John David Biglands1, Abdulghani Larghat2,
Sven Plein2, Derek R. Magee3, Roger D. Boyle3,
Aleksandra Radjenovic4,5
1Division of Medical
Physics, University of Leeds, LEEDS, Yorkshire, United Kingdom; 2Division
of Cardiovascular & Neuronal Remodelling, University of Leeds, LEEDS,
Yorkshire, United Kingdom; 3School of Computing, University of Leeds,
LEEDS, Yorkshire, United Kingdom; 4School of Medicine, University
of Leeds, LEEDS, Yorkshire, United Kingdom; 5Academic Section of
Musculoskeletal Disease, University of Leeds, LEEDS, Yorkshire, United
Kingdom
Estimating myocardial blood flow (MBF) from DCE-MRI datasets requires contours to be drawn outlining the myocardium. Manual contouring is time consuming and numerous automated contouring algorithms have been developed. A variety of metrics are employed to evaluate such algorithms in terms of either geometric contour errors or errors in the estimated MBF. This study aimed to establish the relationship between a selection of geometric distance metrics and MBF error and found that these measures did not correlate. Future automated myocardial contouring algorithms should be evaluated in terms of both geometric distance metrics and MBF to enable objective cross-comparison.
Keywords