Denoising magnetic resonance spectroscopy (MRS) data using stacked autoencoder for improving signal-to-noise ratio and speed of MRS.
Magnetic resonance spectroscopy (MRS)
denoising
metabolite quantification
signal to noise ratio (SNR)
sparse representation
stack auto-encoder (SAE)
Journal
Medical physics
ISSN: 2473-4209
Titre abrégé: Med Phys
Pays: United States
ID NLM: 0425746
Informations de publication
Date de publication:
Dec 2023
Dec 2023
Historique:
revised:
05
10
2023
received:
30
03
2023
accepted:
25
10
2023
medline:
6
12
2023
pubmed:
10
11
2023
entrez:
10
11
2023
Statut:
ppublish
Résumé
While magnetic resonance imaging (MRI) provides high resolution anatomical images with sharp soft tissue contrast, magnetic resonance spectroscopy (MRS) enables non-invasive detection and measurement of biochemicals and metabolites. However, MRS has low signal-to-noise ratio (SNR) when concentrations of metabolites are in the range of millimolar. Standard approach of using a high number of signal averaging (NSA) to achieve sufficient SNR comes at the cost of a long acquisition time. We propose to use deep-learning approaches to denoise MRS data without increasing NSA. This method has potential to reduce the acquisition time as well as improve SNR and quality of spectra, which could enhance the diagnostic value and broaden the clinical applications of MRS. The study was conducted using data collected from the brain spectroscopy phantom and human subjects. We utilized a stack auto-encoder (SAE) network to train deep learning models for denoising low NSA data (NSA = 1, 2, 4, 8, and 16) randomly truncated from high SNR data collected with high NSA (NSA = 192), which were also used to obtain the ground truth. We applied both self-supervised and fully-supervised training approaches and compared their performance of denoising low NSA data based on improvement in SNR. To prevent overfitting, the SAE network was trained in a patch-based manner. We then tested the denoising methods on noise-containing data collected from the phantom and human subjects, including data from brain tumor patients. We evaluated their performance by comparing the SNR levels and mean squared errors (MSEs) calculated for the whole spectra against high SNR "ground truth", as well as the value of chemical shift of N-acetyl-aspartate (NAA) before and after denoising. With the SAE model, the SNR of low NSA data (NSA = 1) obtained from the phantom increased by 28.5% and the MSE decreased by 42.9%. For low NSA data of the human parietal and temporal lobes, the SNR increased by 32.9% and the MSE decreased by 63.1%. In all cases, the chemical shift of NAA in the denoised spectra closely matched with the high SNR spectra without significant distortion to the spectra after denoising. Furthermore, the denoising performance of the SAE model was more effective in denoising spectra with higher noise levels. The reported SAE denoising method is a model-free approach to enhance the SNR of MRS data collected with low NSA. With the denoising capability, it is possible to acquire MRS data with a few NSA, shortening the scan time while maintaining adequate spectroscopic information for detecting and quantifying the metabolites of interest. This approach has the potential to improve the efficiency and effectiveness of clinical MRS data acquisition by reducing the scan time and increasing the quality of spectroscopic data.
Sections du résumé
BACKGROUND
BACKGROUND
While magnetic resonance imaging (MRI) provides high resolution anatomical images with sharp soft tissue contrast, magnetic resonance spectroscopy (MRS) enables non-invasive detection and measurement of biochemicals and metabolites. However, MRS has low signal-to-noise ratio (SNR) when concentrations of metabolites are in the range of millimolar. Standard approach of using a high number of signal averaging (NSA) to achieve sufficient SNR comes at the cost of a long acquisition time.
PURPOSE
OBJECTIVE
We propose to use deep-learning approaches to denoise MRS data without increasing NSA. This method has potential to reduce the acquisition time as well as improve SNR and quality of spectra, which could enhance the diagnostic value and broaden the clinical applications of MRS.
METHODS
METHODS
The study was conducted using data collected from the brain spectroscopy phantom and human subjects. We utilized a stack auto-encoder (SAE) network to train deep learning models for denoising low NSA data (NSA = 1, 2, 4, 8, and 16) randomly truncated from high SNR data collected with high NSA (NSA = 192), which were also used to obtain the ground truth. We applied both self-supervised and fully-supervised training approaches and compared their performance of denoising low NSA data based on improvement in SNR. To prevent overfitting, the SAE network was trained in a patch-based manner. We then tested the denoising methods on noise-containing data collected from the phantom and human subjects, including data from brain tumor patients. We evaluated their performance by comparing the SNR levels and mean squared errors (MSEs) calculated for the whole spectra against high SNR "ground truth", as well as the value of chemical shift of N-acetyl-aspartate (NAA) before and after denoising.
RESULTS
RESULTS
With the SAE model, the SNR of low NSA data (NSA = 1) obtained from the phantom increased by 28.5% and the MSE decreased by 42.9%. For low NSA data of the human parietal and temporal lobes, the SNR increased by 32.9% and the MSE decreased by 63.1%. In all cases, the chemical shift of NAA in the denoised spectra closely matched with the high SNR spectra without significant distortion to the spectra after denoising. Furthermore, the denoising performance of the SAE model was more effective in denoising spectra with higher noise levels.
CONCLUSIONS
CONCLUSIONS
The reported SAE denoising method is a model-free approach to enhance the SNR of MRS data collected with low NSA. With the denoising capability, it is possible to acquire MRS data with a few NSA, shortening the scan time while maintaining adequate spectroscopic information for detecting and quantifying the metabolites of interest. This approach has the potential to improve the efficiency and effectiveness of clinical MRS data acquisition by reducing the scan time and increasing the quality of spectroscopic data.
Types de publication
Journal Article
Langues
eng
Sous-ensembles de citation
IM
Pagination
7955-7966Subventions
Organisme : NIBIB NIH HHS
ID : R01 EB032680
Pays : United States
Organisme : NIBIB NIH HHS
ID : R56 EB033332
Pays : United States
Organisme : NIH HHS
Pays : United States
Organisme : NIH HHS
Pays : United States
Commentaires et corrections
Type : UpdateOf
Informations de copyright
© 2023 American Association of Physicists in Medicine.
Références
Dienel GA. Brain glucose metabolism: integration of energetics with function. Physiol Rev. 2018;99(1):949-1045.
van Ewijk PA, Schrauwen-Hinderling VB, Bekkers SC, Glatz JF, Wildberger JE, Kooi ME. MRS: a noninvasive window into cardiac metabolism. NMR Biomed. 2015;28(7):747-766.
Valkovič L, Chmelík M, Krššák M. In-vivo 31P-MRS of skeletal muscle and liver: a way for non-invasive assessment of their metabolism. Anal Biochem. 2017;529:193-215.
Wang T, Zhu XH, Li H, et al. Noninvasive assessment of myocardial energy metabolism and dynamics using in vivo deuterium MRS imaging. Magn Reson Med. 2021;86(6):2899-2909.
Lingvay I, Esser V, Legendre JL, et al. Noninvasive quantification of pancreatic fat in humans. J Clin Endocrinol Metab. 2009;94(10):4070-4076.
Hu HH, Kim HW, Nayak KS, Goran MI. Comparison of fat-water MRI and single-voxel MRS in the assessment of hepatic and pancreatic fat fractions in humans. Obesity. 2010;18(4):841-847.
Utriainen M, Komu M, Vuorinen V, et al. Evaluation of brain tumor metabolism with [11C] choline PET and 1H-MRS. J Neurooncol. 2003;62(3):329-338.
Horská A, Barker PB. Imaging of brain tumors: mR spectroscopy and metabolic imaging. Neuroimaging Clin. 2010;20(3):293-310.
Sharma U, Jagannathan NR. Metabolism of prostate cancer by magnetic resonance spectroscopy (MRS). Biophys Rev. 2020;12(5):1163-1173.
Trigui R, Mitéran J, Walker PM, Sellami L, Hamida AB. Automatic classification and localization of prostate cancer using multi-parametric MRI/MRS. Biomed Signal Process Control. 2017;31:189-198.
Haddadin IS, McIntosh A, Meisamy S, et al. Metabolite quantification and high-field MRS in breast cancer. NMR Biomed. 2009;22(1):65-76.
Stanwell P, Gluch L, Clark D, et al. Specificity of choline metabolites for in vivo diagnosis of breast cancer using 1H MRS at 1.5 T. Eur Radiol. 2005;15(5):1037-1043.
Mueller S, Trabesinger A, Boesiger P, Wieser H. Brain glutathione levels in patients with epilepsy measured by in vivo 1H-MRS. Neurology. 2001;57(8):1422-1427.
Simister RJ, McLean MA, Barker GJ, Duncan JS. Proton MRS reveals frontal lobe metabolite abnormalities in idiopathic generalized epilepsy. Neurology. 2003;61(7):897-902.
Hall EL, Stephenson MC, Price D, Morris PG. Methodology for improved detection of low concentration metabolites in MRS: optimised combination of signals from multi-element coil arrays. Neuroimage. 2014;86:35-42.
Löffler R, Sauter R, Kolem H, Haase A, von Kienlin M. Localized spectroscopy from anatomically matched compartments: improved sensitivity and localization for cardiac31P MRS in humans. J Magn Reson. 1998;134(2):287-299.
Snaar J, Teeuwisse W, Versluis M, et al. Improvements in high-field localized MRS of the medial temporal lobe in humans using new deformable high-dielectric materials. NMR Biomed. 2011;24(7):873-879.
Considine EC. The search for clinically useful biomarkers of complex disease: a data analysis perspective. Metabolites. 2019;9(7):126.
Lam F, Li Y, Guo R, Clifford B, Liang ZP. Ultrafast magnetic resonance spectroscopic imaging using SPICE with learned subspaces. Magn Reson Med. 2020;83(2):377-390.
Lam F, Ma C, Clifford B, Johnson CL, Liang ZP. High-resolution 1H-MRSI of the brain using SPICE: data acquisition and image reconstruction. Magn Reson Med. 2016;76(4):1059-1070.
Chang SG, Yu B, Vetterli M. Adaptive wavelet thresholding for image denoising and compression. IEEE Trans Image Process. 2000;9(9):1532-1546.
Abramovich F, Sapatinas T, Silverman BW. Wavelet thresholding via a Bayesian approach. J R Stat Soc Series B Stat Methodol. 1998;60(4):725-749.
Donoho DL, Johnstone JM. Ideal spatial adaptation by wavelet shrinkage. Biometrika. 1994;81(3):425-455.
Donoho DL, Johnstone IM. Adapting to unknown smoothness via wavelet shrinkage. J Am Statist Assoc. 1995;90(432):1200-1224.
Kaur P, Singh G, Kaur P. A review of denoising medical images using machine learning approaches. Curr Med Imaging Rev. 2018;14(5):675-685.
Ji B, Hosseini Z, Wang L, Zhou L, Tu X, Mao H. Spectral wavelet-feature analysis and classification assisted denoising for enhancing magnetic resonance spectroscopy. NMR Biomed. 2021;34(6):e4497.
Vincent P, Larochelle H, Lajoie I, Bengio Y, Manzagol P-A, Bottou L. Stacked denoising autoencoders: learning useful representations in a deep network with a local denoising criterion. J Mach Learn Res. 2010;11(12):3371-3408.
Vincent P, Larochelle H, Bengio Y, Manzagol P-A, Extracting and composing robust features with denoising autoencoders. In: 25th International Conference on Machine Learning (ICML); 2008:1096-1103.
Nurmaini S, Darmawahyuni A, Sakti Mukti AN, Rachmatullah MN, Firdaus F, Tutuko B. Deep learning-based stacked denoising and autoencoder for ECG heartbeat classification. Electronics. 2020;9(1):135.
Xiong P, Wang H, Liu M, Lin F, Hou Z, Liu X. A stacked contractive denoising auto-encoder for ECG signal denoising. Physiol Meas. 2016;37(12):2214.
Xiong P, Wang H, Liu M, Liu X. Denoising autoencoder for eletrocardiogram signal enhancement. J Med Imaging & Health Infor. 2015;5(8):1804-1810.
Liu X, Wang H, Li Z, Qin L. Deep learning in ECG diagnosis: a review. Knowl Based Syst. 2021;227:107187.
Lei Y, Ji B, Liu T, Curran WJ, Mao H, Yang X. Deep learning-based denoising for magnetic resonance spectroscopy signals. In: Medical Imaging 2021: Biomedical Applications in Molecular, Structural, and Functional Imaging. Vol. 11600. SPIE; 2021:16-21.
Provencher S, LCModel & LCMgui User's Manual. LCModel Version 62-4. In. 2012.
Hu H, Bai J, Xia G, Zhang W, Ma Y. Improved baseline correction method based on polynomial fitting for Raman spectroscopy. Photonic Sens. 2018;8(4):332-340.
Xi Y, Rocke DM. Baseline correction for NMR spectroscopic metabolomics data analysis. BMC Bioinf. 2008;9(1):324.
Stanford TE, Bagley CJ, Solomon PJ. Informed baseline subtraction of proteomic mass spectrometry data aided by a novel sliding window algorithm. Proteome Sci. 2016;14(1):19.
Shamaei A, Starcukova J, Pavlova I, Starcuk Jr Z. Model-informed unsupervised deep learning approaches to frequency and phase correction of MRS signals. Magn Reson Med. 2022;89(3):1221-1236.
Jirayucharoensak S, Pan-Ngum S, Israsena P. EEG-based emotion recognition using deep learning network with principal component based covariate shift adaptation. Sci World J. 2014;2014:627892.
Sohn J, Nie W, Avery SM, et al. Wavelet-based protoacoustic signal denoising for proton range verification. In: Medical Imaging 2020: Ultrasonic Imaging and Tomography. Vol. 11319. SPIE; 2020:14-19.
Yao S, Hu Z, Zhang X, et al. Feasibility study of range verification based on proton-induced acoustic signals and recurrent neural network. Phys Med Biol 2020;65(21):215017.
Freijo C, Herraiz JL, Sanchez-Parcerisa D, Udias JM. Dictionary-based protoacoustic dose map imaging for proton range verification. Photoacoustics. 2021;21:100240.
Mafi M, Martin H, Cabrerizo M, Andrian J, Barreto A, Adjouadi M. A comprehensive survey on impulse and Gaussian denoising filters for digital images. Signal Process. 2019;157:236-260.
Lee G, Gommers R, Waselewski F, Wohlfahrt K, O'Leary A. PyWavelets: a Python package for wavelet analysis. J Open Source Software. 2019;4(36):1237.
Van Rossum G, Drake F. Python 3 Reference Manual. CreateSpace; 2009.
Virtanen P, Gommers R, Oliphant TE, et al. SciPy 1.0: fundamental algorithms for scientific computing in Python. Nat Methods. 2020;17(3):261-272.
Lin F, Chen K, Wang X, Cao H, Chen D, Chen F. Denoising stacked autoencoders for transient electromagnetic signal denoising. Nonlin Processes Geophys. 2019;26(1):13-23.
Huang W, Sun F-C. Building feature space of extreme learning machine with sparse denoising stacked-autoencoder. Neurocomputing. 2016;174:60-71.
Thirukovalluru R, Dixit S, Sevakula RK, Verma NK, Salour A. Generating feature sets for fault diagnosis using denoising stacked auto-encoder. In: 2016 IEEE International Conference on Prognostics and Health Management (ICPHM). IEEE; 2016:1-7.
Bender A, Auer DP, Merl T, et al. Creatine supplementation lowers brain glutamate levels in Huntington's disease. J Neurol. 2005;252:36-41.
Near J, Edden R, Evans CJ, Paquin R, Harris A, Jezzard P. Frequency and phase drift correction of magnetic resonance spectroscopy data by spectral registration in the time domain. Magn Reson Med. 2015;73(1):44-50.