In many areas of life, measurements are made to describe the world. Any method that estimates the physical units of a property is a measurement tool. For example, cooking scales are used to ensure the correct quantities of ingredients in a recipe.
The performance of any measurement tool can be assessed in terms of accuracy and precision, to estimate the theoretical limits of the measurement and inform the interpretation of subsequent measurements. Quantitative Magnetic Resonance (qMR) methods are measurement tools, which means they too have accuracy and precision that can be characterised.
National Physical Laboratory (NPL) (Teddington, UK) provides general guidance on measurement science, including accuracy and precision12. The NPL also provides other resources3 for measurement science which should be applied to qMRI to ensure good practice.
Accuracy
Accuracy describes how close the measured value is to the ‘true’ value, sometimes called the ground truth or the reference value. The reference value may be determined using a trusted gold standard method under strict laboratory conditions. High accuracy means the estimated value is very similar to the ground truth. If many measurements are taken, accuracy is estimated by comparing the measurement mean with the true value. For example, the mean measurement of MR parameter T2 could be taken from a region of selected image voxels in a phantom vial and compared to the value provided by the phantom manufacturer.
Precision
Precision refers to the measurement spread of repeated measurements. In qMR, this can be defined as the standard deviation across an area of voxels (i.e. a ‘Region of Interest’ (ROI)).
Which is more useful?
A method can have high precision but low accuracy. In other words, the measured values could be very similar each time the measurement is performed but the measured values may be very different to the true value.
In quantitative medical imaging, high precision is often prioritised over accuracy as it suggests high measurement consistency. High precision means the method is sensitive to small parameter changes, thereby allowing subtle biological changes to be detected. In this case, any inaccuracy can be accounted for by applying a correction factor to adjust the measurements.