Blood-threshold CMR volume analysis of functional univentricular heart

Francesco Secchi, Marco Alì, Marcello Petrini, Francesca Romana Pluchinotta, Andrea Cozzi, Mario Carminati, Francesco Sardanelli

Research output: Contribution to journalArticlepeer-review


Purpose To validate a blood-threshold (BT) segmentation software for cardiac magnetic resonance (CMR) cine images in patients with functional univentricular heart (FUH). Materials and methods We evaluated retrospectively 44 FUH patients aged 25 ± 8 years (mean ± standard deviation). For each patient, the epicardial contour of the single ventricle was manually segmented on cine images by two readers and an automated BT algorithm was independently applied to calculate end-diastolic volume (EDV), end-systolic volume (ESV), stroke volume (SV), ejection fraction (EF), and cardiac mass (CM). Aortic flow analysis (AFA) was performed on throughplane images to obtain forward volumes and used as a benchmark. Reproducibility was tested in a subgroup of 24 randomly selected patients. Wilcoxon, Spearman, and Bland-Altman statistics were used. Results No significant difference was found between SV (median 57.7 ml; interquartile range 47.9-75.6) and aortic forward flow (57.4 ml; 48.9-80.4) (p = 0.123), with a high correlation (r = 0.789, p < 0.001). Intra-reader reproducibility was 86% for SV segmentation, and 96% for AFA. Inter-reader reproducibility was 85 and 96%, respectively. Conclusion The BT segmentation provided an accurate and reproducible assessment of heart function in FUH patients.

Original languageEnglish
Pages (from-to)331-337
Number of pages7
JournalRadiologia Medica
Issue number5
Publication statusPublished - Jan 22 2018


  • Cardiac magnetic resonance
  • Congenital heart disease
  • Image segmentation
  • Univentricular heart
  • Ventricular function

ASJC Scopus subject areas

  • Radiology Nuclear Medicine and imaging


Dive into the research topics of 'Blood-threshold CMR volume analysis of functional univentricular heart'. Together they form a unique fingerprint.

Cite this