To evaluate the influence of ECA in automatic myocardial edge detection under normal conditions and with abnormal perfusion and also evaluate the reproducibility of semi-automatic processing.
A total of 100 MPSs, 50 with ECA, were analyzed. Each subgroup included 25 cases with perfusion abnormalities. The cases were processed automatically and by 4 independent operators with different levels of experience. Commercial QGS and QPS softwares were used with tools to mask and relocate the left ventricle area. Functional parameters (final diastolic and systolic volumes and ejection fraction) and perfusion parameters such as the reversibility perfusion score and rest perfusion defect extension were analyzed. The data were compared with Pearson's correlation and Student's test.
Interobserver correlation significantly worsened with the presence of ECA and was moderately affected by perfusion abnormalities. More experienced observers presented better correlation. Reproducibility was greater for the functional perfusion parameters, independently of the observer's experience.
ECA significantly affects automatic edging delimitation, affecting the MPS values. Interobserver reproducibility with manual processing was more altered regarding funtional parameters than in the perfusion scores. Perfusion abnormalities did not interfere with software reproducibility, and when present, better correlation was found. If ECA is not present, manual intervention should be avoided.