Proof of concept study: Phenotyping of cardiac plaques by means of Radiomics
Date
2023-10-23
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Saudi Digital Library
Abstract
Objectives: To assess the statistical reliability of radiomic features across two
different platforms, Pyradiomics and LIFEx, and to evaluate the effectiveness of the
Virtual Monoenergetic Imaging (VMI) protocol for visualizing vessels and plaques
in cardiac imaging.
Methods: The statistical reliability of radiomic features was assessed in 11
Regions of Interest (ROIs) within the Catphan 700 phantom. These features were
calculated using two IBSI-compliant platforms (Pyradiomics and LIFEx). The effectiveness
of the Virtual Monoenergetic Imaging (VMI) protocol in visualizing vessels
and plaques within cardiac imaging was evaluated using a CTDI phantom, which
contained three different tubes and a plaque. The impact of user-defined calculation
settings and the effectiveness of VMI were assessed by calculating the intraclass
correlation coefficient (ICC) and confidence intervals.
Results: The reliability of radiomic features calculated by the two software
platforms was excellent (ICC CI > 0.9) for only 8 out of the 17 features when
comparing the two platforms but significantly improved to (ICC CI > 0.9) for
12 out of the 17 features when harmonized settings were applied. Failure to use
harmonized settings resulted in lower reliability for most features. When assessing
feature classes, the reliability improved to some extent. In the second part of the
experiment, the ICC values varied among different vessels and a plaque, highlighting
the influence of VMI on the reliability of radiomics features.
Conclusion: The significant impact of user-defined calculation settings on IBSI compliant
platforms emphasizes the need for clear reporting of these parameters
to enhance reproducibility and comparability. Collaborative initiatives within the
radiomics community are essential to establish standardized and harmonized calculation
settings, potentially leading to more robust and reliable radiomic analyses in
medical imaging.
Description
Keywords
Radiomics, Plaque, Monochromatic, CAD, LIFEx, Pyradiomics