Synopsis

Keywords: Neuro: Brain, Transferable skills: Metrology of MRI, Transferable skills: Statistics

Introduction

Challenges and Unmet Needs

• Measurement Errors and Variation: For any qMRI method, critical considerations include the accuracy and precision over the range of anticipated measurement values. Multiple sources of measurement errors and noise can lead to biased estimates and/or increased variance. Common types of measurement error and noise include the inherent electronic receiver noise of the MRI system, physiologic noise, scanner stability, patient motion, image artifacts, and operator errors. These effects diminish the interpretability and confidence of qMRI measurements. Further, qMRI measurements that deviate from linear dependence on the true values create greater challenges with interpretation.

• Systematic qMRI Error Factors: Magnetic field properties have significant effects on qMRI measurements. Obviously, scanner magnetic field strength (B0) can have a range of effects on qMRI including changes in SNR, NMR relaxation rates, and artifacts. Spatial inhomogeneity in B0 will lead to EPI distortion, frequency shifts in spectral components (fat/water, MR spectroscopy metabolites), and susceptibility artifact signal attenuation. Errors in the RF field amplitude (B1) will affect RF pulse flip angles, inversion and refocusing pulses, which can lead to modeling errors for relaxometry and magnetization transfer measures. Uncorrected nonlinear gradient fields (G) will lead to spatial distortion and errors in the prescribed diffusion-weighting. These errors are compounded when considering data from multiple sites, scanners and protocols.
• Standardization: A significant challenge in the field is a lack of standardized acquisition protocols and analysis pipelines, which can lead to variability in qMRI measurements across studies and sites. Even within a single study, scanner upgrades can lead to changes in qMRI error and variation. Establishment of standardized protocols, phantoms, and quality control measures is essential. Methods for harmonizing qMRI data across sites and protocols are an active area of research and may help to mitigate site-specific differences but create challenges for expansion to broad generic applications.
• Individual Variation: On top of sources of measurement variability and errors described above are inter-subject variabilities in brain anatomy and physiology. A major objective of many PHPM applications is detecting individual differences of qMRI measures across the distributions associated with healthy and patient populations. The detectable effect size for qMRI measures is ultimately limited by sources of variance associated with the measurements, inter-site variation, and population variation.
• Biological Specificity, Interpretation and Validation: For many qMRI measures, including diffusion, relaxometry, and magnetization transfer imaging, the actual specificity or relationship to specific biological or pathological properties is indirect or unclear. This is despite the widespread labeling of qMRI-based measures with names suggestive of biologic features like myelin water fraction, neurite density, etc. Further, since qMRI methods are noninvasive, it is challenging to evaluate the biological/pathological specificity to qMRI measurements. Unfortunately, there is limited validation against histopathology and other reference standards. Large-scale multi-center studies validating quantitative MRI biomarkers against gold-standard reference standards are necessary to establish their clinical utility and reliability.
• Data Analysis Complexity: In many cases, qMRI processing and analyses are complex and require advanced specialized pipelines, which are barriers for routine use in PHPM. Development of user-friendly, standardized software tools for automated data processing and analysis would facilitate the integration of quantitative MRI into routine clinical workflows.
• Cost and Accessibility: High costs associated with MRI scanners and specialized software, coupled with limited access in rural and underserved areas, restrict the widespread adoption of qMRI.

Future Directions

• Advanced Imaging Technologies: The continued development of novel imaging systems, sequences and contrast mechanisms will reduce qMRI measurement errors and variance, and will expand the range of qMRI biomarkers including more biologically specific measures. Particularly exciting developments include the recent introduction of dedicated ultrahigh performance neuroimaging systems, qMRI methods that are faster and more robust to motion, and more advanced qMRI models with greater biological specificity.
• Open-Source MRI Sequences and Protocols: The recent development of open-source MRI pulse sequence and protocol tools will enable multi-site qMRI studies using identical pulse sequences and scanner settings. This will help to reduce sources of inter-site variation.
• Big Data Analytics: Large-scale multi-modal neuroimaging datasets and advanced analytics will be extremely valuable for generating normative distributions, identifying novel biomarkers, disease subtypes, and therapeutic targets for PHPM.
• Machine Learning and Artificial Intelligence: Leveraging machine learning and artificial intelligence algorithms for automated image curation, analysis, disease classification, and predictive modeling based on quantitative MRI data.
• Point-of-Care Imaging: Development of lower cost and portable MRI systems for point-of-care imaging would enable rapid and cost-effective assessment of neurological conditions in diverse clinical settings.

Acknowledgements

References