Diffeomorphic Deformation
Diffeomorphic deformation, the process of smoothly and reversibly warping one object (e.g., image, shape, time series) onto another while preserving topology, is a crucial technique in various scientific fields. Current research focuses on developing efficient and accurate algorithms, often leveraging deep learning architectures like convolutional neural networks (CNNs) and transformers, to achieve diffeomorphic transformations, particularly for image registration and time series analysis. These advancements are significantly impacting medical image analysis (e.g., improving organ segmentation and radiotherapy planning) and other areas requiring robust and topology-preserving data alignment.
Papers
Towards Positive Jacobian: Learn to Postprocess Diffeomorphic Image Registration with Matrix Exponential
Soumyadeep Pal, Matthew Tennant, Nilanjan Ray
A training-free recursive multiresolution framework for diffeomorphic deformable image registration
Ameneh Sheikhjafari, Michelle Noga, Kumaradevan Punithakumar, Nilanjan Ray