Deep Learning Applied to Multi-Structure Segmentation in 2D Echocardiography: A Preliminary Investigation of the Required Database Size

Leclerc, Sarah; Smistad, Erik; Grenier, Thomas; Lartizien, Carole; Østvik, Andreas; Espinosa, Florian; Jodoin, Pierre-Marc; Løvstakken, Lasse; Bernard, Olivier

doi:10.1109/ultsym.2018.8580136

Cited by 13 publications

(11 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…• U-net is robust to image quality, and able to segment several structures without dropping performance at any instant (ED and ES) or view (AP4C and AP2C). It also generalizes well with only a 250 patients training set (half of the dataset), while still benefiting from additional training data, unlike SRF (Leclerc et al, 2018). • Encoder-decoder networks outperform state-of-the-art non-deep learning methods and its accuracy is within the inter-expert scores.…”

Section: Main Conclusionmentioning

confidence: 95%

Deep Learning Segmentation in 2D echocardiography using the CAMUS dataset : Automatic Assessment of the Anatomical Shape Validity

Leclerc¹,

Smistad²,

Østvik³

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

We recently published a deep learning study on the potential of encoder-decoder networks for the segmentation of the 2D CAMUS ultrasound dataset. We propose in this abstract an extension of the evaluation criteria to anatomical assessment, as traditional geometric and clinical metrics in cardiac segmentation do not take into account the anatomical correctness of the predicted shapes. The completed study sheds a new light on the ranking of models.

show abstract

Section: Main Conclusionmentioning

confidence: 95%

Deep Learning Segmentation in 2D echocardiography using the CAMUS dataset : Automatic Assessment of the Anatomical Shape Validity

Leclerc¹,

Smistad²,

Østvik³

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…This network classifies each pixel into one of four classes: 1) background, 2) left ventricle, 3) myocardium and 4) left atrium. The network was trained with dice loss and the Adam optimizer using a dataset consisting of A4C and A2C images from both ED and ES of 500 patients [6]. Fig.…”

Section: B Segmentationmentioning

confidence: 99%

“…Another essential step for automating cardiac measurements is image segmentation. This has been an active research area for several decades for both 2D and 3D ultrasound, and recently it has been shown that deep neural networks can also perform this efficiently and accurately [3][4][5][6]. In addition to view classification and segmentation, the estimation of end-diastole and end-systole, as well as extraction of apex and base landmarks is needed.…”

Section: Introductionmentioning

confidence: 99%

Fully Automatic Real-Time Ejection Fraction and MAPSE Measurements in 2D Echocardiography Using Deep Neural Networks

Smistad¹,

Østvik²,

Salte³

et al. 2018

2018 IEEE International Ultrasonics Symposium (IUS)

Self Cite

View full text Add to dashboard Cite

Cardiac ultrasound measurements such as left ventricular volume, ejection fraction (EF) and mitral annular plane systolic excursion (MAPSE) are time consuming and highly observer dependent. In this work, we investigate if deep neural networks can be used to fully automate cardiac ultrasound measurements in real-time while scanning. One neural network was used for identifying and separate the cardiac views while a second neural network performed segmentation of the left ventricle. By using TensorFlow, FAST and the highly optimized cuDNN backend real-time runtime of the entire pipeline was achieved with an average frames per second of 43, thus enabling these measurements to be performed while an operator is scanning. The measurement accuracy was evaluated using a Bland-Altmann analysis on a dataset of 75 patients resulting in (−13.7 ± 8.6)% for EF and (−0.9 ± 4.6) mm for MAPSE. It is concluded that deep learning can be used to fully automate these measurements, however more work remains to improve the accuracy.

show abstract

“…Unet architecture was also introduced for heart segmentation in some researches. [ 18 ] The main difference between Unet and fully convolutional networks is the symmetry of Unet architecture along with adding concatenation operation in the decoder path, as is explained in section 2-1. Subsequently, various improvements in Unet are addressed in numerous articles that were mainly focusing on the innovations in the following three areas (i) changing the encoder network to extract more abstract features, (ii) modifying the up-sampling strategy, and (iii) changing the skip connections.…”

Section: Introductionmentioning

confidence: 99%

“…Subsequently, various improvements in Unet are addressed in numerous articles that were mainly focusing on the innovations in the following three areas (i) changing the encoder network to extract more abstract features, (ii) modifying the up-sampling strategy, and (iii) changing the skip connections. For example, in[ 18 ] Unet++ was introduced. In the Unet network, data in skip connection path transmit directly from the encoder to the decoder, whereas, in Unet++, it passes through a series of Conv-blocks and transfers the feature maps.…”

Section: Introductionmentioning

confidence: 99%

Residual learning: A new paradigm to improve deep learning-based segmentation of the left ventricle in magnetic resonance imaging cardiac images

et al. 2021

View full text Add to dashboard Cite

Background: Recently, magnetic resonance imaging (MRI) has become a useful tool for the early detection of heart failure. A vital step of this process is a valid measurement of the left ventricle's properties, which seriously depends on the accurate segmentation of the heart in captured images. Although various schemes have been tested for this segmentation so far, the latest proposed methods have used the concept of deep learning to estimate the range of the left ventricle in cardiac MRI images. While deep learning methods can lead to better results than their classical alternatives, but unfortunately, the gradient vanishing and exploding problems may hamper their efficiency for the accurate segmentation of the left ventricle in MRI heart images. Methods: In this article, a new concept called residual learning is utilized to improve the performance of deep learning schemes against gradient vanishing problems. For this purpose, the Residual Network of Residual Network (i.e., Residual of Residual) substructure is utilized inside the main deep learning architecture (e.g., Unet), which provides more significant detection indexes. Results and Conclusion: The proposed method's performances and its alternatives were evaluated on Sunnybrook Cardiac Data as a reliable dataset in the left ventricle segmentation. The results show that the detection parameters are improved at least by 5%, 3.5%, 8.1%, and 11.4% compared to its deep alternatives in terms of Jaccard, Dice, precision, and false-positive rate indexes, respectively. These improvements were made when the recall parameter was reduced to a negligible value (i.e., approximately 1%). Overall, the proposed method can be used as a suitable tool for more accurate detection of the left ventricle in MRI images.

show abstract

Deep Learning Applied to Multi-Structure Segmentation in 2D Echocardiography: A Preliminary Investigation of the Required Database Size

Cited by 13 publications

References 7 publications

Deep Learning Segmentation in 2D echocardiography using the CAMUS dataset : Automatic Assessment of the Anatomical Shape Validity

Deep Learning Segmentation in 2D echocardiography using the CAMUS dataset : Automatic Assessment of the Anatomical Shape Validity

Fully Automatic Real-Time Ejection Fraction and MAPSE Measurements in 2D Echocardiography Using Deep Neural Networks

Residual learning: A new paradigm to improve deep learning-based segmentation of the left ventricle in magnetic resonance imaging cardiac images

Contact Info

Product

Resources

About