1
Concepts: Measurement in MRI1
Paul S. Tofts
Brighton and Sussex Medical School
In this chapter, concepts of quantitative magnetic resonance imaging (qMRI) are introduced. The traditions of measurement science and MRI meet in qMRI. MRI has undergone a paradigm shift, as a sophisticated device for visualisation within the brain also offers the option for high-quality measurement, giving access to biological and physiological information, often at the microstructural level. qMRI can measure the effects of disease and the response to treatment. In clinical trials, it is often used as an objective and early indicator of treatment efficacy. The long tradition of scientific measurement is followed to the invention of MRI. The ‘longitude problem’ in the 1700s posed a similar scientific challenge to that of qMRI – how to develop a technology that would have a dramatic impact on human well-being. Principles of good study design are laid out; attention to measurement error (the instrumental standard deviation) can enormously reduce the size and cost of a study. A table of recommended MRI books is given in this chapter.
Contents
1.1 Introduction
Measurement science and MRI come together • Limits to progress • Using this book
1.2 History of measurement
Early measurement • The longitude problem: John Harrison • Scientific societies • Units of measurement • Mathematical physics • Scientific medicine • Early qMRI
1.3 Measurement in medical imaging
Images, partial volume and maps • Study design • Usefulness of an MR parameter
1.4 The Future of quantitative MRI
Technology and methodology • International standardisation and certification
References
1.1 Introduction
1.1.1. Measurement science and MRI come together
Measurement science has been around a long time; MRI2 has been around for about 35 years. This book is about the blending of the two paradigms.
We have come to expect to be able to measure certain quantities with great accuracy, precision and convenience. Instruments for mass, length and time are all conveniently available, and we expect the results to be reproducible when measured again and also to be comparable with measurements made by others in other locations. In the human body we expect to measure some parameters (height, weight, blood pressure) ourselves, recognising that some of these parameters may have genuine biological variation with time. More invasive measurements (e.g. blood alcohol level or blood sugar level) are also expected to have a well-defined normal range and to be reproducible. In physics, chemistry, electrical engineering and the manufacturing industry, there is a strong tradition of measurement, international agreements on standards and training courses for laboratory practitioners. International standards of length, mass and time have been in existence for many years. Secondary standards are produced, which can be traced back to the primary standards. National and international bodies provide coordination.
As individual scientists we may have a passionate desire to use our talents for the benefit of mankind, preferring to devote our energy to finding better ways of helping our fellow humans to be healthy than to improving weapons for their destruction. In this context, developing measurement techniques in MRI constitutes a perfect application of traditional scientific skills to a modern problem.
MRI is now widespread and accepted as the imaging method of choice for the brain (and for many body studies). It is generally used in a qualitative way, with the images being reported non-numerically by radiologists. Many MRI machines now have independent workstations, connected to the scanner and the database of MR images, which enable and encourage simple quantitative analysis of the images in their numerical (i.e. digital) form. However, the data collection procedure often prevents proper quantification being carried out; variation in machine parameters such as transmitter output gain, flip angle value (and its spatial variation), receiver gain, and image scaling may all be acceptable for qualitative analysis but cause irreversible confusion in images to be quantified. Researchers may be unaware of good practice in quantification and collect or analyse data in an unsuitable way, even though the MRI machine is capable of more.
The process of quantifying or measuring parameters in the brain necessarily takes more time and effort than a straightforward qualitative study. More MRI scanner time is needed, and considerable physics development effort and computing resources may be needed to set up the procedure. In addition, analysis can be very time-consuming, and support of the procedure is required to measure and maintain its reliability over time. Processes have to be found that are insensitive to operator procedure (whether in the data collection or image analysis) and to scanner imperfections (such as radiofrequency non-uniformity from a particular head coil), that provide good coverage of the brain in a reasonable time and that are stable over study times, which may extend to decades.
The benefits of quantification are that fundamental research into biological changes in disease, and their response to potential treatments, can proceed in a more satisfactory way. Problems of bias, reproducibility and interpretation are substantially reduced. MRI can move from a process of picture-taking, where reports are made on the basis of unusually bright, dark, small or large objects, to a process of measurement, in the tradition of scientific instrumentation, where a whole range of quantities can be tested to see whether they lie in a normal range and whether they have changed from the time of a previous examination.
In this book, the intention is to demonstrate the merging of these two traditions or paradigms, that is measurement and MRI, to form the field of quantitative MRI (qMRI).3 The MRI measurement process is analysed, often in great detail. Limits to accuracy and precision are identified, as far as possible, with the intention of identifying methods that are reliable and yet practical in a clinical MRI scanning environment. The biological meaning of the many MR parameters that are available is explored, and clinical examples are given where MR parameters are altered in disease. Often these changes have been observed qualitatively, and they serve to encourage us to improve the measurement techniques, in order that more subtle effects of disease can be seen, earlier than currently possible, and in tissue that is currently thought to be normal as judged by conventional MRI. The ideal is to obtain push-button (turnkey) techniques for each of the many MRI parameters in this book, such that an MRI radiographer (technologist) can measure each of these parameters reliably and reproducibly, with a minimum of human training or intervention, in the same way that we can currently step onto a weighing machine and obtain a digital readout of our mass. In the case of qMRI, the output would be considerably richer, perhaps showing images of abnormal areas (computed from large databases of normal image data sets), changes from a previous MRI exam, possible interpretations (diagnoses) and an indication of certainty for each piece of information. The advances in the pre-scan and the spectroscopy MR procedures, which used to be very time-consuming and operator-dependent and are now available as fully automated options, show how this might be possible.
Thus, MRI has been undergoing a paradigm shift4 in how it is viewed and used. In the past it was used for forming qualitative images (the ‘happy-snappy MRI camera’, taking pictures); in the future it may be increasingly used as a scientific instrument to make measurements of clinically relevant quantities. The dichotomy can be seen in the MRI literature; radiological descriptions often speak of signal hyperintensity in a sequence with a particular weighting, whilst studies using physical measurements often report localised concentration values, normal ranges, age and gender effects, and reproducibility. As measurement becomes more precise and analysis enables clinically relevant information to be extracted from a myriad of information, it will become possible, in principle, to make measurements on an individual patient to characterise the state of their tissue, guiding the choice of treatment and measuring its effect. The issues involved in bringing qMRI into the radiological clinic were well summarised in an editorial in the American Journal of Neuroradiology (McGowan, 2001; Box 1.1).
As part of this ongoing paradigm shift, our view of what MRI can tell us is changing. When it started, information was largely anatomical (anatomical MRI), in the sense that relat...