4886
Development of a phantom for assessment of signal-to-noise ratio in whole-body diffusion-weighted MRI1MRI Unit, The Royal Marsden NHS Foundation Trust, London, United Kingdom, 2Division of Radiotherapy and Imaging, The Institute of Cancer Research, London, United Kingdom, 3Medical Physics, St George's University Hospitals NHS Foundation Trust, London, United Kingdom
Synopsis
Keywords: Phantoms, Whole Body, Whole-body diffusion-weighted imaging
Motivation: Poor signal-to-noise ratio (SNR) of whole-body diffusion-weighted images (WB-DWI) impacts the diagnostic exam quality in whole-body MRI. Evaluating SNR of WB-DWI using healthy volunteers is challenging when developing imaging protocols for multi-centre studies.
Goal(s): Develop a phantom for assessing whether a proposed WB-DWI protocol will provide adequate SNR in patient examinations.
Approach: A phantom was developed which replicated relevant MR properties of WB-MRI patients. We measured SNR using the phantom and qualitatively graded SNR in subjects.
Results: Good correlation was found between the phantom and the subject data and a discrimination threshold between good and poor quality exams was determined.
Impact: A phantom can be used to assess the SNR of WB-DWI protocols and shows good correlation with qualitative image quality, enabling faster, quantitative optimisation of SNR in WB-DWI protocols when setting up multi-centre studies.
Introduction
Diffusion-weighted imaging (DWI) plays a crucial role in whole-body MRI (WB-MRI) for oncology. WB-MRI is recommended for metastatic prostate cancer, myeloma, and general screening1–3, aiding lesion identification, staging of disease and assessment of treatment response.While guidelines specify key imaging protocol parameters, centres must adapt other parameters due to hardware and software variations, for example maximum gradient strength4.
Signal-to-noise ratio (SNR) of b=900s/mm2 images typically used in WB-DWI can significantly impact image quality, influencing radiological assessment and lesion detection, as observed in a multi-centre myeloma study5.
Volunteers are often used for qualitative SNR evaluation, but have limitations in availability, variability, and patient representation. The aim of this study was to develop a phantom for rapid, quantitative SNR assessment in b=900s/mm2 DWI to facilitate WB-DWI protocol optimisation without volunteer imaging, comparing quantitative SNR measurements to qualitative radiological scores from healthy volunteers and patients.
Methods
Images were acquired on a 1.5T scanner (MAGNETOM Sola, Siemens Healthineers, Erlangen, Germany).A phantom was constructed using three concentric Perspex cylinders reflecting abdominal size, composition, and coil loading (Figure 1). Single-station diffusion-weighted images were acquired (Figure 2). The baseline protocol was from the clinical WB-MRI protocol at our institution with common variations based on previous multi-centre WB-DWI studies4 and the QIBA DWI profile6. SNR was measured using methods 1 and 2 from the NEMA standard7 (Figure 2). Repeated measurements for protocols 1-4 were made with an identical set-up (short-term) and a reconstructed set-up the following day (long-term).
All volunteer and patient studies were approved by a national research ethics committee. All volunteers gave written consent and all patients gave verbal consent for additional research sequences as part of their clinical WB-MRI examinations.
Single station images, centred on the L5 vertebral body, were acquired (Figure 2) from 10 healthy volunteers (male/female:2/8, median age=40years, range=28-58years) and 6 patients referred for clinical WB-MRI (male/female:6/0, median age=78years, range=67-92years). One radiologist and one MR physicist (5 and 10 years' WB-MRI experience, respectively) qualitatively graded the SNR of the b=900s/mm2 images from the volunteer and patient examinations, relative to a good clinical b=900s/mm2 WB-DWI image, on a three-point Likert scale (good/borderline/poor).
Repeatability and performance of the SNR phantom measurements were assessed using Bland-Altman analysis and receiver-operating characteristics (ROC), respectively (MATLAB2023b, MathWorks, Natick, MA).
Results
Both phantom SNR measurement methods showed agreement to the expected proportionality of SNR, with good repeatability (Figure 3).The effect of the different protocol variations on qualitative SNR in patients and volunteers is illustrated in Figure 4.
Figure 5A illustrates that protocol variations producing higher phantom SNR measurements in phantoms were associated with better scores in vivo. The ROC curves in Figure 5B demonstrate that analysis A produced a perfect discriminator (phantom SNR threshold = 58.1), whereas the other analyses lead to less specific and sensitive thresholds.
Discussion and Conclusions
In WB-DWI, the b=900s/mm2 images are a key component of the examination, it is essential that they have sufficient SNR to be clinically useful.The agreement between the SNR measured in the phantom and the expected dependency on the imaging parameters demonstrates that SNR can be successfully measured using the methods tested on the phantom.
The ideal result provided by this phantom would be a threshold SNR that can be used to discriminate between protocols which produced images with good and poor SNR. Measurements of SNR made with the phantom can then be assessed against this threshold to optimise protocols. The variability of the qualitative scores for a given protocol variation demonstrates the pitfalls encountered with exclusively using volunteers in setting up WB-DWI protocols in multi-site studies. Age and gender are known to affect ADC quantification in WB-DWI studies8, which is dependent on signal intensity and thus SNR. This work demonstrates the need for a phantom SNR measurement which encapsulates the inter-subject variation and can be used to consistently and objectively assess WB-DWI SNR. A limitation of the phantom is that it does not account for motion experienced by living subjects, however choosing an SNR measurement method that is less susceptible to motion will help to reduce this limitation when making quantitative measurements in vivo.
Further work will acquire additional data from volunteer and patient examinations for scoring and quantitative measurements of SNR in vivo, to better define a discrimination threshold and will evaluate inter-scanner reproducibility.
In conclusion, a phantom has been developed for quantitative assessment of SNR in WB-DWI that correlates with qualitative assessment of image quality, to enable rapid, quantitative assessment of SNR to support the optimisation of protocols in multi-centre studies and to provide on-going quality assurance of existing protocols.
Acknowledgements
This study represents independent research funded by the National Institute for Health and Care Research (NIHR) Biomedical Research Centre at The Royal Marsden NHS Foundation Trust and The Institute of Cancer Research, London, and by the Royal Marsden Cancer Charity. The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care.References
1. Padhani AR, Lecouvet FE, Tunariu N, et al. METastasis Reporting and Data System for Prostate Cancer: Practical Guidelines for Acquisition, Interpretation, and Reporting of Whole-body Magnetic Resonance Imaging-based Evaluations of Multiorgan Involvement in Advanced Prostate Cancer. Eur Urol. 2017;71(1):81. doi:10.1016/J.EURURO.2016.05.033
2. Messiou C, Hillengass J, Delorme S, et al. Guidelines for acquisition, interpretation, and reporting of whole-body MRI in myeloma: Myeloma response assessment and diagnosis system (MY-RADS). Radiology. 2019;291(1):5-13. doi:10.1148/RADIOL.2019181949
3. Petralia G, Koh DM, Attariwala R, et al. Oncologically relevant findings reporting and data system (ONCO-RADS): Guidelines for the acquisition, interpretation, and reporting of whole-body MRI for cancer screening. Radiology. 2021;299(3):494-507. doi:10.1148/RADIOL.2021201740/ASSET/IMAGES/LARGE/RADIOL.2021201740.FIG3B.JPEG
4. Rata M, Blackledge M, Scurr E, et al. Implementation of Whole-Body MRI (MY-RADS) within the OPTIMUM/MUKnine multi-centre clinical trial for patients with myeloma. Insights Imaging. 2022;13(1). doi:10.1186/S13244-022-01253-0
5. Keaveney S, Dragan A, Rata M, et al. Image quality in whole-body MRI using the MY-RADS protocol in a prospective multi-centre multiple myeloma study. Insights into Imaging 2023 14:1. 2023;14(1):1-14. doi:10.1186/S13244-023-01498-3
6. QIBA Profile: Magnetic Resonance Diffusion-Weighted Imaging (DWI) of the Apparent Diffusion Coefficient (ADC).; 2022. https://qibawiki.rsna.org/images/b/b0/QIBA_DWIProfile_Stage3_15Dec2022_v3.pdf. Accessed October 23, 2023.
7. NEMA Standards Publication MS 1-2008 (R2014, R2020) Determination of Signal-to-Noise Ratio (SNR) in Diagnostic Magnetic Resonance Imaging.; 2021. www.nema.org.
8. Lavdas I, Rockall AG, Castelli F, et al. Apparent diffusion coefficient of normal abdominal organs and bone marrow from whole-body DWI at 1.5 T: The effect of sex and age. American Journal of Roentgenology. 2015;205(2):242-250. doi:10.2214/AJR.14.13964/SUPPL_FILE/08_14_13964_SUPPDATA_S10_CMYK.EPS
9. Winfield JM, Douglas NHM, Desouza NM, Collins DJ. Phantom for assessment of fat suppression in large field-of-view diffusion-weighted magnetic resonance imaging. Phys Med Biol. 2014;59(9):2235. doi:10.1088/0031-9155/59/9/2235
10. NEMA Standards Publication MS 9-2008 (R2014, R2020) Characterization of Phased Array Coils for Diagnostic Magnetic Resonance Images.; 2021. www.nema.org.
Figures
Figure 2 - A summary of the key sequence parameters of the baseline diffusion sequence adapted from the clinical protocol, a list of the variations made to the baseline protocol, and the methodology and analysis pipeline for collecting quantitative measures of SNR in the phantom and qualitative SNR in volunteers and patients. NSA = number of signal averages (per b-value), FoV= field-of-view.10
Figure 3 – A: SNR measured in the phantom, relative to the baseline protocol SNR, for each set of protocol variations using the “no RF” and subtraction methods of SNR estimation. The relationships between the varied parameters and SNR have been linearised, according to the expected relationship, to ease interpretation of goodness of fit for each SNR method. Error bars, which are present on all the values, show the standard error on the mean pixel value. B: Short- and long-term coefficient of variation (CoV) repeatability results for protocol variations 1-4.
Figure 4 – Examples of b=900s/mm2 images from the protocol variations employed in patients and volunteers with the corresponding final qualitative SNR quality score, age, gender, and BMI. The windowing per image is automatically set by the PACS system. (Bright, vertical streaks seen in some of the images are from a known interference artefact.)
Figure 5 – A: Percentage split of poor/borderline/good scores for each protocol variation overlaid with phantom SNR measured with the “no RF” method. B: ROC curves for the three analyses performed, using quantitative results from the “no RF” method. AUC = area under the ROC curve.