Keywords
satellite sign - prediction - hematoma expansion - spontaneous intracranial hemorrhage
- computed tomography
Introduction
Intracerebral hemorrhage (ICH) is a common condition that is associated with a mortality
rate as high as 40% at 1 month.[1] Hematoma expansion (HE), which occurs in about 30% of spontaneous ICH patients,
has been identified as a factor that causes neurological deterioration and poor outcome.[2] In contrast to other predictors, such as age, underlying disease, initial volume
of ICH, and location, HE is the most important modifiable predictor that can change
the outcome of ICH.[3] Several previous studies found and reported the spot sign on computed tomography
angiography (CTA) to be a good predictor of HE.[4]
[5]
[6]
In the setting of our clinical practice, CTA is not currently part of the routine
workup, and most ICH patients have a follow-up CT only when they develop clinical
deterioration. The saying “time is brain: the outcome of stroke is generally worse
the longer treatment is delayed,”[7] has been and continues to be a fundamental precept in ischemic stroke care. We agree
that time is a key factor that relates to poor outcome in ICH patients, which means
that early identification of patients at high risk for HE by initial non-contrast
CT (NCCT) is important for anti-HE management.
Several NCCT markers of HE have been reported, such as irregular shape, heterogeneous
density, intra-hematoma hypodensities, and hematoma sedimentation levels; however,
the optimal predictor of HE on NCCT is still being debated.[8]
[9]
[10]
[11]
[12] In 2017, Shimoda et al reported the presence of a satellite sign, which they defined
as a small hematoma completely separate from the main hematoma on at least one slice,
on the initial NCCT that they found to be associated with a significantly worse functional
outcome in ICH patients.[13] Later, Yu et al, using the ABC/2 method to estimate hematoma volume, suggested the
satellite sign an independent predictor of HE. Although spot sign from CTA has higher
predictive accuracy, the satellite sign is still an acceptable predictor of HE when
CTA is unavailable.[14]
At our center, most spontaneous ICH patients receiving conservative treatment will
have a follow-up CT if they demonstrate clinical deterioration or at least within
7 days. No previous study has investigated the efficacy of satellite sign for predicting
HE in our population. Moreover, we postulated that using CT image analysis software
to measure hematoma volume would yield more precise data and decrease inter-reader
variability. Accordingly, the aim of this study was to investigate the potential of
satellite sign for prediction of HE in spontaneous ICH patients who had follow-up
NCCT within 7 days after the initial CT scan by using CT image analysis software to
determine hematoma volume.
Materials and Methods
Patients
We retrospectively reviewed cases recorded in the ICH database at our hospital during
January 2013 to December 2018. Cases were eligible for inclusion if they met the following
criteria: (1) age of 18 years or older, (2) spontaneous ICH confirmed by CT scan,
(3) onset-to-CT time 12 hours or lesser, and (4) follow-up CT scan was conducted within
168 hours or 7 days after the initial CT scan. The exclusion criteria were: (1) secondary
ICH, such as hemorrhagic transformation from a cerebral infarct, cerebrovascular anomalies,
tumor, or trauma; (2) unavailable baseline CT or follow-up CT; and/or (3) surgical
evacuation before follow-up CT scan. A flow diagram describing the patient enrollment
process is shown in [Fig. 1]. The following baseline information of spontaneous ICH patients was collected: demographic
variables (age and gender), medical history (hypertension, chronic kidney disease),
medication (antiplatelet drug), admission Glasgow Coma Scale score, admission blood
pressure (BP) levels (systolic and diastolic), and laboratory data (platelet count,
prothrombin time [PT], activated partial thromboplastin time [aPTT], and creatinine).
Fig. 1 Flow diagram of the patient enrollment process.
The protocol for this study was approved by our Institutional Review Board (SIRB)
(COA no. 737/2561). The requirement to obtain written informed consent was waived
due to the retrospective nature of this study.
CT Imaging Technique
The CT scans performed on patients included in this study were performed on a 64-slice
CT scanner (GE LightSpeed VCT) (GE Discovery) or a 256-MDCT (GE Revolution CT) (all
GE Healthcare, Chicago, Illinois, United States). The CT imaging parameters were as
follows: 120 kVp; 200 to 490 mA; section thickness, 1.25 mm; pitch, 1:1.
Image Interpretation
ICH was diagnosed using axial 1.25-mm thick section images obtained from brain CT,
with CT density profiles within the range established for hemorrhage based on Hounsfield
units (HUs). The location of ICH was classified into the following four areas: lobar
(cortex and subcortical white matter), deep (basal ganglia, thalamus, internal capsule,
and deep periventricular white matter), brain stem, and cerebellum. When hemorrhage
volume was large and had spread to more than two areas, the location of ICH was defined
on the basis of the area of the main hemorrhage. The onset-to-CT time and the duration
from the initial CT to the follow-up CT were also recorded.
Measurement of the Main Hemorrhage Size
Hematoma volumes on CT scans were measured by Stroke VCAR (Volume Computer-Assisted
Reading) software (GE Healthcare). Stroke VCAR is a CT image analysis software package
that is designed to evaluate NCCT images. Hematoma segmentation is semiautomated and
is triggered by a user click. The exam is loaded and the region-of-interest (ROI)
of the suspected hematoma is selected. The software then performs an automated segmentation
of the ROI if the volumetric parameter falls within the hematoma HU range ([Fig. 2]). The software also provides the user with a tool called Smart Mesh to edit the
hematoma segmentation results ([Fig. 3]). In cases where that hemorrhagic lesion cannot be detected automatically by the
software, ROIs can be manually drawn by tracing the perimeters of the hematoma on
each slice throughout the hemorrhagic lesion. The ROIs in every slice are then summed
after adjusting for slice thickness to determine the hematoma volume. Accuracy for
intracerebral hematoma is 85%, with an average variation of 15% underestimation (Stroke
VCAR User Guide Direction 5486219-1EN, Revision 3). Intraventricular hemorrhage (IVH),
subarachnoid hemorrhage, and satellite hemorrhage, when present, were not included
in the size measurement.
Fig. 2 Automated segmentation of the region of interest (ROI) based on the hematoma Hounsfield
unit (HU) range.
Fig. 3 Smart Mesh to edit the hematoma segmentation results. Blue highlight indicates the volume to be added to the segmented object. Red highlight indicates the volume to be removed.
The minimum slice thickness and volume values supported by the hematoma segmentation
are as shown in [Table 1].
Table 1
Minimum slice thickness and volume values supported by the hematoma segmentation
|
Slice Thickness (St) Range (in mm)
|
Minimum Supported Size (Volume in cm3)
|
|
St < 2.5
|
2
|
|
2.5 ≤ St ≤ 5
|
5
|
Due to the minimum support size of the software, patients with hematoma volume less
than 2 mL (on initial CT) were excluded from this study.
Hematoma Expansion
HE was defined by two methods. HE was first inspected and measured by a radiologist
on the axial plane (expansion vs. non-expansion). HE was then evaluated using image
analysis software to obtain the quantitative results (mL and percentage change). The
association between satellite sign and each 1% increase in hematoma volume was calculated.
Substantial HE (SHE), defined as absolute growth greater than 12.5 mL or relative
growth greater than 33% in hematoma size, was also used to evaluate potential association
with satellite sign.[15]
Definition of Satellite Sign
Hemorrhages were classified into two types according to the presence or absence of
a satellite sign around the main hemorrhage ([Fig. 4]). From a previous study, the criteria for satellite sign on NCCT were: (1) presence
of a small hematoma completely separate from main hematoma on at least one slice,
(2) the largest transverse diameter of the small hematoma less than 10 mm, and (3)
the minimal distance from the small hematoma to the main hematoma from 1 to 20 mm.[1] When satellite sign was questionable on axial images, coronal and sagittal reformatted
images were also evaluated to obtain a more precise result. To check the accuracy
of the determination of a satellite sign, two radiologist observers (D.S. and W.P.)
independently evaluated each case. Any disagreement between the two reviewers was
resolved by consensus.
Fig. 4 Example of positive satellite sign on non-contrast computed tomography (NCCT) in
patient with intracerebral hemorrhage that had hematoma expansion on the follow-up
NCCT on day 5.
Statistical Analysis
Baseline characteristics were compared between patients with and without HE. Normally
distributed continuous data are presented as mean and standard deviation and were
compared using t-test. Nonnormally distributed data are presented as median and interquartile range
(IQR) and were compared using Mann–Whitney U test. Categorical values were compared using chi-square analysis and are shown at
number and proportion. Multivariable logistic regression was performed to identify
factors independently associated with HE. The results of that analysis are shown as
adjusted odds ratio and 95% confidence interval. Inter-reader reliability for the
identification of satellite sign was analyzed using κ values. A p-value of less than 0.05 indicates statistical significance. All statistical analyses
were performed using SPSS 22.0 and Excel 2013.
Results
Patient Characteristics
There were 1,367 patients that were diagnosed with ICH during the study period. Of
those, 229 patients had ischemic stroke with hemorrhagic transformation, 123 patients
had isolated extra-axial hemorrhage, and 160 patients had secondary causes. The remaining
855 patients with a diagnosis of primary ICH were screened for inclusion eligibility.
Of those, 713 patients were excluded for the following reasons: no baseline cranial
CT within 12 hours of symptom onset (n = 65), no follow-up CT within 168 hours after the baseline CT (n = 397), hematoma evacuation before the follow-up CT (n = 231), and initial hematoma volume less than 2 mL, which is the minimum volume that
can be measured by the image analysis software (n = 20). The remaining 142 patients were included in this study.
The characteristics of the study cohort are shown in [Table 2]. The mean age of the patients was 62.1 ± 13.7 years, and 63.4% were men. The median
time from onset to baseline CT was 3 hours (range: 0–12). The median time from follow-up
to baseline CT was 44.5 hours (range: 3–168). The median baseline hematoma volume
was 16.7 mL (range: 2.1–79.5).
Table 2
Characteristics of the entire study cohort, and compared between patients without
and with hematoma expansion (HE)
|
Characteristics
|
Total (n = 142)
|
Without HE (n = 97)
|
With HE (n = 45)
|
p-Value
|
|
Age (y), mean (SD)
|
62.1
|
(13.7)
|
62.9
|
(12.6)
|
60.40
|
(15.8)
|
0.315
|
|
Male gender, n (%)
|
90
|
(63.4)
|
63
|
(64.9)
|
27
|
(60.0)
|
0.569
|
|
Hypertension, n (%)
|
114
|
(80.3)
|
79
|
(81.4)
|
35
|
(77.8)
|
0.609
|
|
Chronic kidney disease,[a]
n (%)
|
8
|
(5.6)
|
4
|
(4.1)
|
4
|
(8.9)
|
0.252
|
|
Antiplatelet use, n (%)
|
27
|
(19.0)
|
17
|
(17.5)
|
10
|
(22.2)
|
0.507
|
|
SBP (mm Hg), mean (SD)
|
170.9
|
(35.5)
|
173.3
|
(33.7)
|
165.9
|
(38.9)
|
0.253
|
|
DBP (mm Hg), mean (SD)
|
96.4
|
(20.1)
|
97.9
|
(19.0)
|
93.1
|
(22.1)
|
0.186
|
|
GCS score, mean (SD)
|
13.0
|
(3.0)
|
13.0
|
(3.0)
|
13.0
|
(3.0)
|
0.947
|
|
Platelet count (× 109/L), mean (SD)
|
234.2
|
(87.7)
|
241.3
|
(82.3)
|
218.7
|
(97.7)
|
0.154
|
|
PT (s), median (IQR)
|
11.9
|
(9.9–76.7)
|
11.9
|
(9.9–59.0)
|
12.0
|
(10.3–76.7)
|
0.440
|
|
aPTT (s), mean (SD)
|
26.0
|
(7.8)
|
24.6
|
(4.9)
|
29.2
|
(11.3)
|
0.001
|
|
Creatinine (mg/dL), median (IQR)
|
0.9
|
(0.4–12.1)
|
0.8
|
(0.4–8.6)
|
0.9
|
(0.4–12.1)
|
0.202
|
|
Onset time[b] (h), median (IQR)
|
3.0
|
(1–12)
|
3.0
|
(1–12)
|
2.5
|
(1–11)
|
0.343
|
|
Follow-up time[c] (h), median (IQR)
|
44.5
|
(3–168)
|
60.0
|
(4–168)
|
22.0
|
(3–168)
|
0.001
|
|
Baseline hematoma volume (mL), median (IQR)
|
16.7
|
(2.1–79.5)
|
11.7
|
(2.1–79.5)
|
27.4
|
(2.9–59.3)
|
0.001
|
|
Hematoma location, n (%)
|
0.019
|
|
Lobar
|
36
|
(25.4)
|
19
|
(19.6)
|
17
|
(37.8)
|
|
Deep
|
94
|
(66.2)
|
68
|
(70.1)
|
26
|
(57.8)
|
|
Brainstem
|
2
|
(1.4)
|
1
|
(1.0)
|
1
|
(2.2)
|
|
Cerebellum
|
10
|
(7.0)
|
9
|
(9.3)
|
1
|
(2.2)
|
|
Lobar (vs. non-lobar)
|
36
|
(25.4)
|
19
|
(19.6)
|
17
|
(37.8)
|
0.020
|
|
Satellite sign, n (%)
|
62
|
(43.7)
|
36
|
(37.1)
|
26
|
(57.8)
|
0.021
|
|
Intraventricular hemorrhage, n (%)
|
65
|
(45.8)
|
50
|
(51.5)
|
15
|
(33.3)
|
0.043
|
Abbreviations: aPTT, activated partial thromboplastin time; CT, computed tomography;
DBP, diastolic blood pressure; GCS, Glasgow Coma Scale; IQR, interquartile range;
SBP, systolic blood pressure; SD, standard deviation; PT, prothrombin time.
Note: A p-value of < 0.05 indicates statistical significance, as highlighted in bold values.
a Chronic kidney disease was defined as estimated glomerular filtration rate < 60 mL/min.
b Time from onset to baseline CT.
c Time from baseline CT to follow-up CT.
Hematoma Expansion
[Table 2] lists the results of univariate comparisons between patients with and without HE.
Patients with HE had significantly higher mean aPTT (29.2 ± 11.3 vs. 24.6 ± 4.9, respectively;
p = 0.001) and median baseline hematoma volume (27.4 [IQR: 2.9–59.3] vs. 11.7 mL [IQR:
2.1–79.5]; p = 0.001). Lobar location was significantly more common in patients with HE (37.8
vs. 19.6%; p = 0.02). Time from baseline CT to follow-up CT was significantly shorter in patients
with HE (22 vs. 60 hours; p = 0.001). ICH patients with IVH had significantly less HE than patients without IVH
(p = 0.043).
Satellite Sign
The prevalence of satellite sign in this study was 43.7%. The prevalence of satellite
sign was significantly higher in patients with HE than in those without HE (57.8%,
p = 0.021). Overall agreement for identifying a satellite sign between the two readers
in this study was very good (Cohen's kappa: 0.81). The results of analysis for association
between satellite sign and other factors are shown in [Table 3]. Patients with satellite sign had significantly higher aPTT (p = 0.001) and baseline volume (31.3 mL, p = 0.001). Time from baseline CT to follow-up CT was significantly shorter in patients
with satellite sign (p = 0.001). Patients with IVH had a significantly lower prevalence of satellite sign
than patients without IVH (p = 0.001). Lobar location was significantly more common in patients with satellite
sign (p = 0.001). Multivariate analysis revealed satellite sign to be independently associated
with HE ([Table 4]).
The sensitivity, specificity, and accuracy of satellite sign for predicting HE was
57.8, 62.9, and 61.3%, respectively. From image analysis software, the cutoff of more
than 9% relative growth in hematoma volume on follow-up NCCT showed the strongest
association with satellite sign (p = 0.024), with a sensitivity of 55%, specificity of 64.6%, and accuracy of 60.5%.
However, satellite sign was not statistically significant associated with SHE (p = 0.941).
Table 3
Characteristics of the entire study cohort, and compared between patients without
and with satellite sign
|
Characteristics
|
Total
(n = 142)
|
Without satellite sign
(n = 80)
|
With satellite sign
(n = 62)
|
p-Value
|
|
Age (y), mean (SD)
|
62.1
|
(13.7)
|
60.6
|
(13.0–58.0)
|
64.0
|
(14.5)
|
0.315
|
|
Male gender, n (%)
|
90
|
(63.4)
|
56
|
(70.0)
|
34
|
(54.8)
|
0.063
|
|
Hypertension, n (%)
|
114
|
(80.3)
|
67
|
(83.8)
|
47
|
(25.8)
|
0.238
|
|
Chronic kidney disease,[a]
n (%)
|
8
|
(5.6)
|
3
|
(3.8)
|
5
|
(8.1)
|
0.269
|
|
Antiplatelet use, n (%)
|
27
|
(19.0)
|
13
|
(16.3)
|
14
|
(22.6)
|
0.340
|
|
SBP (mm Hg), mean (SD)
|
170.9
|
(35.5)
|
171.9
|
(33.7)
|
169.9
|
(37.9)
|
0.253
|
|
DBP (mm Hg), mean (SD)
|
96.4
|
(20.1)
|
98.9
|
(19.1)
|
93.2
|
(21.0)
|
0.186
|
|
GCS score, mean (SD)
|
13.0
|
(3.0)
|
13.0
|
(3.0)
|
13.0
|
(3.0)
|
0.947
|
|
Platelet count (×109/L), mean (SD)
|
234.1
|
(87.7)
|
237.0
|
(82.5)
|
230.2
|
(94.6)
|
0.154
|
|
PT (s), median (IQR)
|
11.9
|
(9.9–76.7)
|
11.8
|
(10.3–76.7)
|
12.1
|
(9.9–66.5)
|
0.440
|
|
aPTT (s), mean (SD)
|
26.0
|
(7.8)
|
25.1
|
(5.8)
|
27.2
|
(9.8)
|
0.001
|
|
Creatinine (mg/dL), median (IQR)
|
0.9
|
(0.4–12.1)
|
0.9
|
(0.4–8.6)
|
0.9
|
(0.4–12.1)
|
0.202
|
|
Onset time[b] (h), median (IQR)
|
3.0
|
(1–12)
|
3.0
|
(1–12)
|
3.0
|
(1–12)
|
0.343
|
|
Follow-up time[c] (h), median (IQR)
|
44.5
|
(3–168)
|
51.5
|
(3–168)
|
39.0
|
(4–168)
|
0.001
|
|
Baseline hematoma volume (mL), median (IQR)
|
16.7
|
(2.1–79.5)
|
9.9
|
(2.1–57.4)
|
31.3
|
(3.4–79.5)
|
0.001
|
|
Hematoma location, n (%)
|
0.001
|
|
Lobar
|
36
|
(25.4)
|
12
|
(15.0)
|
24
|
(38.7)
|
|
Deep
|
94
|
(66.2)
|
58
|
(72.5)
|
36
|
(58.1)
|
|
Brainstem
|
2
|
(1.4)
|
2
|
(2.5)
|
0
|
(0.0)
|
|
Cerebellum
|
10
|
(7.0)
|
8
|
(10.0)
|
2
|
(3.2)
|
|
Lobar (vs. non-lobar)
|
36
|
(25.4)
|
12
|
(15.0)
|
24
|
(38.7)
|
0.001
|
|
Intraventricular hemorrhage, n (%)
|
65
|
(45.8)
|
46
|
(57.5)
|
19
|
(30.6)
|
0.001
|
Abbreviations: aPTT, activated partial thromboplastin time; CT, computed tomography;
DBP, diastolic blood pressure; GCS, Glasgow Coma Scale; IQR, interquartile range;
PT, prothrombin time; SBP, systolic blood pressure; SD, standard deviation.
Note: A p-value of < 0.05 indicates statistical significance, as highlighted in bold values.
a Chronic kidney disease was defined as estimated glomerular filtration rate < 60 mL/min.
b Time from onset to baseline CT.
c Time from baseline CT to follow-up CT.
Table 4
Multivariate logistic regression model for factors that predict hematoma expansion
|
Factors
|
OR
|
95% CI
|
p-Value
|
|
Lower
|
Upper
|
|
Activated partial thromboplastin time (aPTT)
|
1.078
|
1.000
|
1.163
|
0.050
|
|
Lobar hematoma location
|
1.525
|
0.600
|
3.876
|
0.375
|
|
Follow-up time[a] for every 1-h increase
|
0.991
|
0.982
|
0.999
|
0.032
|
|
Baseline hematoma volume for every 1-mL increase
|
1.013
|
0.986
|
1.040
|
0.349
|
|
Satellite sign
|
2.495
|
1.162
|
5.354
|
0.019
|
|
Intraventricular hemorrhage
|
0.785
|
0.338
|
1.822
|
0.573
|
Abbreviations: CI, confidence interval; CT, computed tomography; OR, odds ratio.
Note: A p-value of < 0.05 indicates statistical significance, as highlighted in bold values.
a Time from baseline CT to follow-up CT.
Discussion
HE is the most important modifiable predictor that can change the outcome of ICH.[3] Our population had a mean age of 62.1 years, and 63.4% were men. Univariate analysis
revealed high aPTT, short follow-up time, large initial hematoma volume, and absence
of IVH to be significantly associated with HE. However, multivariate analysis showed
only follow-up time and satellite sign to be independent predictors of HE (this will
be discussed in more detail below). Other factors, such as time from onset to baseline
CT, hypertension, chronic kidney disease, antiplatelet use, BP on admission, platelet
count, PT, and creatinine, were not found to be significantly associated with HE in
this study. These results are similar to many previous ICH studies.[16]
[17]
[18]
[19]
No generally accepted cutoff for change in hematoma volume has yet been established
for HE. In 1997, Brott et al studied 142 ICH patients who underwent a baseline CT
within 3 hours after clinical onset, and that had follow-up CTs at 1 and 20 hours
after baseline CT. They prospectively selected 33% change in hematoma volume as measured
by image analysis to be the cutoff for HE, and they found it to be significantly associated
with early neurological deterioration. Subsequent studies have used this cutoff to
define SHE. Furthermore, they found that most ongoing bleeding occurred within 3 to
4 hours after hemorrhage onset, and they found this to also be significantly associated
with early neurological deterioration.[20] Thus, early identification of patients at risk for HE is crucial. Many NCCT predictors
have been proposed, but they are still controversial, and they have overlapping definitions.[8]
[9]
[10]
[11]
[12]
In 2017, Shimoda et al coined the term “satellite sign” to describe a new NCCT marker,
which was defined as a small hematoma completely separate from the main hematoma on
at least one slice. They retrospectively reviewed 241 ICH patients who underwent NCCT
within 12 hours after clinical onset, and they found the satellite sign to be associated
with a significantly worse functional outcome in ICH patients.[13]
Later, Yu et al studied 153 ICH patients who underwent a baseline CT within 6 hours
after clinical onset and a follow-up CT within 24 hours after baseline CT. They used
the ABC/2 method to calculate the hematoma volume, and they defined absolute growth
greater than 12.5 ml or relative growth greater than 33% in hematoma size on follow-up
CT as HE, which occurred in 37 patients or 24% of all study patients. The sensitivity,
specificity, positive predictive value, and negative predictive value of satellite
sign for HE prediction in their study was 59.46, 68.97, 37.93, and 84.21%, respectively.
They reported the satellite sign to be an independent predictor of spontaneous ICH,
and they suggested satellite sign as being an acceptable predictor of HE when CTA
is unavailable.[14]
The clear definition, the ease of use, and the potential for predicting HE reported
in the previous study of satellite sign compelled us to investigate this NCCT sign
among our ICH population. None has been done before. The very good agreement for determining
a satellite sign between the two readers in our study (Cohen's kappa: 0.81) confirmed
that satellite sign has clear definition and is practical for use in routine clinical
practice. In 142 patients, we found satellite sign to be significantly associated
with HE, but not significantly associated with SHE (defined as absolute growth > 12.5 ml
or relative growth > 33% in hematoma size). So, the cutoff of HE is the key factor
that influences differences in results among studies. Our patients had NCCT within
12 hours from onset, which is the same timing that was used in the Shimoda et al study.
However, in our country, there is no definite guideline for optimal timing of follow-up
CT. Most ICH patients with conservative treatment will undergo follow-up CT when they
develop clinical deterioration or at least within 7 days after the initial CT. Accordingly,
we included patients with follow-up CT within 7 days in our study since it reflects
real-life clinical practice in the study. This timing was markedly different from
many previous ICH studies, especially from a study by Brott et al that first defined
SHE in ICH patients having 1- and 20-hour follow-up CTs.
In our study, the median time from baseline CT to follow-up CT was 46 hours (range
3–168 hours). Fifty-three patients (37%) had follow-up CT within 24 hours, and 43.3%
of those patients had HE. The other 89 patients (63% of all patients) had a 22.6%
prevalence of HE. These data confirm the finding of several previous studies in ICH
that ongoing bleed occurred during the first several hours.[16]
[17]
[18]
[19] However, there were 20 patients (44.4% of patients with HE) who had HE after 24 hours,
and 10 of those (22.2% of patients with HE) had percentage expansion less than 33%.
In addition, our study cohort had less baseline hematoma volume than the cohort of
Yu et al (16.7 vs. 24.7 mL, respectively), and lower baseline hematoma volume normally
expands less and less often than higher baseline hematoma volume. As such, the cutoff
of more than 33% hematoma growth was not the optimal cutoff value in our study.
Qualitative measurement to determine expansion versus non-expansion was initially
and independently performed by two radiologists, after which image analysis software
was used to obtain quantitative results, which is the same quantitative method used
by Brott et al. We decided to use image analysis software because we felt that it
would deliver more precise data and decrease inter-reader variability.
The association between satellite sign and each 1% increase in hematoma volume was
calculated, and we found the cutoff of greater than 9% relative growth in hematoma
volume on follow-up NCCT have the strongest association with satellite sign (p = 0.024). The sensitivity, specificity, and accuracy were 55.0, 64.6, and 60.5%.
The 9% cutoff found in this study is lower than the previously reported cutoff of
33%.
Limitations
This study has some limitations. First, our study's retrospective design made it susceptible
to incomplete or missing data. Second, the fact that our data came from a single center
that is also a national tertiary referral center suggests that our findings may not
be immediately generalizable to all other care settings. Third, we did not investigate
the accuracy of Stroke VCAR in this study. Fourth, our small sample size may have
limited our study's ability to identify all significant differences and correlations.
Fifth, other factors during hospitalization, such as BP control and antithrombotic
drug use, were not included in our analysis. Sixth and last, an insufficient amount
of data prevented us from investigating association between the new HE cutoff value
and clinical outcomes.
Conclusion
Satellite sign, a new NCCT predictor, was found to be significantly associated with
HE. With different context of Thai population, HE was found in smaller baseline hematoma
volume. Satellite sign was found more common in lobar hematoma (p = 0.001) and less common in patients with IVH (p = 0.001). Quantitative measurement by image analysis software revealed HE greater
than 9%, different from standard greater than 33%, to have the strongest association
with satellite sign. Using software analysis might better help detection HE. Further
studies to validate satellite sign for predicting HE and to identify an optimal cutoff
in our population that is correlated with clinical outcomes are warranted.
