Automatically Segmenting the Left Atrium from Cardiac Images Using Successive 3D U-Nets and a Contour Loss

Abstract

Radiological imaging offers effective measurement of anatomy, which is useful in disease diagnosis and assessment. Previous study [1] has shown that the left atrial wall remodeling can provide information to predict treatment outcome in atrial fibrillation. Nevertheless, the segmentation of the left atrial structures from medical images is still very time-consuming. Current advances in neural network may help creating automatic segmentation models that reduce the workload for clinicians. In this preliminary study, we propose automated, two-stage, threedimensional U-Nets with convolutional neural network, for the challenging task of left atrial segmentation. Unlike previous two-dimensional image segmentation methods, we use 3D U-Nets to obtain the heart cavity directly in 3D. The dual 3D U-Net structure consists of, a first U-Net to coarsely segment and locate the left atrium, and a second U-Net to accurately segment the left atrium under higher resolution. In addition, we introduce a Contour loss based on additional distance information to adjust the final segmentation. We randomly split the data into training datasets (80 subjects) and validation datasets (20 subjects) to train multiple models, with different augmentation setting. Experiments show that the average Dice coefficients for validation datasets are around 0.91 -0.92, the sensitivity around 0.90-0.94 and the specificity 0.99. Compared with traditional Dice loss, models trained with Contour loss in general offer smaller Hausdorff distance with similar Dice coefficient, and have less connected components in predictions. Finally, we integrate several trained models in an ensemble prediction to segment testing datasets.

3D Automatic Segmentation of Aortic Computed Tomography Angiography Combining Multi-View 2D Convolutional Neural Networks

“…Then, the coarse segmentation is used to extract a bounding box of the aortic mask and keep only the area around the aorta. The bounding box coordinates are used to get the ROI with higher resolution, as proposed by Jia et al 3 (b) Single-view Segmentation. Each CTA scan is parsed into 2D axial, sagittal, and coronal views.…”

“…9 In our work, we have adopted another approach exploited for left atrium automatic segmentation. 3 Here, a first 3D U-Net was exploited to extract the region of interest from MR images, then a second 3D U-Net performed a finer segmentation on MR cropped at higher resolution.…”

Style Data Augmentation for Robust Segmentation of Multi-modality Cardiac MRI

“…In this study, we implement the dual U-Net strategy [6], where two networks are trained independently but can be used consecutively in the segmentation pipeline. The first network is trained to segment the target from the low reso-lution inputs, as the original image has to be shrunk down to reduce memory consumption.…”

“…Firstly, we proposed a minimalist U-Net inspired network, tailored to accelerate the convergence speed and to decrease memory usage. Secondly, we adopt the dual network strategy [6], which allows for the segmentation of high resolution targets.…”

Fully Automated Electrophysiological Model Personalisation Framework from CT Imaging

“…The deep learning approach we applied is based on a previously described methodology to segment the left atrium [6]. It relies on the use of two successive specialised U-nets [5].…”