Eventos Anais de eventos
COBEM 2023
27th International Congress of Mechanical Engineering
Nuclear Magnetic Resonance Image Reconstruction Using Reconfigurable Simulated System
Submission Author:
Amauri da Costa Júnior , DF
Co-Authors:
Amauri da Costa Júnior, Gerardo Pizo
Presenter: Amauri da Costa Júnior
doi://10.26678/ABCM.COBEM2023.COB2023-1502
Abstract
Magnetic resonance imaging has been one of the most researched and used forms of obtaining medical images in recent years due to its range of imaging possibilities. More efficient reconstructions arise every day and this work will address hardware reconstruction. The reconstruction of resonance images must be done quickly with quality and precision, this process is usually carried out by computers, however, the use of algorithms executed in hardware can prove to be efficient due to its high speed and precision in time. The objective of this work is to present the theory involving the magnetic resonance imaging process as well as explain the steps for creating the simulation that performs the reconstruction of the images. The article presents the steps for acquisition and reconstruction of resonance images through an explanation of the theory involved, and also the steps for creating the algorithm that simulates the reconfigurable system, for example an FPGA. The workflow performed is presented, the various procedures applied to the images and, finally, a comparison between the images reconstructed by the algorithm and images reconstructed in the traditional way. The performance of the algorithm is evaluated by the similarity between the image reconstructed by it and the image reconstructed in the traditional way, through the time necessary to carry out the reconstruction and also through the precision in the time of the algorithm, that is, if it manages to carry out the same tasks without variations in time. The algorithm managed to produce images with a structural similarity index greater than 80% compared to images reconstructed in a standard way and also achieved high accuracy in time, reconstructing images of equal sizes at the same speed regardless of the image. The algorithm was also able to reconstruct the images at high speed, producing over 4000 images per second. The article presents a theoretical explanation as well as a new approach in the reconstruction of magnetic resonance images that obtained promising results, which can be deepened in future studies of real-time imaging and implementation in FPGAs.
Keywords
MRI, fpga, medical imaging, VHDL, Reconstruction, Reconfigurable
