Operator induced variability in left ventricular measurements with cardiovascular magnetic resonance is improved after training.
Karamitsos TD., Hudsmith LE., Selvanayagam JB., Neubauer S., Francis JM.
BACKGROUND: Accurate and reproducible measurement of left ventricular (LV) mass and function is a significant strength of Cardiovascular Magnetic Resonance (CMR). Reproducibility and accuracy of these measurements is usually reported between experienced operators. However, an increasing number of inexperienced operators are now training in CMR and are involved in post-processing analysis. The aim of the study was to assess the interobserver variability of the manual planimetry of LV contours amongst two experienced and six inexperienced operators before and after a two months training period. METHODS: Ten healthy normal volunteers (5 men, mean age 34+/-14 years) comprised the study population. LV volumes, mass, and ejection fraction were manually evaluated using Argus software (Siemens Medical Solutions, Erlangen, Germany) for each subject, once by the two experienced and twice by the six inexperienced operators. The mean values of experienced operators were considered the reference values. The agreement between operators was evaluated by means of Bland-Altman analysis. Training involved standardized data acquisition, simulated off-line analysis and mentoring. RESULTS: The trainee operators demonstrated improvement in the measurement of all the parameters compared to the experienced operators. The mean ejection fraction variability improved from 7.2% before training to 3.7% after training (p=0.03). The parameter in which the trainees showed the least improvement was LV mass (from 7.7% to 6.7% after training). The basal slice selection and contour definition were the main sources of errors. CONCLUSIONS: An intensive two month training period significantly improved the accuracy of LV functional measurements. Adequate training of new CMR operators is of paramount importance in our aim to maintain the accuracy and high reproducibility of CMR in LV function analysis.