Balancing Speed and Accuracy in Cardiac Magnetic Resonance Function Post-Processing: Comparing 2 Levels of Automation...

Area of Science:

• Cardiovascular Imaging
• Medical Imaging Analysis
• Computational Cardiology

Background:

• Manual contour-tracing for cardiac function assessment on cardiac magnetic resonance (CMR) short-axis cines is time-consuming and prone to variability.
• Accurate tracing of basal contours remains a significant challenge in automated CMR analysis.
• Existing automated methods show potential but require further refinement for clinical reliability.

Purpose of the Study:

• To compare the speed, reproducibility, and accuracy of three automated cardiac function post-processing software packages (Level 1) against manual assessment.
• To evaluate the efficacy of a standardized manual adjustment protocol combined with software-specific error correction (Level 2) in improving automated measurements.
• To determine if Level 2 automation can achieve accuracy comparable to manual assessments.

Main Methods:

• Three Level 1 automated software packages were used for cardiac function assessment in 65 healthy subjects.
• Manual contour-tracing was performed independently for comparison.
• Level 2 involved manual adjustments to automated basal tracings with standardized error correction protocols.
• Accuracy was defined as a difference less than or equal to the maximum manual inter-observer disagreement (6%).

Main Results:

• Level 1 (2.1 ± 1.0 min) and Level 2 (5.2 ± 1.3 min) automated methods were significantly faster than manual assessment (21.1 ± 2.9 min).
• Level 1 automation exhibited wide limits of agreement compared to manual tracing.
• Level 2 automation, particularly with the most reliable software, demonstrated substantially improved accuracy for key metrics like left ventricular end-diastolic volume (98% vs. 53%) and ejection fraction (98% vs. 60%) compared to Level 1.

Conclusions:

• Level 1 automated cardiac function post-processing is rapid and reproducible but shows variable accuracy.
• Level 2 automation, incorporating manual adjustments and error correction, effectively balances speed with improved accuracy.
• This refined automated approach holds promise for more efficient and reliable clinical cardiac function assessment.