Document Type : Original Research
Authors
1 Department of Biomedical Engineering, University of Eyvanekey, Eyvanekey, Iran
2 School of Electrical and Computer Engineering, University of Tehran, Tehran, Iran
3 Cognitive Systems Laboratory, Control and Intelligent Processing Center of Excellence (CIPCE), School of Electrical and Computer Engineering, University of Tehran, Tehran, Iran
Abstract
Background: Magnetic Resonance Spectroscopy (MRS) enables noninvasive assessment of brain metabolites, but is often limited by low Signal to Noise Ratio (SNR) and structured artifacts, such as residual water peaks. Conventional denoising methods, including Hankel Singular Value Decomposition (HSVD) and Empirical Mode Decomposition (EMD), address specific noise types but have complementary limitations.
Objective: This study aimed to develop and evaluate a hybrid HSVD and Complex Empirical Mode Decomposition (CEMD) framework for in vivo proton MRS of the Alzheimer’s brain.
Material and Methods: This retrospective observational in vivo methodological study analyzed single-voxel proton MRS (¹H-MRS) data acquired from 20 Alzheimer’s patients using a 3T Magnetic Resonance Imaging (MRI) scanner. HSVD was first applied to suppress structured residuals, followed by CEMD to remove stochastic high-frequency noise. Performance was assessed using spectral SNR, residual water amplitude, and metabolite linewidths, and compared with conventional averaging, HSVD-only, and CEMD-only approaches.
Results: The hybrid HSVD+CEMD approach achieved the highest SNR (3.24±1.03), reduced linewidths (18.83±4.11 Hz), and preserved key metabolite peaks, outperforming individual methods.
Conclusion: Combining model-based and data-driven denoising provides robust noise reduction while maintaining spectral fidelity, enhancing reliable metabolite quantification in clinical MRS applications, including Alzheimer’s studies.
Keywords
)