Dataset Information

Seg-CapNet: A Capsule-Based Neural Network for the Segmentation of Left Ventricle from Cardiac Magnetic Resonance Imaging.

ABSTRACT: Deep neural networks (DNNs) have been extensively studied in medical image segmentation. However, existing DNNs often need to train shape models for each object to be segmented, which may yield results that violate cardiac anatomical structure when segmenting cardiac magnetic resonance imaging (MRI). In this paper, we propose a capsule-based neural network, named Seg-CapNet, to model multiple regions simultaneously within a single training process. The Seg-CapNet model consists of the encoder and the decoder. The encoder transforms the input image into feature vectors that represent objects to be segmented by convolutional layers, capsule layers, and fully-connected layers. And the decoder transforms the feature vectors into segmentation masks by up-sampling. Feature maps of each down-sampling layer in the encoder are connected to the corresponding up-sampling layers, which are conducive to the backpropagation of the model. The output vectors of Seg-CapNet contain low-level image features such as grayscale and texture, as well as semantic features including the position and size of the objects, which is beneficial for improving the segmentation accuracy. The proposed model is validated on the open dataset of the Automated Cardiac Diagnosis Challenge 2017 (ACDC 2017) and the Sunnybrook Cardiac Magnetic Resonance Imaging (MRI) segmentation challenge. Experimental results show that the mean Dice coefficient of Seg-CapNet is increased by 4.7% and the average Hausdorff distance is reduced by 22%. The proposed model also reduces the model parameters and improves the training speed while obtaining the accurate segmentation of multiple regions.

Supplementary information

The online version contains supplementary material available at 10.1007/s11390-021-0782-5.

SUBMITTER: Cao YJ

PROVIDER: S-EPMC8044657 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:ObjectivesTo determine the test-retest reproducibility and observer variability of CMR-derived LA function, using (i) LA strain (LAS) and strain rate (LASR), and (ii) LA volumes (LAV) and emptying fraction (LAEF).MethodsSixty participants with and without cardiovascular disease (aortic stenosis (AS) (n = 16), type 2 diabetes (T2D) (n = 28), end-stage renal disease on haemodialysis (n = 10) and healthy volunteers (n = 6)) underwent two separate CMR scans 7-14 days apart. LAS and LASR, corresponding to LA reservoir, conduit and contractile booster-pump function, were assessed using Feature Tracking software (QStrain v2.0). LAEF was calculated using the biplane area length method (QMass v8.1). Both were assessed using 4- and 2-chamber long-axis standard steady-state free precession cine images, and average values were calculated. Intra- and inter-observer variabilities were assessed in 10 randomly selected participants.ResultsThe test-retest reproducibility was moderate to poor for all strain and strain rate parameters. Overall, strain and strain rate corresponding to reservoir phase (LAS_r, LASR_r) were the most reproducible, yielding the smallest coefficient of variance (CoV) (29.9% for LAS_r, 28.9% for LASR_r). The test-retest reproducibility for LAVs and LAEF was good: LAVmax CoV = 19.6% ICC = 0.89, LAVmin CoV = 27.0% ICC = 0.89 and total LAEF CoV = 15.6% ICC = 0.78. The inter- and intra-observer variabilities were good for all parameters except for conduit function.ConclusionThe test-retest reproducibility of LA strain and strain rate assessment by CMR utilising Feature Tracking is moderate to poor across disease states, whereas LA volume and emptying fraction are more reproducible on CMR. Further improvements in LA strain quantification are needed before widespread clinical application.Key points• LA strain and strain rate assessment using Feature Tracking on CMR has moderate to poor test-retest reproducibility across disease states. • The test-retest reproducibility for the biplane method of assessing LA function is better than strain assessment, with lower coefficient of variances and narrower limits of agreement on Bland-Altman plots. • Biplane LA volumetric measurement also has better intra- and inter-observer variability compared to strain assessment.

Dataset Information

Seg-CapNet: A Capsule-Based Neural Network for the Segmentation of Left Ventricle from Cardiac Magnetic Resonance Imaging.

Supplementary information

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets