Virtual native enhancement instead of late gadolinium enhancement

Virtual native enhancement instead of late gadolinium enhancement Late gadolinium enhancement (LGE) cardiovascular magnetic resonance (CMR) imaging is considered the gold standard for non-invasive myocardial tissue characterization [1]. Artificial intelligence was used to develop a CMR virtual native enhancement (VNE) imaging which does not need an intravenous contrast as in LGE. VNE uses a deep learning model with multiple streams of convolutional neural networks to enhance existing signals of native T1 maps and cine imaging of cardiac structure and function to produce LGE equivalent images. T1 maps are pixel wise maps of tissue T1 relaxation times. The technology was developed in CMR datasets from the Hypertrophic Cardiomyopathy Registry. Datasets were randomized into two independent groups for deep learning and testing as is usual with artificial intelligence programs. Test data of VNE and LGE were scored by experienced persons to assess image quality, myocardial lesion quantification and visuospatial agreement. 4093 triplets of matched T1 maps, cine and LGE datasets were obtained from 1348 HCM patients. 2695 datasets were used for the development of VNE method and 345 for independent testing. Significantly better image quality than LGE was obtained by VNE. VNE correlated with LGE in detecting and quantifying both hyperintensity and intermediate intensity myocardial lesions in 326 datasets (121 patients). A remarkable feature was that native CMR images can be acquired with...
Source: Cardiophile MD - Category: Cardiology Authors: Tags: Machine Learning and AI in Cardiology Source Type: blogs