1082From Tuscan Trabeculae to Florentine Fractals – A Novel Approach to Quantification by CMR

Abstract

Background: Several mechanisms have been implicated in abnormal trabecular development leading on to left ventricular noncompaction(LVNC).We need a systematic yardstick, capable of assessing trabeculae in embryonic mice used in the study of cardiogenesis and also in human hearts imaged clinically by cardiovascular magnetic resonance(CMR). Because trabeculae are complex and irregular, a fractal approach for their quantification would have merit. Objectives: To: A) Investigate trabecular changes in the embryonic mouse heart using high resolution episcopicmicroscopy(HREM) and fractal geometry; B) Apply fractal analysis to CMR as a means for describing human trabeculae. Methods: Fractal analysis was performed on embryonic mouse hearts (1500slices,n = 12 per stage) starting from the time of cardiac septation till after completion of compaction (E14.5-18.5). Image processing of HREM data(Fig.1) permitted calculation of the fractal dimension(FD) for left and right ventricles(LV,RV). The influence of different slicing planes( 0,10,20degrees) on FD was evaluated. A similar technique was developed for use with human CMR datasets. The analysis was applied to short-axis cine volumetric studies in 105 healthy volunteers (self-defined white/black) and 30 LVNC cases. Using an in-house macro, the fractal method was compared to perimetry as another semiautomated measure of trabecular complexity. Intra-observer(n = 60), inter-observer(n = 60) and interstudy reproducibility of fractal analysis(n = 24) was examined. Impact of varying slice thickness(8,7,6mm;n = 50) on FD was also tested. Results: In mouse, FD was higher for RV than LV, did not change significantly following a 10degree reslice, and within the LV,FD varied from base to apex and fell with development (Fig.2)(E14.5 to 18.5,p < 0.0001). Human CMR analysis showed differences between LVNC and healthy volunteers, especially in the apical third(p < 0.00001;for LVNC:1.392 + /-0.010)(Fig.3). Maximal apical FD had the highest accuracy in predicting LVNC. FD showed minimal variability between observers(intra/inter-observer intraclass correlation coefficient[ICC]:0.97,0.96), between repeat studies(ICC:0.94) and between slice thicknesses. Conclusions: These preliminary insights from the fractal characterisation of human and embryonic murine hearts suggest interesting, previously unquantifiable patterns of trabecular complexity. This biological signal could provide a clinically useful and robust metric for improved trabecular quantification by CMR.