Comparing knowledge-based 3D Reconstruction to Conventional MRI Algorithms Measuring Left Cardiac Chamber Volumes in Pediatrics

Objectives: Compare accuracy of 3D knowledge-based reconstruction (3D KBR) algorithm to standard measurement of left atrial (LA) and left ventricle (LV) volumes. Background: Accurate measurement of LV and LA volume is essential in assessing cardiac function. Cardiac magnetic resonance imaging (CMR) is the gold standard, but analysis is relatively time consuming. Our study compared analysis time and agreement of 3D KBR algorithm to conventional CMR. Methods: CMR studies of children aged 3-17 years with iron-overload were analyzed. DiCOM data was uploaded into the 3D KBR software calculated the LA and LV volumes in end systole and diastole, and ejection fraction. LA volumes were calculated using biplane method. LV measurements were calculated using Simpson's method (using a short axis stack) (SAX) technique. These methods were compared using intraclass coefficients (ICC) and Bland-Altman plots. Results: 71 CMR studies of 31 patients were analyzed. No mean bias between SAX and VMS (Ventripoint software) measurement of LV end diastolic volume (EDV), biplane and VMS measurements of LA end systolic volume (ESV) or LA EDV were found. A small positive bias in VMS LV ESV; with moderate agreement in LV EDV, LA ESV and LV ejection fraction (EF)/ LA EF and wider limits of agreement in LV ESV and LA EDV. Excellent correlation between SAX and VMS in measuring LV volumes, biplane and VMS LA volumes. Interobserver agreement for VMS LV and LA volumes were excellent. VMS LV analysis time was 2.43 min and VMS LA analysis time was 1.46 min. Conclusion: 3DKBR offers a time efficient alternative with comparable accuracy to the current LV and LA measurements used in clinical practice.