on decision‐making for thrombolysis. Methods Consecutive patients assessed via telemedicine from April 2013 to June 2018. Clinical outcomes were measured, and decision‐making compared using theoretical models for reperfusion therapy applied without mCT guidance. Symptomatic intracranial hemorrhage (sICH) was assessed according to Safe Implementation of Treatments in Stroke Thrombolysis Registry criteria. Results A total of 334 patients were assessed, 240 received mCT, 58 were thrombolysed (24.2%). The mean age of thrombolysed patients was 70 years, median baseline National Institutes of Health Stroke Scale was 10 (IQR 7‐18) and 23 (39.7%) had a large vessel occlusion. 1.7% had sICH and 3.5% parenchymal hematoma. Three months poststroke, 55% were independent, compared with 70% in the non‐thrombolysed group. Conclusion Implementation of CTP in rural centers was feasible and led to high thrombolysis rates with low rates of sICH.


| INTRODUC TI ON
Almost 20 years ago, the term telestroke was coined to define the emergent use of telemedicine in acute stroke. 1 Using a camera and having access to brain computed tomography (CT), neurologists at remote sites were able to determine whether a patient would be candidate for reperfusion treatment with intravenous thrombolysis.
It was subsequently shown that telestroke achieved similar safety and outcome results to those in comprehensive stroke centers. [2][3][4] Importantly, however, acute stroke treatment has changed dramatically over the last few years, endovascular thrombectomy (EVT) is now the standard of care in large vessel occlusion (LVO) strokes.
Moreover, recent trials supporting the use of multimodal imaging (including brain noncontrast CT, CT, or MR angiography and perfusion imaging) in the 24-hour window 5,6 have driven a need for more sophisticated imaging-based patient selection. There is also data indicating that such imaging may allow better patient selection and lower rates of symptomatic intracranial hemorrhage (sICH) in those receiving intravenous thrombolysis. 7,8 John Hunter Hospital is the comprehensive stroke centre for the Hunter and New England regions of New South Wales, Australia.
Since April 2013, a telestroke network, where multimodal CT (mCT) was performed routinely at all sites has been developed, aiming to identify potential candidates for thrombolysis and EVT. 9 Although mCT has typically been restricted to comprehensive stroke centers, recently published trials 5,6 provided a strong rationale for its use to aid EVT transfer decision-making from regional and rural centers.
We aimed to a)describe our initial 5-year experience applying multimodal CT imaging in telestroke and b)determine the influence of mCT on thrombolysis decision-making. Our principal hypothesis was that the use of mCT implemented in regional hospitals and supported by telestroke would deliver more refined patient selection for thrombolysis-specifically that it would allow selection of those most likely to benefit from therapies-based on presence of a vessel occlusion and "target" mismatch-, and also of those unlikely to benefit, such as stroke mimics, large infarct cores, or small perfusion lesions where the natural history is excellent. 10

| Telestroke network
We established a telestroke network to provide acute stroke ser-  As part of the network, the local hospitals were equipped with cameras and the physicians were trained in the face arm speech time (FAST) scale. mCT was introduced and performed routinely by trained radiology technicians. Scans were interpreted in the acute phase by the stroke neurologist. For more details, we direct the reader to our published pilot phase experience. 9

| Imaging protocol and data collection
The mCT imaging protocol included brain noncontrast CT, CT angiography (CTA), and CT perfusion (CTP) at baseline and either NCCT or MRI at 24-48 hours. The spoke sites used different CT scanners with a-to z-axis coverage between 80 and 150 mm. A 40 mL bolus of iodinated contrast at a rate of 6 mL/s was used to acquire CT perfusion images, lasting between 60 and 72 seconds (depending on the individual scanner protocol). Extracranial CTA was performed afterward, using another 50 mL of contrast agent (rate of 6 mL/s). Intracranial CTA was reconstructed from the CTP acquisition.
All imaging was postprocessed using the commercial software MIStar (Apollo Medical Imaging Technology), which automatically generated cerebral blood volume, cerebral blood flow (CBF), mean transit time and delay time (DT) maps, as well as infarct core and penumbra maps. Penumbra was defined as the tissue with a DT >3 seconds and relative CBF >30% of normal tissue. 12 Ischemic core was defined as the tissue with a DT >3 seconds and a relative CBF <30% of the contralateral hemisphere. 13

| Thrombolysis decision: multimodal CT versus standard clinical/NCCT criteria
We hypothesized that the use of mCT would allow selection of those most likely to benefit from reperfusion therapies-based on presence of a vessel occlusion and "target" mismatch-, and also of those unlikely to benefit, such as stroke mimics. Our local protocol was that a thrombolysis decision was based on both standard guideline-based clinical criteria 14 plus mCT imaging decision assistance using the presence or absence of "salvageable tissue", defined as at least 15 mL of penumbra assessed by automated perfusion software. The decision for thrombolysis was made by the treating telestroke vascular neurologist in a "real world" clinical practice setting. A patient was con- A vessel occlusion was defined as any visible occlusion identified by the treating stroke neurologist.

| Non-thrombolysed patients
In order to gauge the potential influence of CTP decision-making on patient outcomes, we assessed outcome not only in those who received thrombolysis, but also in those in whom the use of CTP led to the decision not to thrombolyse and in whom there were no "standard" clinical and NCCT exclusion criteria for thrombolysis. This group differs from the standard clinical/NCCT criteria group mentioned in the above section, since reflecting most current thrombolysis guidelines, we did not apply a NIHSS threshold.

| Statistical analysis
Results are presented as mean ± standard deviation (SD) and me-

| mCT-based thrombolysis
Of the 240 patients that underwent mCT, thrombolysis was given in 58 (24.2%), 16 were transferred for EVT (7%), and seven received combined therapies ( Table 1 and Table S1 of

| Non-thrombolysed patients
The characteristics of the 240 mCT-assessed patients and the various subgroups within are shown in Table 1 There were 108 patients with no standard clinical contraindications to thrombolysis but who were excluded based on CTP. Just three of these patients had a LVO. One of these was not considered suitable to treatment due to poor baseline function and severe comorbidities, another was not treated with thrombolysis but transferred for clot retrieval and the third patient was asymptomatic at presentation.

| D ISCUSS I ON
We describe our first five-year experience of a telestroke network with routine use of multimodal CT. Intravenous thrombolysis was delivered to 58 patients, 17.4% of all the calls received during this period. The thrombolysed patients had moderate/severe strokes (median baseline NIHSS of 10), and 74% had a visible vessel occlusion on CTA. Of these patients, 55% were independent three months after stroke. Interestingly, more than 80% of the nonthrombolysed group were independent 3 months after stroke.
A combination of (a) very small or absent perfusion lesions, (b) low percentage of large vessel occlusion (10%), and (c) presence of mimics (44 patients) is the probable explanation for the high rates of good outcome in the non-thrombolysed group. These data suggest that mCT can identify patients traditionally eligible for thrombolysis, but who actually have an excellent natural history, including stroke mimics and those who are likely already in the process of spontaneous reperfusion (those with significant neurologic deficits but very small cortical perfusion lesions, often in locations not corresponding to their deficits). Our data add to previous observations suggesting that patients with small perfusion deficits may do as well or even better without thrombolysis, calling into question strategies that reward high thrombolysis rates in a relatively undifferentiated population of patients presenting with acute focal neurologic deficit. 7,10 We compared our thrombolysis outcomes with those from SITS-MOST registry 16 and other telestroke networks. The three-month rate of independence was 55% in SITS-MOST, identical to this study.
SITS-MOST had a slightly more severe population (baseline NIHSS 12 versus 10 in our population), but lower rate of vessel occlusions.
Of the SITS-MOST patients, 16% were severely impaired after stroke (bedridden/dead), compared with 13.8% in our study. Remarkably, the rate of sICH was 1.7% in our study, compared with 7.3% in SITS-MOST. This low rate is even more remarkable when one considers that it does not include the many patients who were excluded from thrombolysis based on small or absent perfusion lesions, who would be expected to have very low rates of sICH. A similar very low rate decision-making. In this randomized clinical trial and meta-analysis, multimodal CT was able to identify best responders to thrombolysis in the 4.5-9 hours window, achieving better functional outcomes, despite a higher rate of symptomatic hemorrhage.
A limitation in this observational study is the small part of the data was collected retrospectively (26%), and with relatively small numbers of patients receiving reperfusion therapies, which may affect reliability of estimates of outcomes such as sICH.
Nevertheless, our results highlight the feasibility of using mCT in smaller centers lacking stroke neurologists, in particular its potential for reperfusion therapy decision-making. The data regarding hemorrhage rates were very favorable. Process of care times (acknowledging room for improvement), and outcomes compared favorably with published data, and of note, the outcomes looked even better when based on all those patients who would have been eligible for therapy based on standard clinical and NCCT criteria.
In conclusion, our article shows the feasibility of mCT implementation in a rural telestroke network, suggesting enhanced safety of mCT-based imaging selection.

ACK N OWLED G M ENTS
We