Abstract 1850: Clinical Utility of Automatic Assessment of Left Ventricular Ejection Fraction using Computer Assisted Border Detection Software
Ejection fraction (EF) provides important prognostic and therapeutic information in patients with heart disease but quantification of EF requires planimetry and is time consuming. As a result, visual assessment is frequently used but is subjective and requires extensive experience. New computer softwares to assess EF automatically are now available and could be used routinely in busy digital labs (>15,000 studies/year) and in core labs running large clinical trials. We tested Siemens AutoEF® software to determine whether it correlated with both formal planimetry and visual estimation of EF
Methods: Siemens AutoEF® is based on automated border detection and artificial intelligence. An expert (EX) and a non expert (NEX) (cardiology fellow) assessed EF visually by reviewing transthoracic echograms from consecutive patients. An experienced sonographer quantified EF on all studies using Simpson’s method of disks. The readers were blinded to each other, AutoEF and EF by planimetry. Wall motion abnormalities, LV dilation, LV hypertrophy, and mitral disease were noted
Results: Ninety nine echograms were analyzed. Visual assessment by EX(R=0.87) and NEX(R=0.80) correlated more closely with quantified Simpson’s method than did Auto EF (R=0.69). The correlation between AutoEF and Simpson’s was better in normal hearts (R=0.72) but was weaker in the presence of LVH (R=0.43), LV dilatation (R=0.47) and mitral calcification (R=0.5). However, the correlation between EX, NEX visual assessment of EF and quantitated EF by Simpson’s method remained strong in all 3 conditions mentioned above. The correlations between AutoEF and visually assessed EF by EX and NEX were only modest (R=0.64 and R=0.60 respectively).
Conclusion: The discrepancies in EF estimates between AutoEF and Simpson’s method preclude use of AutoEF on a routine basis. Visual assessment of EF, with its strong correlation with quantitative EF, underscores its continued clinical utility.