Purpose: The aim of this study was to compare the performance of quantitative methods, either semiautomated or automated, for left ventricular (LV) nonviable tissue analysis from cardiac magnetic resonance late gadolinium enhancement (CMR-LGE) images. Materials and Methods: The investigated segmentation techniques were: (i) n-standard deviations thresholding; (ii) full width at half maximum thresholding; (iii) Gaussian mixture model classification; and (iv) fuzzy c-means clustering. These algorithms were applied either in each short axis slice (single-slice approach) or globally considering the entire short-axis stack covering the LV (global approach). CMR-LGE images from 20 patients with ischemic cardiomyopathy were retrospectively selected, and results from each technique were assessed against manual tracing. Results: All methods provided comparable performance in terms of accuracy in scar detection, computation of local transmurality, and high correlation in scar mass compared with the manual technique. In general, no significant difference between single-slice and global approach was noted. The reproducibility of manual and investigated techniques was confirmed in all cases with slightly lower results for the nSD approach. Conclusions: Automated techniques resulted in accurate and reproducible evaluation of LV scars from CMR-LGE in ischemic patients with performance similar to the manual technique. Their application could minimize user interaction and computational time, even when compared with semiautomated approaches.

Comparison of image processing techniques for nonviable tissue quantification in late gadolinium enhancement cardiac magnetic resonance images

CAIANI, ENRICO GIANLUCA
2016

Abstract

Purpose: The aim of this study was to compare the performance of quantitative methods, either semiautomated or automated, for left ventricular (LV) nonviable tissue analysis from cardiac magnetic resonance late gadolinium enhancement (CMR-LGE) images. Materials and Methods: The investigated segmentation techniques were: (i) n-standard deviations thresholding; (ii) full width at half maximum thresholding; (iii) Gaussian mixture model classification; and (iv) fuzzy c-means clustering. These algorithms were applied either in each short axis slice (single-slice approach) or globally considering the entire short-axis stack covering the LV (global approach). CMR-LGE images from 20 patients with ischemic cardiomyopathy were retrospectively selected, and results from each technique were assessed against manual tracing. Results: All methods provided comparable performance in terms of accuracy in scar detection, computation of local transmurality, and high correlation in scar mass compared with the manual technique. In general, no significant difference between single-slice and global approach was noted. The reproducibility of manual and investigated techniques was confirmed in all cases with slightly lower results for the nSD approach. Conclusions: Automated techniques resulted in accurate and reproducible evaluation of LV scars from CMR-LGE in ischemic patients with performance similar to the manual technique. Their application could minimize user interaction and computational time, even when compared with semiautomated approaches.
Cardiac imaging techniques; Computerassisted; Image processing; Magnetic resonance imaging; Myocardial infarction; Aged; Algorithms; Cicatrix; Contrast Media; Female; Heart Ventricles; Humans; Image Enhancement; Image Processing, Computer-Assisted; Magnetic Resonance Imaging; Male; Meglumine; Myocardial Infarction; Reproducibility of Results; Organometallic Compounds; Radiology, Nuclear Medicine and Imaging; Pulmonary and Respiratory Medicine
File in questo prodotto:
File Dimensione Formato  
inpress-Carminati-2016-Comparisonof image processing techniques for nonviable tissue quantificationin LGECMR images.pdf

embargo fino al 31/05/2017

Descrizione: Articolo
: Post-Print (DRAFT o Author’s Accepted Manuscript-AAM)
Dimensione 702.74 kB
Formato Adobe PDF
702.74 kB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11311/1012501
Citazioni
  • ???jsp.display-item.citation.pmc??? 5
  • Scopus 17
  • ???jsp.display-item.citation.isi??? 17
social impact