SUPPLEMENTARY DATA
Automatic Measurement Of The Myocardial Interstitium:
Synthetic Extracellular Volume Quantification without Hematocrit Sampling.
Treibel TA et al.
The following data has been provided as supporting material due to limited space in
the main manuscript. Although not essential for the understanding of the paper, the
authors felt that this is of benefit to the reader and will complement the main
manuscript.
1) Overlap with previously published cohorts.
2) Supplementary description of statistical analysis.
3) Supplementary Tables:

Table S1: Synthetic/Conventional ECV correlations with Clinical Parameters.

Table S2: In multivariable modeling, conventional and synthetic ECV remained
associated with HHF, death, or the combined endpoint of HHF/death.
4) Supplementary Figures:

Figure S1: Hematocrit Variability.

Figure S2: Validation cohort - measured vs synthetic Hct correlation and
Bland-Altman comparison.

Figure S3: Outcome cohort - measured vs synthetic Hct correlation and
Bland-Altman comparison.

Figure S4: Local synthetic Hct derivation for Outcome Cohort.

Figure S5: Conventional versus synthetic ECV in the outcome cohort.

Figure S6: Small ECV bias without local Hct:R1 calibration.
4) Supplementary Video: Separate MP4 file.
5) Sequence parameters.
1
1. Overlap with previously published cohorts.
The ECV data of the following cohorts have been published previously:
-
The AS histology cohort was previously published as indicated (White SK et al
JACC Imaging 2013).
-
The outcome cohort has kindly been provided by Dr. Schelbert and is currently
under peer review in JAHA.
-
In the proof-concept-cohort, the ATTR amyloid patients (n=74) were part of a
larger, previously published amyloid cohort (Fontana M et al Radiology 2015).
All other cohorts have not been published previously.
2
2. Supplementary description of statistical analysis
The log-rank test with ECV (categorized arbitrarily in 5% increments that were
considered clinically relevant intervals) and Cox regression (ECV expressed as a
continuous percentage variable) examined associations between ECV and outcomes.
To benchmark ECV against EF, we compared univariable Cox regression chi square
(χ2) values. Initial multivariable Cox regression models constrained the number of
covariates to yield ~10 events per predictor variable and stratified for hospitalization
status and heart failure stage. These disease severity variables do not provide insight
into etiology. Since we were not interested in quantifying their association with
outcomes, their use as stratification variables still permitted risk adjustment for
hospitalization and baseline heart failure stage. We performed further risk adjustment
for key clinical covariates including EF (which governs eligibility for treatments), age
(frailty marker), myocardial infarction size (determines extent of irreversible
myocardial damage), and renal function (important for volume homeostasis) all of
which are important risk markers. We also adjusted for gender since ECV appears
higher in women. All Cox regression models included ECV as a continuous variable,
and expressed hazard ratios for 5% ECV increments which we thought were clinically
meaningful intervals for ease of interpretation. We tested for interactions between
variables by adding terms into the model that were their product. Nonsignificant time
interaction terms for ECV confirmed the proportional hazards assumption. ECV did
not interact with MI size or the presence of nonischemic scar evident on LGE. Yet,
we found a significant interaction between EF and ECV for all outcomes. Thus, we
created additional Cox models stratifying by hospitalization status and clinically
relevant EF categories (EF<30%, 30%≤EF<50%, 50%≤EF) while adjusting for heart
failure stage, age, myocardial infarction size, and renal function.
3
3-1. Supplementary Table S1: Synthetic and Conventional ECV both correlate
well with Clinical Parameters.
ECV Method
Conventional
Sequence
Synthetic
Conventional
ShMOLLI
Synthetic
MOLLI
NT-pro-BNP
0.60
0.60
0.62
0.63
Systolic blood pressure
-0.64
-0.64
-0.62
-0.63
LV ejection fraction
-0.57
-0.61
-0.56
-0.60
Stroke volume indexed
-0.36
-0.38
-0.31
-0.34
Left atrial area indexed
0.39
0.41
0.44
0.44
LV mass index
0.49
0.49
0.52
0.52
Correlation coefficient R, all values significant with a p-value <0.001; ShMOLLI = Shortened
MOdified Look-Locker Inversion recovery; MOLLI = Modified Look-Locker Inversion recovery.
4
3-2. Supplementary Table S2: In multivariable modeling, conventional and synthetic
ECV remained associated with HHF, death, or the combined endpoint of HHF/death.
Hazard ratios and Chi-Square (χ2) for conventional and synthetic ECV are shown unadjusted
(univariable) and in a multivariable model. The multivariable model stratified by heart failure
stage and hospitalization status, and adjusted for LVEF, age, glomerular filtration rate,
myocardial infarction size, and gender.
χ2
Outcome
Cox regression model covariate
ECV (5% increase)
HHF (n=55)
Death (n=74)
HHF or Death (n=111)
p-value
Hazard ratio
(CI 95%)
Conventional – univariable
2.41 (1.93-3.02)
59.6
<0.001
Synthetic – univariable
2.30 (1.87-2.81)
64.5
<0.001
Conventional – multivariable model
1.77 (1.32-2.36)
14.9
<0.001
Synthetic – multivariable model
1.74 (1.32-2.30)
15.2
<0.001
Conventional – univariable
2.13 (1.74-2.61)
53.4
<0.001
Synthetic – univariable
1.90 (1.55-2.31)
39.8
<0.001
Conventional – multivariable model
1.87 (1.45-2.40)
23.9
<0.001
Synthetic – multivariable model
1.70 (1.34-2.16)
18.9
<0.001
Conventional – univariable
2.25 (1.91-2.64)
96.0
<0.001
Synthetic – univariable
2.06 (1.77-2.40)
89.2
<0.001
Conventional – multivariable model
1.85 (1.50-2.27)
33.9
<0.001
Synthetic – multivariable model
1.69 (1.38-2.75)
24.9
<0.001
ECV = extracellular volume fraction; HHF = hospitalization for heart failure; LVEF = left ventricular
ejection fraction.
5
4) Supplemental Figures:
Figure S1: Variability In Repeated Hematocrit Samples.
Repeat sampling variability was tested in 44 patients who underwent two samples a
median of 4 hours apart. Test:retest variability of hematocrit was higher than expected
(n=44, variability 10% with hct:hct R2=0.86, p<0.001).
6
Figure S2: Validation cohort - measured vs synthetic Hct correlation and BlandAltman comparison.
7
Figure S3: Outcome cohort - measured vs synthetic Hct correlation and BlandAltman comparison.
8
Figure S4: Local synthetic Hct derivation for Outcome Cohort
Applying the UK synthetic Hct derivation to the US Outcome cohort, a bias of 2%
was found (see figure S6). Therefore, a local calibration was obtained for the
Pittsburgh cohort.
9
Figure S5: Conventional versus synthetic ECV in the outcome cohort.
In the outcome cohort, conventional and synthetic ECV had similar ranges (16.647.8% and 16.2-50.9%, respectively) with excellent correlation (R2 = 0.82, p<0.001).
10
Figure S6: Small ECV bias without local Hct:R1 calibration.
Applying the UK synthetic Hct derivation to the US Outcome cohort, a bias of 2%
was found.
11
5) Supplementary Video: Separate MP4 file.
This Video is a live screen capture of an inline synthetic ECV tool prototype, showing
the instantaneous ECV map output in a patient with a chronic infarct. The synthetic
ECV mapping module was created as a part of the T1 mapping reconstruction
pipeline to generate automatic synthetic ECV maps inline (“on-the-fly”) as soon as
post contrast T1 maps are available. The additional processing time is less than 1
second per slice for finding paired pre-contrast T1 mapping data, performing image
co-registration, and generating the blood mask and ECV map. The user is able to
draw regions of interest on the scanner console, where pixel values represent
percentage ECV.
6) Sequence parameters
Imaging for the proof-of-concept and histology cohorts was performed on a 1.5 T
Siemens Avanto scanner (Siemens Medical Solutions, Erlangen, Germany; system
software version VB17) using a 32-channel cardiac coil array. Both, ShMOLLI and
MOLLI variants were run for these cohorts. Complete sequence details can be found
on the following pages as detailed below:
a. Pages 7 & 8 – ShMOLLI for Step 1 and 2 (pre- and post-contrast).
b. Pages 9 & 10 – pre-contrast 5s(3s)3s MOLLI used for Step 1 and 2
The pre-contrast MOLLI was a 5s(3s)3s variant, indicating 2 inversions with
acquisition of images for at least 5 seconds, followed by a recovery of at least
3 s, and a second inversion with images acquired for at least 3 s.
c. Pages 11 & 12 – post-contrast 4s(1s)3s(1s)2s MOLLI used for Step 1 and 2
The post-contrast MOLLI was a 4s(1s)3s(1s)2s variant, indicating 3
inversions with acquisition of images for at least 4 seconds, followed by a
recovery of at least 1 second, a second inversion with images acquired for at
least 3 s, followed by a recovery of at least 1 second, and a final acquisition of
2 seconds.
12
Imaging for the Outcome cohort was performed on a 1.5 T Siemens Espree scanner
(Siemens Medical Solutions, Erlangen, Germany) using a 32-channel cardiac coil
array. Complete sequence details for the MOLLI variants run for the outcome cohort
can be found on the following pages as detailed below:
d. Pages 13 & 14 – post-contrast 5(2)2 MOLLI used for Step 3
The pre-contrast MOLLI was a 5(2)2 variant, indicating 2 inversions with
acquisition of images for at least 5 heart beats, followed by a recovery of at
least 2 heart beats, and a second inversion with images acquired for at least 2
heart beats.
e. Pages 15 & 16 – post-contrast 4(1)3(1)2 MOLLI used for Step 3
The post-contrast MOLLI was a 4(1)3(1)2 variant, indicating 3 inversions
with acquisition of images for at least 4 heart beats, followed by a recovery of
at least 1 heart beat, a second inversion with images acquired for at least heart
beats, followed by a recovery of at least 1 heart beat, and a final acquisition of
2 heart beats.
13
14
15
16
17
18
19
20
21
22
23