Model versus non-model based left ventricular volumetry - A matter of imaging modality or quantification software?

Objective: To compare the modalities 3D-echocardiography (RT3DE) and cardiac magnetic resonance as well as semiautomatic non-model-based and model-based quantification software (SWP-MRI and SWT-MRI) regarding accuracy and agreement of left ventricular functional indices. Methods: 9 asymmetrically shaped gel phantoms (range: 20-350ml), 24 healthy children (age=11.4±3.3y) and 11 patients with abnormally shaped left ventricles (22.0±17.0y) were prospectively investigated. 3D-echocardiography was performed using a Vivid 7 ultrasound machine (V3 matrix transducer); postprocessing was done with a model-based analysis strategy (SWT-echo). CMR datasets were obtained using a multi-slice multi-phase steady‑state-free-precision acquisition (TR/TE/flip=2.8msecs/1.4msecs/60°) with a 1.5T MR system. Volume quantification was done using the same model-based software for CMR as well as non model-based software based on the summation of discs method. Agreement of EDV, ESV and EF between SWT-echo, SWP-mri vs. SWT-mri was determined by Bland Altman analysis. Results: Phantom study revealed high accuracy (<1%) for SWT-echo and SWP-mri as well as a moderate underestimation for SWT‑mri (13%). Agreement between SWP-mri and SWT-echo was superior in volunteers [mean; limits-of- agreement: EDV(5.3%; -20.1 to 30.8%), ESV(-1.3%; -41.6 to 38.9%), EF(4.0%; -12.0 to 19.9%)] with only slight underestimation by  RT3DE in patients [EDV(11.5%; ‑18.5 to 41.4%), ESV(13.0%; -5.4 to 31.5%), EF(‑6.9%; -49.9 to 36.1%)]. Comparing SWT-echo with SWT-mri revealed volume underestimation of EDV (9.8; -20.5 to 40.0%) and overestimation of ESV (-9.6; -60.1 to 41.0%) in volunteers by SWT-mri resulting in underestimation of EF (12.6;-9.6 to 34.9). In patients minor differences between SWT-echo and SWT-mri were observed [EDV (0.6%; -28.2 to 29.4%), ESV (-2.4%; -38.2 to 33.4%), EF(9.3%; -35.7 to 54.3%)]. Compared to our reference SWP-mri both model-based techniques moderately underestimated EDV (SWT-MRI 12.1%; ‑2.1 to 26.4%, SWT-echo 11.5%; -18.5 to 41.4%) and ESV (SWT-mri 10.6%; -21.2 to 42.4%, SWT-echo 13.0%; -5.4 to 31.5%)  resulting in quite precise  EF(SWT-MRI  2.4%, -23.7 to 28.5%). Conclusion: Accuracy and reliability of left ventricular indices are excellent for RT3DE assessed by model based approach compared to non-model-based CMR approach in phantoms and healthy volunteers with minor volume underestimation in atypically shaped moving ventricles. Minor agreement was present if the model-based CMR software was used for determination of ventricular volumes.