MRI Image Enhancement: A Scoping Review of Traditional Techniques and Deep Learning Approaches

Magnetic Resonance Imaging (MRI) enhancement techniques, including traditional methods and deep learning approaches, improve image quality by reducing artefacts, noise, and enhancing resolution, yet no single method universally excels across scenarios. This PRISMA-guided scoping review screened 1337 articles and 67 patents from ScienceDirect, CINAHL, PubMed, and Lens.org, including 36 articles, 7 patents, and 2 grey literature sources on MRI-specific enhancement (artefact reduction, noise minimization, resolution improvement). Studies without empirical validation or deep learning details were excluded following title/abstract and full-text screening. Traditional techniques like NLM filtering and motion correction effectively reduce noise and artefacts but falter in dynamic imaging. Deep learning accelerates acquisition, boosts signal-to-noise ratios, and enhances diffusion-weighted imaging, though challenges persist in generalizability and transparency. Hybrid approaches combining both, show promise for overcoming limitations. The review maps strengths and gaps in MRI enhancement, recommending hybrid traditional deep learning methods and deeper model interpretability to advance clinical diagnostics.