Variables of interest

Identifying the target anastomotic territory for the hemorrhage, the primary variable of interest, was a systematic process consisting of the following 3 steps: 1) identification of the hemorrhagic site associated with the present hemorrhagic event; 2) identification of the target collateral vessels responsible for the hemorrhage; and 3) identification of the target anastomotic territory fed by the target collateral vessels. Assessment of imaging scans for the target collateral vessel was performed via the baseline angiography through consensus reading by 2 raters, each with 5 years’ experience in MMD imaging evaluation, who were blinded to treatment and clinical outcome. In cases of disagreement between the 2 raters, a third investigator with > 10 years’ experience in MMD imaging resolved any discrepancies between the 2 raters. Both raters had participated in a training session involving 15 representative cases (including 5 recurrent cases) in which the target collateral vessels had been confirmed by angiography.

We defined the target anastomotic territory as the feeding territory from the terminal portion of the target collateral vessel, located beyond (with collateral anastomosis) or at (without collateral anastomosis) the hemorrhagic site, as seen on frontal and lateral views in the later arterial phase of angiography. The target anastomotic territory was thus classified as the anastomotic territory of the medial medullary artery, lateral medullary artery, multiple medullary arteries, or a nonmedullary artery based on the anatomical and radiological evidence [12, 13]; medullary arteries were defined as arteries supplying the cerebral white matter. The anastomotic territory of the medial medullary artery was defined as the area fed by the cortical branches of the anterior cerebral artery (ACA) and the medial branches of the posterior cerebral artery (PCA) (parietooccipital artery). The anastomotic territory of the lateral medullary artery was defined as the area fed by the cortical branches of the middle cerebral artery (MCA). The anastomotic territory for multiple medullary arteries was defined as the area fed by the cortical branches of any two of the above-mentioned major arteries (ACA, MCA, and PCA). The anastomotic territory of the nonmedullary artery was defined as the area located at the basal ganglia region and thalamus and fed by the basal perforators (medial and lateral striate arteries, thalamic perforators, and choroidal arteries). Schematic illustrations of these definitions are shown in Fig. 1.

Schematic illustrations demonstrating imaging definitions of the medial, lateral, and nonmedullary artery anastomotic territories in different target collateral vessel patterns. Med = medullary; A = artery; LSA = lenticulostriate artery; ThA = thalamic artery; ChA = choroidal artery

Previously reported angiographic definitions of the periventricular collateral vessels, which derive from the lenticulostriate artery (LSA), thalamic artery (ThA), and choroidal artery (ChA), were adapted to the present study [10]. The target collateral vessel was defined as an abnormally extended periventricular collateral vessel corresponding to the distribution of the presumed hemorrhagic site on angiography. The hemorrhagic site was generally identified with noncontrast-enhanced CT performed during the acute phase of the hemorrhage. When the hemorrhage presented as diffuse ventricular hemorrhage, susceptibility-weighted imaging (SWI) was performed within 2 months of the hemorrhagic event to identify the presumed hemorrhage origin, which was defined as an abnormal hypointense area > 10 mm that overlapped with the signal of the abnormally extended collateral vessel on SWI and was also apparent on T2-weighted imaging. The initial presumed hemorrhagic site was considered to be responsible for the recurrent hemorrhage when the recurrent hemorrhage occurred in the same hemisphere containing the initial target collateral vessel and corresponded to the distribution of recurrent hemorrhagic sites.

Other baseline variables were also evaluated, such as the occurrence of posterior hemorrhage [8], the presence of ruptured collateral aneurysm, the Suzuki stage of the disease [1], and involvement of the PCA (Mugikura stages II-IV) [14].