Deep-learning cardiac motion analysis for human survival prediction

Research output: Contribution to journalArticle

Standard

Deep-learning cardiac motion analysis for human survival prediction. / Bello, Ghalib A.; Dawes, Timothy J. W.; Duan, Jinming; Biffi, Carlo; De Marvao, Antonio; Howard, Luke S. G. E.; Gibbs, J. Simon R.; Wilkins, Martin R.; Cook, Stuart A.; Rueckert, Daniel; O’regan, Declan P.

In: Nature Machine Intelligence, Vol. 1, No. 2, 11.02.2019, p. 95-104.

Research output: Contribution to journalArticle

Harvard

Bello, GA, Dawes, TJW, Duan, J, Biffi, C, De Marvao, A, Howard, LSGE, Gibbs, JSR, Wilkins, MR, Cook, SA, Rueckert, D & O’regan, DP 2019, 'Deep-learning cardiac motion analysis for human survival prediction', Nature Machine Intelligence, vol. 1, no. 2, pp. 95-104. https://doi.org/10.1038/s42256-019-0019-2

APA

Bello, G. A., Dawes, T. J. W., Duan, J., Biffi, C., De Marvao, A., Howard, L. S. G. E., Gibbs, J. S. R., Wilkins, M. R., Cook, S. A., Rueckert, D., & O’regan, D. P. (2019). Deep-learning cardiac motion analysis for human survival prediction. Nature Machine Intelligence, 1(2), 95-104. https://doi.org/10.1038/s42256-019-0019-2

Vancouver

Author

Bello, Ghalib A. ; Dawes, Timothy J. W. ; Duan, Jinming ; Biffi, Carlo ; De Marvao, Antonio ; Howard, Luke S. G. E. ; Gibbs, J. Simon R. ; Wilkins, Martin R. ; Cook, Stuart A. ; Rueckert, Daniel ; O’regan, Declan P. / Deep-learning cardiac motion analysis for human survival prediction. In: Nature Machine Intelligence. 2019 ; Vol. 1, No. 2. pp. 95-104.

Bibtex

@article{36ff9b2df63c4b479495dbc4203969f8,
title = "Deep-learning cardiac motion analysis for human survival prediction",
abstract = "Motion analysis is used in computer vision to understand the behaviour of moving objects in sequences of images. Optimizing the interpretation of dynamic biological systems requires accurate and precise motion tracking as well as efficient representations of high-dimensional motion trajectories so that these can be used for prediction tasks. Here we use image sequences of the heart, acquired using cardiac magnetic resonance imaging, to create time-resolved three-dimensional segmentations using a fully convolutional network trained on anatomical shape priors. This dense motion model formed the input to a supervised denoising autoencoder (4Dsurvival), which is a hybrid network consisting of an autoencoder that learns a task-specific latent code representation trained on observed outcome data, yielding a latent representation optimized for survival prediction. To handle right-censored survival outcomes, our network used a Cox partial likelihood loss function. In a study of 302 patients, the predictive accuracy (quantified by Harrell{\textquoteright}s C-index) was significantly higher (P = 0.0012) for our model C = 0.75 (95% CI: 0.70–0.79) than the human benchmark of C = 0.59 (95% CI: 0.53–0.65). This work demonstrates how a complex computer vision task using high-dimensional medical image data can efficiently predict human survival.",
author = "Bello, {Ghalib A.} and Dawes, {Timothy J. W.} and Jinming Duan and Carlo Biffi and {De Marvao}, Antonio and Howard, {Luke S. G. E.} and Gibbs, {J. Simon R.} and Wilkins, {Martin R.} and Cook, {Stuart A.} and Daniel Rueckert and O{\textquoteright}regan, {Declan P.}",
year = "2019",
month = feb
day = "11",
doi = "10.1038/s42256-019-0019-2",
language = "English",
volume = "1",
pages = "95--104",
journal = "Nature Machine Intelligence",
issn = "2522-5839",
publisher = "Nature Publishing Group",
number = "2",

}

RIS

TY - JOUR

T1 - Deep-learning cardiac motion analysis for human survival prediction

AU - Bello, Ghalib A.

AU - Dawes, Timothy J. W.

AU - Duan, Jinming

AU - Biffi, Carlo

AU - De Marvao, Antonio

AU - Howard, Luke S. G. E.

AU - Gibbs, J. Simon R.

AU - Wilkins, Martin R.

AU - Cook, Stuart A.

AU - Rueckert, Daniel

AU - O’regan, Declan P.

PY - 2019/2/11

Y1 - 2019/2/11

N2 - Motion analysis is used in computer vision to understand the behaviour of moving objects in sequences of images. Optimizing the interpretation of dynamic biological systems requires accurate and precise motion tracking as well as efficient representations of high-dimensional motion trajectories so that these can be used for prediction tasks. Here we use image sequences of the heart, acquired using cardiac magnetic resonance imaging, to create time-resolved three-dimensional segmentations using a fully convolutional network trained on anatomical shape priors. This dense motion model formed the input to a supervised denoising autoencoder (4Dsurvival), which is a hybrid network consisting of an autoencoder that learns a task-specific latent code representation trained on observed outcome data, yielding a latent representation optimized for survival prediction. To handle right-censored survival outcomes, our network used a Cox partial likelihood loss function. In a study of 302 patients, the predictive accuracy (quantified by Harrell’s C-index) was significantly higher (P = 0.0012) for our model C = 0.75 (95% CI: 0.70–0.79) than the human benchmark of C = 0.59 (95% CI: 0.53–0.65). This work demonstrates how a complex computer vision task using high-dimensional medical image data can efficiently predict human survival.

AB - Motion analysis is used in computer vision to understand the behaviour of moving objects in sequences of images. Optimizing the interpretation of dynamic biological systems requires accurate and precise motion tracking as well as efficient representations of high-dimensional motion trajectories so that these can be used for prediction tasks. Here we use image sequences of the heart, acquired using cardiac magnetic resonance imaging, to create time-resolved three-dimensional segmentations using a fully convolutional network trained on anatomical shape priors. This dense motion model formed the input to a supervised denoising autoencoder (4Dsurvival), which is a hybrid network consisting of an autoencoder that learns a task-specific latent code representation trained on observed outcome data, yielding a latent representation optimized for survival prediction. To handle right-censored survival outcomes, our network used a Cox partial likelihood loss function. In a study of 302 patients, the predictive accuracy (quantified by Harrell’s C-index) was significantly higher (P = 0.0012) for our model C = 0.75 (95% CI: 0.70–0.79) than the human benchmark of C = 0.59 (95% CI: 0.53–0.65). This work demonstrates how a complex computer vision task using high-dimensional medical image data can efficiently predict human survival.

U2 - 10.1038/s42256-019-0019-2

DO - 10.1038/s42256-019-0019-2

M3 - Article

VL - 1

SP - 95

EP - 104

JO - Nature Machine Intelligence

JF - Nature Machine Intelligence

SN - 2522-5839

IS - 2

ER -