An Overview of Deep Learning in Medical Imaging Focusing On MRI
DOI:
https://doi.org/10.32628/IJSRSET2512327Keywords:
Brain tumor detection, computer-aided diagnosis, convolutional neural networks, deep learning, image segmentation, medical image analysis, medical imaging, MRI, MRI reconstruction, radiomics, transfer learning, U-NetAbstract
Deep learning has impacted medical imaging quite a lot, and there is potential to advance the detection and treatment of intricate disorders. MRI is unique among imaging modalities because it is capable of detecting subtle contrasts in soft tissue without the application of ionizing radiation. To aim for better enhanced medical diagnosis, the present paper introduces a fresh analysis of MRI data and the integration of deep learning models. This study shows that advanced neural networks, e.g., Generative Adversarial Networks (GANs) for image reconstruction and U-Net for segmentation, can be employed to make diagnosis automatic, minimize the risk of human error, and facilitate early disease diagnosis of diseases like brain tumors, multiple sclerosis, and neurodegenerative diseases. The study also points to challenges regarding interpretability, lack of data, and the ethical aspects of AI-based
Downloads
References
Heckel et al. (2024) surveyed deep learning techniques for MRI reconstruction, including end-to-end networks, pre-trained models, and self-supervised methods. These techniques enhance image quality, speed up scans, and increase robustness to variability in data. https://arxiv.org/abs/2404.15692
Oscanoa et al. (2023) presented an in-depth review of deep learning-based reconstruction for cardiac MRI, focusing on state-of-the-art unrolled networks and their applications in speeding up cardiac MRI acquisitions.https://www.mdpi.com/2306-5354/10/3/334
Pal and Rathi (2022) performed an extensive review and experimental assessment of deep learning techniques for MRI reconstruction, classifying current methods into image-domain and k-space-domain methods. They reviewed different architectures, training schemes, and performance measures, as well as the absence of standardization in evaluation protocols. https://www.melba-journal.org/papers/2022%3A001.html
Ruthotto et al. (2021) underscored the promise of deep learning in multi-modal MRI analysis, such as the integration of structural and functional MRI for enhanced disease detection. Their deep neural network model utilized both modalities to more efficiently classify abnormalities such as multiple sclerosis and tumors.https://www.melba-journal.org/papers/2022%3A001.html
Reyes et al. (2020) examined explainable AI (XAI) methods applied to medical imaging, including saliency maps and layer-wise relevance propagation. They highlighted the lack of current explainability tools for clinical use, citing the disparity between what a model "attends to" and clinical logic.https://www.sciencedirect.com/science/article/pii/S1361841522001177
Bian & Tamilselvam (2024) – A Review of Optimization-Based Deep Learning Models for MRI Reconstruction
Emphasizes deep learning-based MRI reconstruction optimization strategies, analyzinghttps://www.mdpi.com/2673-9909/4/3/59
Spieker et al. (2023) – Deep Learning for Retrospective Motion Correction in MRI: A Comprehensive Review
Reviews deep learning techniques applied to motion artefact correction for MRI scans.. https://arxiv.org/abs/2305.06739
Bi (2023) – Exploring the Power of Generative Deep Learning for Image-to-Image Translation and MRI Reconstruction
Investigates generative deep learning techniques to be used for MRI reconstruction as well as generating multi-contrast MRI. https://arxiv.org/abs/2303.09012
Yang et al. (2024) – Deep Learning-based Human MRI Reconstruction and Preprocessing with Artificial Fourier Transform Network (AFT-Net)
Presents AFT-Net, which is a deep learning approach to direct processing of complex-valued raw MRI data and shows accelerated denoising and reconstruction ability.. https://archive.ismrm.org/2024/4669.html
Zhang et al. (2024) – Deep Learning-based Super-Resolution Reconstruction for Brain Diffusion-weighted MRI
Introduces a deep learning-constrained algorithm that sharpens images in brain diffusion-weighted MRI with preservation of signal-to-noise ratio, supporting clinical diagnosis.. https://archive.ismrm.org/2024/4694.html
Saksena et al. (2024) – Novel Deep Learning Approach Combining Image Reconstruction and Diagnostic Segmentation
Introduces a technique that combines reconstruction and segmentation in training, enhancing image quality and diagnostic value in accelerated MRI.. https://archive.ismrm.org/2024/2787.html
Ma, Q., Lai, Z., Wang, Z., et al. "MRI Reconstruction with Enhanced Self-Similarity Using Graph Convolutional Network https://bmcmedimaging.biomedcentral.com/articles/10.1186/s12880-024-01297-2.
Downloads
Published
Issue
Section
License
Copyright (c) 2025 International Journal of Scientific Research in Science, Engineering and Technology
This work is licensed under a Creative Commons Attribution 4.0 International License.
