Advances in Medical Imaging and Reconstruction

The field of medical imaging and reconstruction is rapidly evolving, with a focus on developing more accurate and efficient methods for image registration, segmentation, and reconstruction. Recent research has explored the use of deep learning techniques, such as convolutional neural networks and transformers, to improve the accuracy and robustness of these methods. Additionally, there is a growing interest in using event-based cameras and other novel sensing technologies to capture high-speed and high-resolution images of dynamic scenes.

Noteworthy papers in this area include: URNet, which introduces an uncertainty-aware refinement network for event-based stereo depth estimation, achieving state-of-the-art results on the DSEC dataset. CPT-4DMR, which proposes a continuous spatial-temporal representation for 4D-MRI reconstruction, demonstrating significant improvements in efficiency and accuracy compared to conventional methods. MoAngelo, which presents a novel framework for motion-aware neural surface reconstruction, achieving superior reconstruction accuracy on the ActorsHQ dataset. SHMoAReg, which introduces a spark deformable image registration method via spatial heterogeneous mixture of experts and attention heads, showing consistent improvements over various methods on two publicly available datasets.

Sources

Ideal Registration? Segmentation is All You Need

MoAngelo: Motion-Aware Neural Surface Reconstruction for Dynamic Scenes

URNet: Uncertainty-aware Refinement Network for Event-based Stereo Depth Estimation

CPT-4DMR: Continuous sPatial-Temporal Representation for 4D-MRI Reconstruction

Event-guided 3D Gaussian Splatting for Dynamic Human and Scene Reconstruction

DeblurSplat: SfM-free 3D Gaussian Splatting with Event Camera for Robust Deblurring

A DyL-Unet framework based on dynamic learning for Temporally Consistent Echocardiographic Segmentation

Adaptive von Mises-Fisher Likelihood Loss for Supervised Deep Time Series Hashing

Anatomically Constrained Transformers for Cardiac Amyloidosis Classification

Learning to Stop: Reinforcement Learning for Efficient Patient-Level Echocardiographic Classification

Diffusion-Augmented Contrastive Learning: A Noise-Robust Encoder for Biosignal Representations

SHMoAReg: Spark Deformable Image Registration via Spatial Heterogeneous Mixture of Experts and Attention Heads

Built with on top of