Efficient Segmentation for Left Atrium With Convolution Neural Network Based on Active Learning in Late Gadolinium Enhancement Magnetic Resonance Imaging

Abstract

Background To propose fully automatic segmentation of left atrium using active learning with limited dataset in late gadolinium enhancement in cardiac magnetic resonance imaging (LGE-CMRI). Methods An active learning framework was developed to segment the left atrium in cardiac LGE-CMRI. Patients (n = 98) with atrial fibrillation from the Korea University Anam Hospital were enrolled. First, 20 cases were delineated for ground truths by two experts and used for training a draft model. Second, the 20 cases from the first step and 50 new cases, corrected in a human-in-the-loop manner after predicting using the draft model, were used to train the next model; all 98 cases (70 cases from the second step and 28 new cases) were trained. An additional 20 LGE-CMRI were evaluated in each step. Results The Dice coefficients for the three steps were 0.85 ± 0.06, 0.89 ± 0.02, and 0.90 ± 0.02, respectively. The biases (95% confidence interval) in the Bland-Altman plots of each step were 6.36% (−14.90–27.61), 6.21% (−9.62–22.03), and 2.68% (−8.57–13.93). Deep active learning-based annotation times were 218 ± 31 seconds, 36.70 ± 18 seconds, and 36.56 ± 15 seconds, respectively. Conclusion Deep active learning reduced annotation time and enabled efficient training on limited LGE-CMRI.

Efficient labelling for efficient deep learning: the benefit of a multiple-image-ranking method to generate high volume training data applied to ventricular slice level classification in cardiac MRI

“…One approach is transfer learning using networks that have been pre-trained on huge datasets in an unsupervised fashion and then fine-tuning for the specific task using a few (much fewer than if training from scratch) manually annotated examples ( 16 ). Another approach is ‘active learning’, in which the examples within the dataset that are most likely to contribute to network learning are selected for manual annotation ( 17 , 18 ). This can be combined with metrics of label uncertainty to identify the data that will yield the most performance-value from manual annotation ( 19 ).…”

Artificial Intelligence in Cardiovascular CT and MR Imaging

“…They demonstrated a mean accuracy of 0.856 ± 0.033 and a mean DSC of 0.702 ± 0.071 for left atrium scar quantification. Even Cho et al used a 3D U-net architecture using a limited dataset of LGE CMR, demonstrating a DSC value of 0.90 [ 97 ].…”

Artificial Intelligence in the Image-Guided Care of Atrial Fibrillation

“…For instance, in the 2018 LA segmentation challenge (LASC) dataset [ 13 ], currently the largest open-source dataset for LA segmentation on late gadolinium-enhanced MRI (LGE-MRI) in patients with AF, the structure of interest was defined as the pixels within the LA endocardial surface, including the MV and the LAA, as well as the extent of the PV sleeves. In the datasets used in other selected publications, the definitions of the structure of interest vary, including solely the LA [ 14 , 15 ] or various combinations of the LA and its substructures [ 16 , 17 , 18 ] on contrast-enhanced CT (CECT) or LGE-MRI. In addition, Jin et al [ 19 ] proposed a model for the segmentation of the LAA on CECT, which is desirable for LAA occlusion procedures [ 20 ].…”

“…Second, the framework utilized dilated residual learning to learn features from the sagittal and coronal views. Methods for direct 3D LA segmentation [ 14 , 26 , 27 ] have also been proposed. Borra et al [ 26 ] demonstrated that the 3D variant of U-net outperforms its 2D counterpart for LA segmentation.…”