Cardiovascular Biomechanics and A.I. Laboratory
Imperial College London, Department of bioengineering

Menu Lab Home People Research Projects Our Codes Publications Position Vacancy	Machine Learning Biomechanics
	Accurate CNN Image Registration of Echocardiography Images We developed a new Convolutional Neural Network (CNN) framework for image registration of echocardiography images, to aid motion tracking and strain computations, that has better performance than existing techniques. Our approach combines shape encoding, adversarial training, and multi-scale registration to achieve this, it could out-perform gold standard traditional techniques such as using the optimal use of algorithms like SimpleElastix. Our Algorithm, ACDLIR, produced displacement fields that more accurately warped the moving image to match the fixed image than alternative algorithms. Reference: - Hasan MK, Zhu H, Yang G, Yap CH. Deep learning image registration for cardiac motion estimation in adult and fetal echocardiography via a focus on anatomic plausibility and texture quality of warped image. Computers in Biology and Medicine. 2025 Mar 1;187:109719.
	Feedback Attention to Enhance 2D and 3D Echo Image Registration Image registration is a potential replacement for speckle tracking echocardiography, as it provides the advantage of not requiring very high-quality images or optimal frame rate that speckle tracking needs. We developed a new strategy to enhance image registration performance, and to enable successful scaling up to 3D echocardiography, which poses additional challenges to registration accuracy. Our approach is to generate a co-attention map that describes the remaining registration error, and to feedback this to the registration network to improve self-supervision. This technique can be applied to a wide range of various networks to improve their performance. Our proposed feedback attention (FBA) enhancement to a deep learning registration network can improve registration performance. Reference: - Hasan MK, Luo Y, Yang G, Yap CH. Feedback Attention to Enhance Unsupervised Deep Learning Image Registration in 3D Echocardiography. IEEE Transactions on Medical Imaging. 2025 Jan 16.
	A Deep Learning Plugin for Motion Enhancement to Any Segmentation Network Image segmentation is an essential image processing step for quantification of cardiac anatomic features. Segmentation is especially difficult for noisy images such as from echocardiography. Motion enhancement is a robust strategy to improve segmentation, as it can bridge over transient loss of signals and presentation of noise. However, motion enhancement can significantly increase computational burden, such as when image dimension is increased by the time dimension. We designed the Temporal Attention Module (TAM), a small and number plug-in that can be added to a wide range of network types to enhance them with motion information to improve their performance. Only 3 time points is needed for optimal performance. The module has a novel architecture based on KQV projection rather than co-attention. Reference: - Hasan MK, Yang G, Yap CH. Motion-enhancement to Echocardiography Segmentation via Inserting a Temporal Attention Module: An Efficient, Adaptable, and Scalable Approach. arXiv preprint arXiv:2501.14929. 2025 Jan 24.