Anatomic Pattern of Intracerebral Hemorrhage Expansion: Relation to CT Angiography Spot Sign and Hematoma Center

Abstract

Background and Purpose

We sought to identify baseline determinants of the anatomic pattern of hematoma expansion in patients with intracerebral hemorrhage and spot sign.

Methods

We coregistered baseline and follow-up CT scans from 15 intracerebral hemorrhage patients and measured growth at each surface node from baseline to follow-up hematoma. We analyzed the effects of proximity to the spot sign or hematoma center on distance of expansion, controlling for covariates.

Results

There was substantial node-to-node variation in extent of expansion around each hematoma surface (mean coefficient of variation for expansion distance 0.43, 95% CI 0.39–0.48), indicating nonuniform expansion. Closer proximity to the hematoma center was independently associated with increased expansion (0.185mm greater expansion for each 1mm closer to the center, p<0.0001). Closer proximity to the spot sign was not independently associated with increased expansion in models including both terms.

Conclusions

Hemorrhages expand nonuniformly around their surface with a tendency for greater expansion closer to their center. These findings provide a novel framework for analyzing mechanisms underlying hemorrhage growth and response to treatment.

Hematoma expansion (HE) is an important contributor to the poor outcome of intracerebral hemorrhage (ICH).¹ A barrier to developing effective treatments for preventing HE is our limited understanding of its underlying mechanisms. Various risks for HE have been identified,² including the CT angiography (CTA) spot sign thought to represent active bleeding,³ but few data exist on the features and mechanisms of this phenomenon. Neuropathological evidence suggest HE could partly entail secondary rupture of surrounding vessels,⁴ potentially offering new approaches to prevention.

In the current study, we develop and apply tools for quantifying the anatomic pattern of HE. We specifically tested the uniformity of expansion around the hematoma surface and its relation to the spot sign and hemorrhage center.

Methods

This study is a retrospective analysis of prospectively collected CT and clinical data from patients enrolled in an ongoing ICH study approved by the Institutional Review Board. Full details are in Supplemental Methods (http://stroke.ahajournals.org).

Subjects met the following criteria: 1) baseline CT/CTA with primary ICH, volume < 100 mL, and at least one spot sign, and 2) follow-up noncontrast CT within 72 hours showing expansion by ≥5% of baseline volume. Patients with >5 mL intraventricular, subarachnid, or subdural extension or surgical evacuation prior to follow-up were excluded. From 1994 to 2010, 63 patients fulfilled the inclusion criteria, 44 were excluded as above, and 4 had unavailable or unsuitable images, leaving 15 for analysis. Baseline CTA and follow-up CTs were acquired and spot signs identified as described.⁵

After coregistration of baseline and follow-up scans and ICH and spot segmentation, HE was analyzed by computation of the expansion distance from baseline to follow-up hematoma surface at each analyzed location (“node”) on the baseline surface and the distances from each node to the volumetric centers of the spot sign and baseline hematoma (Fig. 1).

Figure 1.

Quantitative measurement of expansion pattern. Sections from baseline CTA and follow-up CT (A) demonstrate a thalamic ICH (outlined in blue, spot sign in dark gray) and its evolution (green). Also shown are schematic two- (B) and three- (C) dimensional representations of co-registered ICHs and spot masks and a three-dimensional view (D) of the baseline ICH with thin lines representing expansion along vectors emerging from the baseline ICH surface to the follow-up boundary (thin green line). Expansion in D is coded by a continuous shade scale, largest expansion distances shown as dark.

Statistical analyses assessed the coefficient of variation (standard deviation divided by mean) of each subject’s node-by-node expansion distances and the determinants of expansion distance at a given node by linear mixed effects models controlling for other known determinants of HE.²

Results

Characteristics of the 15 ICH subjects are in Table 1. There was marked nonuniformity of expansion around the surface of the baseline hematomas, with mean node-to-node coefficient of variation 0.43 (95% CI 0.39 to 0.48; Supplemental Table I). The coefficient of variation was was observed to be greater in the 8 non-anticoagulated ICHs than the 7 warfarin-related ICHs (p=0.05).

Table 1.

Subject characteristics

Age (yrs)	71.9±11.0
Male	10 (67%)
Warfarin treatment	7 (47%)
Antiplatelet treatment	1 (7%)
Symptom onset to baseline CT, h	3.0±2.7
Baseline CT to follow-up CT, h	28.5±18.2
Baseline ICH Volume, mL	25.3±15.8
Volume of Expansion, mL	9.3±9.1
Proportional expansion, % baseline	29.1 (14.2, 50.2)
Patients with 2 spots	4 (27%)

Values are mean ± SD or median (interquartile range) as appropriate.

Spot signs tended to be located almost as far from the hematoma center as the median hematoma surface node (center-to-spot distance = 87.0 ± 30.8% of median center-to-node distance), indicating relatively eccentric placement of spots (Fig. 2A and D). Controlling for time from symptom onset to baseline CT, baseline volume, and warfarin use, proximity to hematoma center was associated with increased extent of expansion (Supplemental Table II): For each 1 mm closer distance to the centroid, extent of expansion increased by 0.185 mm. Conversely, expansion increased by only 0.014 mm for each 1 mm closer proximity to the spot sign, and this effect did not remain independent in models including both distance terms (Supplemental Table II). Additional models containing an interaction term for subject-specific quartiles of centroid distance (Supplemental Methods) found the greatest effect of centroid distance on expansion in the quartile of nodes closest to the center.

Figure 2.

Sample expansion patterns. Baseline CTA (A and B) or CT (C) are shown at left, follow-up noncontrast CTs in center, and schematic two-dimensional representations of co-registered images with outlined baseline (orange) and follow-up (green) hematoma boundaries, spot sign centers (red dot), and baseline hematoma centers (yellow plus) at right.

Qualitative inspection (Fig. 2) supported the quantitative findings and suggested no predisposition for expansion in particular anatomic directions, other than the requirement for study inclusion that the hematoma remain largely confined to the brain parenchyma.

Discussion

Our quantitative analysis of the pattern of HE demonstrates that expansion occurs with substantial variation around the surface of the baseline ICH. The only factor found to increase expansion along the surface is proximity to the hematoma center. Although proximity to the spot sign was also weakly associated with increased expansion, this effect appeared entirely due to confounding by proximity to the hematoma center.

The marked node-to-node variations in expansion observed in the current analysis appear consistent with C. Miller Fisher’s neuropathological model of hemorrhage growth occurring via secondary shearing of adjacent vessels triggered by the growing hemorrhage.⁴ The Fisher “avalanche” model has also been supported by the observation of multiple contrast spots within a single hematoma⁶ and by a recent computer simulation⁷ in which the virtual hematomas expanded in a strikingly nonuniform fashion.

The mechanism for increased expansion closer to the hematoma center is unclear and raises the possibility that the physical features of the hematoma-parenchyma interface might favor a spherical shape. The absence of a strong effect of spot sign proximity on expansion, conversely, suggests that this snapshot of ongoing bleeding⁸ may be too limited a sampling to predict the overall pattern of future growth. The short time window sampled by the CTA spot sign may also explain its imperfect ability to predict expansion (only 61% positive predictive value and 78% negative predictive value for significant expansion in the PREDICT study¹).

An important limitation was our inability to quantitatively analyze effects of surrounding tissue structure on expansion. We also note that our results apply only to expansion between baseline and follow-up imaging, though a hyperacute ICH serendipitously captured by neuroimaging also found markedly asymmetric growth.⁹ Finally the results apply only to hemorrhages meeting our inclusion/exclusion criteria.

HE remains an important, potentially modifiable target for acute ICH therapies. The current results add to our understanding of the mechanisms of expansion and indicate that locations close to the hematoma center are more likely to grow. This approach also provides an analytic framework applicable to ongoing¹⁰ and future clinical trials aimed at limiting HE.