Keywords:
Cardiac, Computer applications, Haematologic, MR, Segmentation, Image registration, Image verification, Haematologic diseases
Authors:
P. Triadyaksa, M. Oudkerk, P. E. Sijens; Groningen/NL
Purpose
In clinical studies,
bright-blood magnetic resonance imaging (MRI) has been adopted for measuring myocardial T2* and for non-invasive longitudinal assessment of myocardial iron deposition [1].
A region of interest (ROI) of myocardium is drawn on a selected single short-axis image from a series of multi-gradient echo (MGE) acquisitions,
propagated through the series and fitted to a T2* equation to acquire myocardial transverse relaxation time [1].
A single short axis image is also selected in commercial software packages for T2* calculation,
for example the first or second echo time (TE) image of the MGE image series [2] or the use of a TE showing good contrast between left ventricle (LV),
blood pool (LVBP) and myocardium [3].
The clarity of myocardial contour depends on contrast difference between the myocardium and its surrounding border (LVBP,
right ventricle blood pool [LVBP],
and lung) [4,
5].
At short TE,
short-axis images show little contrast between LV myocardium and LVBP or RVBP as a characteristic of differential transverse magnetization.
At long TE,
the images show poor contrast between LV myocardium and lung and are subject to signal loss at regions adjacent to the posterior vein of the LV [6].
Therefore,
a single short-axis image can not provide optimal contrast between LV myocardium and all of its surroundings that leads to inaccurate,
poorly reproducible myocardium contours which results to unreliable T2* determinations and,
thus,
unreliable iron contents [7].
Analysis of contours hand-drawn on one visually selected MGE image (figure 1) and subsequently propagated through the rest of the MGE series,
continues to be the standard method.
The similarity of LV epicardial and endocardial contours that are drawn by using the same vs.
different TE of MGE images by observers has,
to our knowledge,
not yet been examined.
The alternative hand-drawn contouring proposed in our study is to use a contrast to noise ratio (CNR)-optimized image composed from three short-axis MGE images providing maximum CNR between LV myocardium and its surrounding border areas (left ventricular pool,
right ventricular pool,
lung).
Therefore,
the purpose of this study is to improve inter-observer reproducibility of myocardial T2* assessment in MGE bright blood magnetic resonance imaging using standard CNR-optimized composite images.