Fully automated quantification of biventricular volumes and function in cardiovascular magnetic resonance: applicability to clinical routine settings

Abstract

Background: Cardiovascular magnetic resonance (CMR) represents the clinical gold standard for the assessment of biventricular morphology and function. Since manual post-processing is time-consuming and prone to observer variability, efforts have been directed towards automated volumetric quantification. In this study, we sought to validate the accuracy of a novel approach providing fully automated quantification of biventricular volumes and function in a "real-world" clinical setting. Methods: Three-hundred CMR examinations were randomly selected from the local data base. Fully automated quantification of left ventricular (LV) mass, LV and right ventricular (RV) end-diastolic and end-systolic volumes (EDV/ ESV), stroke volume (SV) and ejection fraction (EF) were performed overnight using commercially available software (suiteHEART®, Neosoft,

Fully automated quantification of left ventricular volumes and function in cardiac MRI: clinical evaluation of a deep learning-based algorithm

“…There is scarce data on the performance and utility of such algorithms in clinical routine. One recent study investigated a similar algorithm by a different vendor in 300 cases randomly selected cases from routine clinical care [13]. Their study found that agreement between fully automated and expert manual segmentation was lower for the right ventricle than for the left ventricle, lower at 3 T, in cases of compromised image quality and in cases of challenging anatomy such as repaired Tetralogy of Fallot [13].…”

“…One recent study investigated a similar algorithm by a different vendor in 300 cases randomly selected cases from routine clinical care [13]. Their study found that agreement between fully automated and expert manual segmentation was lower for the right ventricle than for the left ventricle, lower at 3 T, in cases of compromised image quality and in cases of challenging anatomy such as repaired Tetralogy of Fallot [13]. Our study design builds on these results in several aspects: we focused our investigation on the left ventricle and chose to exclude examinations performed at 3 T, examinations with severe artefacts and substantially altered anatomy (severe congenital heart disease, large left ventricular aneurysms).…”

“…More recently, the dramatic evolution in artificial intelligence-specifically deep learning-has enabled the development of fully automated algorithms for volumetric analysis of CMR datasets. Initials studies indicate substantial time-saving and reduced observer bias compared to manual or semi-automated contouring [8,[9][10][11][12][13][14]. Some of these studies used pre-processed datasets, CMR examinations of healthy individuals from population-based studies or highly selected patient populations.…”

Head‐to‐head comparison of cardiovascular MR feature tracking cine versus acquisition‐based deformation strain imaging using myocardial tagging and strain encoding

“…Artificial intelligence has proven its significant value in medicine in general, 31 and automated analyses using NeoSoft have already demonstrated excellent reproducibility for cardiac volume quantification compared with manual evaluation as the reference standard. 32 Hence, the extension to strain holds great clinical potential and could finally lead to fully automated clinical-routine GLS quantification with improved management of various diseases. 3,[26][27][28] Nevertheless, absolute agreement comparing FT to tagging or fSENC was less convincing.…”

Development of Softcomputing Tool to Evaluate Heart MRI Slices

“…The work of Guttman et al (1997) discusses the procedures to be followed while evaluation the cardiac MRI [8]. Backhaus et al (2019) implemented a technique for a computerized quantification of biventricular volumes and function using the cardiac MRI [9]. The work of Khan et al (2016) presented a detailed review of cardiac MRI evaluation procedures discussed in 100 related articles [10].…”