Abstract 12587: An Improved Approach to Cardiac Volumetrics: A Cardiac MRI Technique to Further Augment Accuracy

Abstract

Introduction: In cardiac MRI (CMR), heart volumes are traditionally measured using contouring methods applied to contiguous image data. Herein, we introduce a new approach, Automatic Threshold and Manual Trimming (ATMT), which is applied to the same contiguous data set. Potentially, the ATMT method can be applied by seed/region-growing algorithms with minimal user supervision. We sought to establish its clinical validity. Hypothesis: We hypothesize that the ATMT approach is more accurate as compared to conventional 'gold standard', cardiac contouring. Methods: Hearts from two populations (N=74) were evaluated: explanted heart transplant (Tx) and a clinical validation cohort ( in vivo ). The transplanted hearts were imaged ex vivo using CMR and then weighed on a high-fidelity scale. Cardiac volume/mass was compared (N=54) to the patient cohort (N=20) and measured non-invasively with stroke volume, independently measured via CMR phase velocity technique. Bland-Altman was applied in a 3-way manner for each group. Results: Bland-Altman analysis for Standard Deviation (SD), Bias and Correlation (R) are summarized in Table 1. When compared with independent measurements (weight/flow), ATMT has lower Bias (close to zero) and SD. Further, any comparison involving cardiac contours has a substantially larger bias term and a higher SD. From the table below, ATMT has consistently higher correlations with the independent measurement than does the contour method. Conclusions: Based on multiple comparison metrics with independent measures, the ATMT approach is more accurate and reproducible for quantification of cardiac volume (integral for EF determination) as compared to standard contouring. Furthermore, ATMT accommodates trabeculae and papillary structures more intuitively than the contouring method. This intrinsic accuracy coupled with the potential for more rapid analysis gives a valid impetus to further develop the ATMT approach further increasing CMR accuracy.