This technology is suggested for use in pulmonary hypertension, a disease with high mortality in which treatment choice depends on individual risk stratification.
Doctors suggest different treatment regimes for patients with pulmonary hypertension by making judgements based on a range of measures. Within this method, there is scope for mistakes to be made.
The technology is a deep-learning algorithm that is trained to find correspondence between heart motion and patient outcome, and which can efficiently predict human survival.
Motion analysis is a technique used in computer vision to understand the behaviour of moving objects in sequences of images. It is possible to predict future events based on the current state of a moving 3D scene by learning correspondences between patterns of motion and subsequent outcomes. Imperial researchers used machine learning techniques to analyse the motion dynamics of the beating heart and created a network – 4Dsurvival – which predicts survival outcomes in pulmonary hypertension patients with greater accuracy than doctors’ measures.
- In a study of 302 patients, the accuracy of survival predictions for 4Dsurvival was 75%, significantly higher than the human benchmark of 59%
Intellectual property information
The technology is protected by a UK priority patent application, number GB1816281.8
Link to published paper(s)
Bello, G.A., Dawes, T.J.W., Duan, J. et al. Deep-learning cardiac motion analysis for human survival prediction. Nat Mach Intell 1, 95–104 (2019). https://doi.org/10.1038/s42256-019-0019-2
Dr Declan O’Regan, Reader in Imaging Sciences, Faculty of Medicine, Imperial College London