CardiacNET: Segmentation of Left Atrium and Proximal Pulmonary Veins from MRI Using Multi-view CNN

Abstract

Anatomical and biophysical modeling of left atrium (LA) and proximal pulmonary veins (PPVs) is important for clinical management of several cardiac diseases. Magnetic resonance imaging (MRI) allows qualitative assessment of LA and PPVs through visualization. However, there is a strong need for an advanced image segmentation method to be applied to cardiac MRI for quantitative analysis of LA and PPVs. In this study, we address this unmet clinical need by exploring a new deep learning-based segmentation strategy for quantification of LA and PPVs with high accuracy and heightened efficiency. Our approach is based on a multi-view convolutional neural network (CNN) with an adaptive fusion strategy and a new loss function that allows fast and more accurate convergence of the backpropagation based optimization. After training our network from scratch by using more than 60K 2D MRI images (slices), we have evaluated our segmentation strategy to the STACOM 2013 cardiac segmentation challenge benchmark. Qualitative and quantitative evaluations, obtained from the segmentation challenge, indicate that the proposed method achieved the state-of-the-art sensitivity (90%), specificity (99%), precision (94%), and efficiency levels (10 seconds in GPU, and 7.5 minutes in CPU).

Fully automated segmentation of the left atrium, pulmonary veins, and left atrial appendage from magnetic resonance angiography by joint‐atlas‐optimization

“…Mortazi et al. used the encoder–decoder U‐net in a multiview framework, and achieved a Dice value of 0.951 using leave‐one‐out cross validation on the same dataset (29 used for training and 1 used for testing) . Deep‐learning‐based approaches have great advantages in accuracy and speed when there is sufficient training data.…”

“…Mortazi et al used the encoder-decoder U-net in a multiview framework, and achieved a Dice value of 0.951 using leave-one-out cross validation on the same dataset (29 used for training and 1 used for testing). 21 Deep-learningbased approaches have great advantages in accuracy and speed when there is sufficient training data. Nevertheless, in scenarios where the annotated dataset is scarce and where the training and testing datasets may have different properties, the atlas-based approach can still be the method of choice to provide reliable and robust segmentation; our proposed joint-atlas-optimization further refines the segmentation at the border, producing more accurate anatomical depiction.…”

“…Mortazi et al. proposed to segment the LA using the encoder–decoder architecture (U‐NET) in a multiview framework and combined with an adaptive fusion strategy …”

“…Mortazi et al proposed to segment the LA using the encoder-decoder architecture (U-NET 20 ) in a multiview framework and combined with an adaptive fusion strategy. 21 Contour-based approaches are generally sensitive to local image quality, and lacking in global shape constraints. Database approaches are potentially the best-performing ones, but would demand sufficient annotated data to train the neural network, which is typically characterized by tens of thousands of parameters.…”

Simultaneous left atrium anatomy and scar segmentations via deep learning in multiview information with attention

“…More recently, convolutional neural networks (CNN) based approaches have been proposed to segment the LA and PV [12] [13] [14][15] [16] and a grand challenge has been held for LA anatomy segmentation [17]. These research studies on LA anatomy segmentation can potentially be useful for LA scars segmentation although to the best of our knowledge, this has not been done to date.…”