Deep learning-based optimization for accurate multimodal medical image registration.
Authors
Affiliations (2)
Affiliations (2)
- Department of Electronics and Communication Engineering, Maulana Azad National Institute of Technology, Bhopal, India. [email protected].
- Department of Electronics and Communication Engineering, Maulana Azad National Institute of Technology, Bhopal, India.
Abstract
In recent years, there has been a growing need of incorporating more than one image registration methods in medical image processing. This has been because diagnosis and treatment planning made effective by these techniques. Our work focuses on a complex implementation of the multimodal U-Net model that we adapted for the purpose of image registration from ADNI, COPDGene and OAI datasets. Our work was concentrated on designing and testing multi-layered transformation models that included rigid, affine, and elastic or non-rigid transformations, enhancing feature extraction methods to create an efficient image registration application. This study thus also supports the assertion that deep learning based multimodal U-Net models are superior to traditional methods placed on efficiency and accuracy of registration processes. In addition to this, it benefits the patients who associated with more advanced and expensive imaging technologies through better imaging outcomes. However, the objections regarding the benefit of this technology should be kept in mind as these are young technologies which may be confronted in conservative environments. It is worth noting that while promising, the implications of these findings are limited and caution is warranted as further studies are necessary to support their generalization.