Automation of gated tomographic left ventricular ejection fraction

Area of Science:

• Nuclear Cardiology
• Cardiovascular Imaging
• Medical Physics

Background:

• Gated SPECT (GSPECT) is established for left ventricular (LV) ejection fraction (EF) measurement.
• Improving measurement precision requires automating observer-defined tomogram processing rules.

Purpose of the Study:

• To develop and validate an automated method for calculating LV ejection fraction (EF) using GSPECT.
• To enhance the precision and reproducibility of LVEF measurements from GSPECT data.

Main Methods:

• Automated LV center estimation using activity centroids.
• Definition of end diastole and end systole based on count rate extremes.
• Endocardial border generation using harmonic functions and thresholding.
• Validation against manual GSPECT, equilibrium gated blood pool, and gated first-pass imaging.

Main Results:

• Automated GSPECT EF showed strong correlation (r=0.87) with equilibrium gated blood pool and gated first-pass methods.
• High interobserver (r=0.92) and intraobserver (r=0.94) reproducibility for automated EF.
• Independent observer agreement with automated results ranged from r=0.93 to r=0.96.

Conclusions:

• Automated GSPECT LV EF measurements correlate well with manual GSPECT and other imaging modalities.
• The automated method demonstrates high reproducibility, supporting its clinical utility.
• Automated processing offers a precise and reliable approach to LVEF assessment in GSPECT.