Session SA1.3

Quantification of Myocardial Oedema and Necrosis during Acute Myocardial Infarct

N Baron, N Kachenoura*, F Beygui, P Cluzel,
P Grenier, A Herment, F Frouin

INSERM
Paris, France

Quantification of myocardial oedema and necrosis during acute infarct (MI) is crucial for the estimation of the functional recovery of the left ventricle (LV) and for patient’s prognosis. The aim of this study was to quantitatively evaluate these two parameters from MRI acquisitions, using a technique based on a fuzzy c-means algorithm.
Methods: Late Gadolinium Enhancement (LGE) sequences and T2 weighted Short Tau Inversion Recovery (STIR)-Black-Blood were acquired in 22 patients hospitalized for an acute MI. To estimate the necrosis volume, a classification method based on a fuzzy c-means algorithm, was applied to the region including the myocardium and the LV cavity. Thereafter, a threshold of the degree of membership to the enhanced class was automatically defined and the volume of the MI was estimated. For the estimation of oedema volume from STIR images, a similar approach restricted to the myocardium was proposed. Results were compared to a manual delimitation of the necrosis and to a semi-automatic thresholding currently reported in the literature: a threshold was defined from a region of interest (ROI) placed in the healthy tissue, as: [mean intensity value (m)] +n*[standard deviation (SD) within the ROI], with n= 2 or 3. All these techniques including ours require a manual delineation of the myocardial borders.
Results: The proposed estimation of the necrosis volume from LGE images was strongly correlated with the semi automatic method, using either the threshold m+2SD (r=0.98, p<10-4) or m+3SD (r=0.93, p<10-4), with a reduced bias (-4.6% and 0.1%). Importantly, it is also highly correlated with the manual delineation of the necrosis (r= 0.90, p<10-4). For the quantification of the oedema volume, our approach was highly correlated with the semi automatic approach using a threshold of m+2SD (r=0.99, p<10-4).
Conclusion: An approach which allows an objective estimation of necrosis and oedema volumes is provided and shown to be strongly correlated to the conventional methods, which are either manual or semi automatic. Combined with an automated detection of the endocardial and epicardial borders, our method may prove clinically useful.

(Abstract Control Number: 268)