Clinical outcome after mechanical thrombectomy is highly variable despite high revascularization rates [1–12]. Clinical outcome, among other factors, is device-dependent, with stent retrievers showing higher rates of recanalization and favorable outcome than devices that cause thrombus fragmentation [5, 10]. This was confirmed by the SWIFT and TREVO 2 studies [5, 10]; however, for both MERCI and the stent retriever groups, the rate of good clinical outcome was lower than in old studies on intraarterial thrombolysis, despite that NIHSS is comparable and treatment selection was less restrictive in PROACT 2, MELT [9, 18], and the Swiss study , comparing intravenous (iv) and intraarterial (ia) thrombolysis (Table 3). The risk of embolic complications is of particular concern in mechanical thrombectomy, as emboli can impair collateral blood supply to the affected territory, accelerate penumbral tissue loss, and cause additional ischemic lesions . This is in accordance with a study demonstrating that the presence of multiple thrombi before initiation of thrombolytic therapy predicts worse outcome .
In the present study, we analyzed a series of patients undergoing protected stent retriever thrombectomy. Mechanical thrombectomy was performed in flow arrest and under aspiration using a balloon-mounted guiding catheter, a distal access catheter, or both. Peri-interventional EEs following thrombectomy were detected on SWI in 22.8 % of patients. Emboli were mostly solitary and showed a similar signal intensity as the primary occluding thrombus. Emboli were exclusively located in the vascular territory distal to the primary arterial occlusion and therefore resulted most likely from fragmentation during thrombus mobilization.
On DWI, new ischemic lesions associated with EE were found in three of 16 EEs. Even in retrospect, stagnation or reversed flow in the vicinity of the EE in addition with a subtle decrease in parenchymal blush could be presumed in four patients only (Figs. 2 and 3). These subtle signs of local flow restriction were visible when the images of the lateral projection of the DSA were evaluated dynamically, but not on the AP view or in the 3D rotation angiography.
The small size of the EE may be one reason for the restricted visibility in the DSA. In contrast to EEs in new vascular territories, which are normally well visible even on frozen DSA images (1–3), downstream EEs migrate only to the watershed zone, where the antegrade flow and pressure are in balance with the reversed flow of the collaterals. Therefore, comparable with embolic basilar artery occlusion, downstream iatrogenic EEs are not fixed and compressed in an artery with a diameter smaller than the EE, which can result in small oscillations of the thrombus and discrete bloodflow around the EE. This may be another reason, why it is difficult to detect downstream EE as a clear-cut branch occlusion. Considering the sensitivity of DSA to motion artifacts, it is not a surprise that indirect discrete signs of flow stagnation may be obliterated and may be detected retrospectively only, after post-processing of the DSA, including motion correction.
Currently, modified TICI is considered the optimal grading system for the evaluation of endovascular recanalization . In view of our finding that the majority of peri-procedural EEs are invisible in DSA, post-treatment SWI may be a valuable complement to DSA for assessment of the success and safety of endovascular procedures.
Peri-interventional emboli, as detected on DSA, were found in up to 48 % of patients treated with the Penumbra system , and an analysis of the histological structure of clots extracted by the Merci retrieval system reported that in 64 %, thrombi were retrieved in multiple fragments . In the MR CLEAN study using various types of intraarterial techniques, iatrogenic embolization to primarily non-affected vascular territories was observed in 8.6 % of patients and 5.6 % had clinical signs of a new ischemic stroke . A blinded core lab found nine new emboli (6 %), two of them in primarily non-occluded arteries (1.4 %), in a multicenter study, using the Solitaire stent retriever with balloon protection in 74 % of interventions . In a study with mandatory use of balloon protection, two emboli were detected in 202 patients (1 %) by the corelab and no embolic event was observed in the subgroup of 119 patients treated with iv thrombolysis in addition to protected stent retriever thrombectomy . The low rate of embolization to primarily non-affected vascular territories, serving as collaterals to the penumbral tissue , might contribute to the better clinical outcomes seen in the latter studies [2, 3] compared to those using predominantly devices and techniques that achieve recanalization through thrombus fragmentation [6, 7, 15–17] or stent retriever thrombectomy without mandatory protection [1, 5, 10] (Table 3).
Modern CT  and MRI techniques  allow pre-interventional visualization of thrombus length and shape, helping the interventionist to deploy the stent retriever over the whole thrombus length. In our study, retraction of the thrombus was performed in flow arrest and simultaneous aspiration using a balloon-mounted guiding catheter. In addition, in patients with anatomically difficult access, a distal access catheter was used. This might explain the fact that distal emboli were only detected in the vascular territory distal to the primary occlusion and not in adjacent territories (e.g., ACA territory). The tiny EEs seen downstream to the primary occlusion site may have been caused by shearing-off of small thrombus parts to perforating arteries during thrombus mobilization, a phenomenon that could be observed in animal models using radiopaque thrombi . Creation of the iatrogenic EE during the passage of the occlusion is unlikely, because we know that the microcatheter never penetrates the thrombus itself but passes the site of occlusion between the vessel wall and the thrombus . In addition, careful microcatheter contrast injection distal to the occlusion is a mandatory part of our stroke protocol, to confirm that the size of the artery is suitable to deploy the stent retriever. However, we have never been able to detect distal EE, neither primary fragmented thrombi, as discussed recently in a theoretical framework , nor iatrogenic EE.
Interestingly, this assumption may be supported by a trend for less frequent EE when distal access catheters were used (6.7 vs 28.6 %), but the difference was not significant (P = 0.150). Distal access catheters may facilitate the mobilization of thrombi by adding the suction power of the aspiration catheter to the traction force of the stent retriever, reducing the risk of distal embolization during thrombus mobilization. In addition, the distance the thrombus has to be retracted through the vessel is minimized, thus protecting the origins of arterial branches that have to be passed during retraction.
The rates of favorable clinical outcome at 3 months were not significantly different in patients with and without emboli (58.3 vs 70.7 %; P = 0.490). This seems to be in contrast to a recent study that demonstrated that primarily fragmented thrombi detected before initiation of thrombolytic therapy predicted worse outcome . This discrepancy could have three reasons. First is that the number of patients with EE was too small to show a significant difference. Second is that emboli seen in our patient cohort were smaller than those in patients with multiple thrombi before initiation of thrombolytic therapy . In fact, only a minority of patients with EE showed signs of tissue damage on DWI that could be attributed to iatrogenic emboli. Last, the reason for the good clinical outcome might be that no embolization to primarily non-affected vascular territories occurred in patients with and without EE.
EEs have been found in a significantly higher percentage of patients who underwent thrombectomy with conscious sedation compared to patients treated with general anesthesia (40.0 vs 13.5 %; P = 0.044). Unlike in general anesthesia, patients with conscious sedation often move during the potentially painful mobilization and retraction of the thrombus. Patient movement might increase the risk of embolization, as it impairs the quality of flow arrest and aspiration and mobilization and retraction of the thrombus cannot be performed in a slow and controlled fashion.
Our study has some limitations. First, not all EEs seen on post-interventional SWI may have actually been related to mechanical thrombectomy itself. New EEs could also be related to ongoing embolization from the primary embolic source of the stroke or migration of thrombus remnants still in place after thrombectomy. Second, the extensive artifacts on SWI generated in the proximity of the skull base limit the evaluation of the adjacent brain parenchyma, the vessels of the posterior circulation, and the intracranial parts of the ICA. Third, the distinction between embolus and hemorrhagic transformation on SWI can be equivocal if the hypointensities are located inside infarcted parenchyma. In such patients, only isolated dot-like lesions were counted as emboli. However, the majority of EEs were detected outside the infarcted territory (n = 9/13), making hemorrhagic transformation very unlikely. Moreover, most EEs were of a slightly elongated shape (Figs. 1, 2, and 3), due to the orientation of the thrombotic material in the vessels. This helps to distinguish them from cerebral microbleeds, usually presenting spherical in SWI . Quantitative susceptibility mapping (QSM) was performed in a few of our patients only but may improve the specificity of this new source of MR image contrast [33, 34].
Fourth, in many patients, the time span between endovascular treatment and post-treatment imaging was quite long. Mean size of the EE as measured on SWI was 5.2 mm. However, the length of the thrombi may be shorter due to the blooming effect  and oscillations of the downstream EE, as described above. It cannot be excluded that ongoing thrombolysis may have caused the disappearance of some peri-interventional EEs on post-treatment SWI. However, EEs were equally distributed in patients treated primarily by endovascular approach (23 %; seven EEs in 30 patients) and the bridging group (22 %; six EEs in 27 patients), but the number of patients is too small, to draw any conclusion.
Fifth, the field strength of the MRI scanners was not standardized across patients because we use several scanners in clinical stroke imaging. Previous studies have shown that the detectability of cerebral microbleeds on SWI improves with increasing field strength . Thus, the detection rate of EE is most likely influenced by field strength as well. Nonetheless, this phenomenon did not seem to have a major impact on our study results, as six EEs (46.2 %) were detected on SWI acquired with a 1.5-T and seven EEs (53.8 %) with a 3-T MRI.