Abstract:
In diagnosis of cardio-vascular patients, analysis of Cardiac Magnetic
Resonance (CMR) images provides the useful functional parameters of the left
ventricle (LV). The gold standard method such as expert technician outlining is time
consuming and recent methods such as mathematical approximation are un-reliable.
This study increases the performance of diagnosis by constructing an automatic
segmentation algorithm. Performance testing employs Repeated-Measure ANOVA
and Bland-Altman analysis for examining the algorithm with short axis CMR images.
The participant studied comprised a group of 10 patients and a group of 10 healthy
subjects. Statistical results display the well agreement in measurement of Ejection
Fraction (EF) by an experience observer and computer (p = 1.000, bias 0.1 %).
Computer detection based on the proposed algorithm was also effective in measuring
EDV (p = 0.125, bias 3.9 cc.) in the patient group. In the healthy subject group, the
computer is well agreeing with the expert observer in measurement of ESV (p =
1.000, bias -0.1 cc.) and myocardium mass (p = 0.187, bias -2.9 g). In conclusion,
computer detection can be reliably used in the assessment of LV functional
parameters.