Implementation and Comparative Analysis of Magnetic Resonance Image (MRI) Reconstruction Using Generative Adversarial Networks (GANs)

Amauri da Costa Júnior; Gerardo Antonio Idrobo Pizo; Cristiano Jacques Miosso

doi:10.3991/ijoe.v22i09.61101

Authors

Amauri da Costa Júnior Universidade de Brasília, Brasília, Brasil
Gerardo Antonio Idrobo Pizo Universidade de Brasília, Brasília, Brasil https://orcid.org/0000-0002-6165-672X
Cristiano Jacques Miosso Universidade de Brasília, Brasília, Brasil

DOI:

https://doi.org/10.3991/ijoe.v22i09.61101

Keywords:

MRI, GAN, reconstruction, Undersampled, U-Net

Abstract

Magnetic resonance imaging (MRI) scans tend to be long procedures, making it difficult for patients to remain motionless. Generative adversarial networks (GANs) are artificial intelligence models widely used in image processing. This paper addresses the use of GANs to aid in the reconstruction of magnetic resonance images made with undersampled single-coil data, aiming to reduce MRI scan time. The GAN improves quality metrics of images reconstructed with different trajectories (vertical Cartesian, random Cartesian, radial, spiral 1, and spiral 4), which present distortions due to not meeting the Nyquist–Shannon sampling theorem. Tests are performed by feeding the trained model with images made from undersampled data to generate enhanced images. The generator receives these images and attempts to produce enhanced versions based on typical MRI scans, while the discriminator compares generated and typical images to determine which are real or fake. At each epoch, both models refine themselves based on the discriminator’s verdict, improving image quality. Metrics such as the structural similarity index and signal-to-noise ratio are extracted from the produced images. In the best tests, increases of 4× in average SSIM and up to 14× in average SNR were achieved.

Author Biographies

Amauri da Costa Júnior, Universidade de Brasília, Brasília, Brasil

Engenheiro Eletricista formado pela Universidade de Brasília (UnB). Seus interesses de pesquisa incluem reconstrução de imagens de ressonância magnética e a aplicação de redes generativas adversárias (GANs) em imagens biomédicas.

Gerardo Antonio Idrobo Pizo, Universidade de Brasília, Brasília, Brasil

Associate Professor at the University of Brasília (UnB), Faculty of Gama (FGA). His research focuses on biomedical engineering, artificial intelligence applied to medical imaging, and signal processing.

Cristiano Jacques Miosso, Universidade de Brasília, Brasília, Brasil

Associate Professor at the University of Brasília (UnB), Faculty of Gama (FGA). His research interests include signal processing, medical image reconstruction, machine learning, and stochastic signal processing.

References

[1] OECD. “Health at a Glance 2023”. 2023. Available: https://www.oecd.org/en/publications/health-at-a-glance-2023_7a7afb35-en.html [Accessed: Sep. 10, 2025].

[2] S. N. Krieger, L. Huber, B. A. Poser, R. Turner, and G. F. Egan. “Simultaneous Acquisition of Cerebral Blood Volume-, Blood Flow-, and Blood Oxygenation-Weighted MRI Signals at Ultra-High Magnetic Field”. Magnetic Resonance in Medicine, 74(2):513–517, 2015.

[3] K. H. Mild, R. Lundström, and J. Wil´en. “Non-Ionizing Radiation in Swedish Health Care—Exposure and Safety Aspects”. International journal of environmental research and public health, 16(7):1186, 2019.

[4] D. Bos, N. Guberina, S. Zensen, M. Opitz, M. Forsting, and A. Wetter. “Radiation Exposure in Computed Tomography”. Deutsches Ärzteblatt International, 120(9):135, 2023.

[5] M. Zaitsev, J. Maclaren, and M. Herbst. “Motion Artifacts in MRI: A Complex Problem With Many Partial Solutions”. Journal of Magnetic Resonance Imaging, 42(4):887–901, 2015.

[6] A. A. Mazzola. “Magnetic Resonance: Principles of Image Formation and Applications in Functional Imaging”. Revista Brasileira de Física Médica, 3(1):117–129, 2009.

[7] A. D. Elster. Methods for filling k-space. Available: https://mriquestions.com/ k-space-trajectories.html. [Accessed: Sep. 15, 2025].

[8] J.P. Felmlee, R.L. Morin, J.R. Salutz, and G.B. Lund. “Magnetic Resonance Imaging Phase Encoding: A Pictorial Essay”. Radiographics, 9(4):717–722, 1989.

[9] H. de Ponteves, and K. Eremenko. “Machine Learning A-Z: AI, Python and R”. Available: https://www.udemy.com/course/machinelearning/. [Accessed: Sep. 11, 2025].

[10] J. Montalt-Tordera, V. Muthurangu, A. Hauptmann, and J. A. Steeden. “Machine Learning in Magnetic Resonance Imaging: Image Reconstruction”. Physica Medica, 83:79–87, 2021.

[11] S. U. Dar, M. Yurt, M. Shahdloo, M. E. Ildız, B. Tınaz, and T. Çukur. “Prior-Guided Image Reconstruction for Accelerated Multicontrast MRI Via Generative Adversarial Networks”. IEEE Journal of Selected Topics in Signal Processing, 14(6):1072–1087, 2020.

[12] NYU Langone Health. “Welcome to the FastMRI dataset”. Available: https://fastmri.med.nyu.edu/. [Accessed: Sep. 15, 2025].

[13] J. Zbontar, F. Knoll, A. Sriram, T. Murrell, Z. Huang, M. J. Muckley, A. Defazio, R. Stern, P. Johnson, M. Bruno, et al. “FastMRI: An open dataset and benchmarks for accelerated MRI”. arXiv preprint arXiv:1811.08839, 2018.

[14] O. Ronneberger, P. Fischer, and T. Brox. “U-net: Convolutional Networks for Biomedical Image Segmentation. In Medical Image Computing and Computer-assisted Intervention–MICCAI 2015: 18th International Conference, Munich, Germany, October 5-9, 2015, proceedings, part III 18, pages 234–241. Springer, 2015.

[15] A. N. Avanaki. “Exact Global Histogram Specification Optimized for Structural Similarity”. Optical Review, 16:613–621, 2009.

[16] A. D. Elster. “MR Quality Control: SNR. How Do You Measure Signal-to-noise in an Image?”. Available: https://mriquestions.com/signal-to-noise.html. [Accessed: Sep. 17, 2025].

[17] TensorFlow. “tf.keras.layers.Dropout”.

Available: https://www.tensorflow.org/api_docs/python/tf/keras/layers/Dropout.

[Accessed: Sep. 25, 2025].

[18] TensorFlow. “tf.keras.Regularizer”.

Available: https://www.tensorflow.org/api_docs/python/tf/keras/Regularizer.

[Accessed: Sep. 25, 2025].