Go Back

Assessment of image quality with phantoms


Phantoms are essential for evaluation of clinical medical imaging systems in order to simulate, as closely as possible, the actual conditions under which these systems function.  Unfortunately, there are many different phantoms, often a phantom for each modality—some modalities have multiple phantoms. 

It is necessary to understand what type of phantom to select for a specific evaluation.  In addition, the medical physicist should be able to carry out either an ROC analysis comparing two different image types, e.g., digital radiography versus screen-film radiography, or low dose versus high dose CT.  A less complicated approach is a features analysis which provides excellent results, assuming the appropriate statistics have been taken into account in designing the study.

Important Points

Understanding image quality comparisons is relatively straightforward if one is comparing two similar modalities, e.g., digital and screen-film radiography, or image noise at different doses in CT.  However, the comparison becomes quite complex when one is trying to compare different modalities, e.g., CT and MRI, as there is little similarity in the image quality metrics and the images provide different information.

Every medical physicist should be able to evaluate resolution and modulation transfer function, image noise (including noise power spectrum), contrast, low contrast such as with mesh patterns, and contrast detail curves.  As noted above, ROC analysis and features analysis are also needed in clinical medical imaging.  Although these may not be used every day in the clinical setting, the medical physicist must understand how to carry out such measurements and, most importantly, the physical principles behind the measurements.

Introduction to References

Metz has been a prolific researcher and author regarding receiver operating characteristic analysis.  The paper by Dorfman, Berbaum, and Metz should be the standard reference for medical physicists on this topic. Metz also provides software for ROC analysis on his website (see Supplemental References).

Schueler et al., provide a good example of the application of features analysis to different screen-film mammography systems.