Community delivery of semiautomated fractal analysis tool in cardiac mr for trabecular phenotyping

Gabriella Captur, Dina Radenkovic, Chunming Li, Yu Liu, Nay Aung, Filip Zemrak, Catalina Tobon-Gomez, Xuexin Gao, Perry M. Elliott, Steffen E. Petersen, David A. Bluemke, Matthias G. Friedrich, James C. Moon

Research output: Contribution to journalArticlepeer-review

10 Scopus citations


Purpose: To report the development of easy-to-use magnetic resonance imaging (MRI) fractal tools deployed on platforms accessible to all. The trabeculae of the left ventricle vary in health and disease but their measurement is difficult. Fractal analysis of cardiac MR images can measure trabecular complexity as a fractal dimension (FD). Materials and Methods: This Health Insurance Portability and Accountability Act (HIPAA)-compliant study was approved by the local Institutional Review Board. Participants provided written informed consent. The original MatLab implementation (region-based level set segmentation and box-counting algorithm) was recoded for two platforms (OsiriX and a clinical MR reporting platform [cvi42, Circle Cardiovascular Imaging, Calgary, Canada]). For validation, 100 subjects were scanned at 1.5T and 20 imaged twice for interstudy reproducibility. Cines were analyzed by the three tools and FD variability determined. Manual trabecular delineation by an expert reader (R1) provided ground truth contours for validation of segmentation accuracy by point-to-curve (P2C) distance estimates. Manual delineation was repeated by R1 and a second reader (R2) on 15 cases for intra/interobserver variability. Results: FD by OsiriX and the clinical MR reporting platform showed high correlation with MatLab values (correlation coefficients: 0.96 [95% CI: 0.95–0.97] and 0.96 [0.95–0.96]) and high interstudy and intraplatform reproducibility. Semiautomated contours in OsiriX and the clinical MR reporting platform were highly correlated with ground truth contours evidenced by low P2C errors: 0.882 ± 0.76 mm and 0.709 ± 0.617 mm. Validity of ground truth contours was inferred from low P2C errors between readers (R1-R1: 0.798 ± 0.718 mm; R1-R2: 0.804 ± 0.649 mm). Conclusion: This set of accessible fractal tools that measure trabeculation in the heart have been validated and released to the cardiac MR community ( to encourage novel clinical applications of fractals in the cardiac imaging domain. Level of Evidence: 3. Technical Efficacy: Stage 1. J. Magn. Reson. Imaging 2017;46:1082–1088.

Original languageEnglish (US)
Pages (from-to)1082-1088
Number of pages7
JournalJournal of Magnetic Resonance Imaging
Issue number4
StatePublished - Oct 2017


  • MRI
  • diagnostic testing
  • etiology
  • myocardial biology
  • myocardial cardiomyopathy disease
  • structure

ASJC Scopus subject areas

  • Radiology Nuclear Medicine and imaging


Dive into the research topics of 'Community delivery of semiautomated fractal analysis tool in cardiac mr for trabecular phenotyping'. Together they form a unique fingerprint.

Cite this