1.
Patient information:
n = 36 SA slices from bright-blood T2* scans of 21 patients (9 haematology [median: 30.11 ms,
range: 9.44-40.71 ms],
12 suspected cardiomyopathy [median: 33.04 ms,
range: 24.18-43.33 ms]).
2.
Highlighted MGE MRI sequence in the study:
Single breath hold,
1.5 T,
8 TE (2.59-18.20 ms,
2.23 ms increment).
3.
Observers who drew myocardium contour:
2 observers with >2 years experience
2 observers with < 1 year experience but with profound knowledge of short axis myocardium anatomy
4.
SA image selection to segment epi- and endomyocardium contour:
Subjective selection of one image (blind,
independent selection by each observer) from a series of short axis MGE of left ventricle (LV) myocardium (Fig.
2).
CNR-optimized composite image by blending three images,
with the same weight factor,
acquired from a series of short axis MGE (Fig.
2) that show the highest CNR [2] between LV myocardium with LV blood pool (LVBP)[3],
right ventricle blood pool (RVBP),
and lung (Fig.
3).
5.
Reproducibility evaluation
Four observers performed hand tracing contour drawing (independently) of short axis LV epi- and endocardial on reference image per slice acquired by using two types of image described in point 4.
The time delay for contour drawing between the two types of image for one observer was more than two week.
The same Software platform,
image quality and screen setting were used in the study.
6.
Statistical analysis of SA image selection:
Inter-observer reproducibility of the epi- and endomyocardium contour by each SA image selection was assessed by Dice Similarity Coefficient (DSC) (0 = no spatial overlap between two sets of binary segmentation,
1 = maximum overlap) and was presented as mean±1SD or median±MAD (median absolute deviation) [3].
Comparison of the two SA image selection methods (total of 36x6=216 DSC matchs) was performed by paired t-test or paired Wilcoxon test (p<0.05 is not significant).
Comparison of myocardial iron T2* calculated from two image selections was assessed by paired Wilcoxon test and results were presented as median±MAD.