Aim. The presence of myocardial scar is a strong predictor of ventricular remodeling, cardiac dysfunction and mortality. Our aim was to assess quantitatively the presence of scar tissue from cardiac-magnetic-resonance (CMR) with late-Gadolinium-enhancement (LGE) images using a deep-learning (DL) approach. Methods. Scar segmentation was performed automatically with a DL approach based on ENet, a deep fully-convolutional neural network (FCNN). We investigated three different ENet configurations. The first configuration (C1) exploited ENet to retrieve directly scar segmentation from the CMR-LGE images. The second (C2) and third (C3) configurations performed scar segmentation in the myocardial region, which was previously obtained in a manual or automatic way with a state-of-the-art DL method, respectively. Results. When tested on 250 CMR-LGE images from 30 patients, the best-performing configuration (C2) achieved 97% median accuracy (inter-quartile (IQR) range = 4%) and 71% median Dice similarity coefficient (IQR = 32%). Conclusions. DL approaches using ENet are promising in automatically segmenting scars in CMR-LGE images, achieving higher performance when limiting the search area to the manually-defined myocardial region.
Automated Scar Segmentation From Cardiac Magnetic Resonance-Late Gadolinium Enhancement Images Using a Deep-Learning Approach
Moccia, Sara;
2018-01-01
Abstract
Aim. The presence of myocardial scar is a strong predictor of ventricular remodeling, cardiac dysfunction and mortality. Our aim was to assess quantitatively the presence of scar tissue from cardiac-magnetic-resonance (CMR) with late-Gadolinium-enhancement (LGE) images using a deep-learning (DL) approach. Methods. Scar segmentation was performed automatically with a DL approach based on ENet, a deep fully-convolutional neural network (FCNN). We investigated three different ENet configurations. The first configuration (C1) exploited ENet to retrieve directly scar segmentation from the CMR-LGE images. The second (C2) and third (C3) configurations performed scar segmentation in the myocardial region, which was previously obtained in a manual or automatic way with a state-of-the-art DL method, respectively. Results. When tested on 250 CMR-LGE images from 30 patients, the best-performing configuration (C2) achieved 97% median accuracy (inter-quartile (IQR) range = 4%) and 71% median Dice similarity coefficient (IQR = 32%). Conclusions. DL approaches using ENet are promising in automatically segmenting scars in CMR-LGE images, achieving higher performance when limiting the search area to the manually-defined myocardial region.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.