Cardiovascular Biomechanics and A.I. Laboratory |
|
Imperial College London, Department of bioengineering |
|
Links
to Our Codes and Datasets |
|||||||
B-Spline
Fourier Motion Tracking Traditional
pair-wise image registration regularized with cyclic motion consistency Reference:
Wiputra et al., Scientific reports, 10(1), 1-14. Traditional
Finite Element Model of the Heart FEniCS Code for traditional Finite Element modelling
of the cardiac ventricle biomechanics Reference:
Green et al., Biomech Model Mechanobiol.
2024 Oct;23(5):1433-48. PINN
Fluid Mechanics Prediction in Vessels Physics-Informed
Neural Network for predicting detailed fluid mechanics in a range of
idealized vascular geometries. Reference:
Chan et al., Comput Biol Med.
2025 May 1;190:110074. PINN
Fluid Mechanics Prediction in the Left Ventricle Physics-Informed
Neural Network for predicting detailed fluid mechanics in the cardiac left
ventricle using 3D Doppler acquisitions as data guidance. Reference:
Wong et al., Comput Methods Programs Biomed. 2025
Feb 13:108671. Deep
Learning Generation of Cranial Aneurysm Shapes (AneuG) 2-stage
Generative AI model for creating synthetic bifurcation cranial aneurysm
shapes for augmenting flow prediction models. Generation can be conditioned
to produce shapes with specific clinically relevant morphological parameters. Reference:
Ding et al. "Two-Stage Generative Model for Intracranial Aneurysm Meshes
with Morphological Marker Conditioning." MICCAI 2025 [accepted June
2025] Large
CFD Dataset of Bifurcation Cranial Aneurysm Large
dataset of steady flow and pulsatile simulations, based on synthetic aneurysm
shapes generated via AneuG Reference:
TBD PINN
Finite Element Model of the Left Ventricle (IMC-PINN-FE) Fast
Physics Informed Neural Network Algorithm for Finite Element Modelling of
ventricular biomechanics, which constraints to adhere to imaged cardiac
motions. Reference:
TBD Deep
Learning Prediction of Unloaded State of the Left Ventricle from Image Reconstruction
(Load2UnNet) Graph
Attention Network to rapidly calculate the unloaded state 3D mesh of the left
ventricle, from the 3D mesh at end-diastole and inputs of the end-diastolic
pressure, cardiac stiffness scale, and helix angle configurations. Reference:
TBD Graph
Hamornic Deformation Encoder for Cardiac Shapes Guided by Segmented Image
Slices via Differential Slicing and Voxelization Reconstruction
of the 3D mesh of the heart’s shape using the graph harmonic mesh morphic
approach, guided by slices of segmented image via the differentiable slicing
and voxelization technique, the first truly differentiable direct supervision
by image. Reference:
Luo et al., arXiv preprint arXiv:2409.02070. 2024
Sep 3. Feedback
Attention Image Registration for Echo/MRI Images 2D
or 3D image registration with a special feedback attention mechanism that enhances
performance. Reference:
Hasan et al. Feedback Attention to Enhance Unsupervised Deep Learning Image
Registration in 3D Echocardiography. IEEE Transactions on Medical Imaging.
2025 Jan 16. Temporal
Attention Module for Enhancing Segmentation 2D
or 3D image segmentation from medical images with temporal attention
enhancement that is achieved with a small, light-weight,
yet highly effective temporal attention module that can be flexibly added to
a range of network architecture. Reference:
Hasan et al., arXiv preprint arXiv:2501.14929. 2025
Jan 24. Link
TB added |
||||||||
|
|