For the quantification of image quality in X-ray imaging, generally exposures of test objects, test phantoms or patients can be used. Of the currently available quantification methods one can observe that methods using test objects or test phantoms can be performed both objectively and subjectively (i.e. by a panel of observers), whereas an exposure of a patient can, until now, only be assessed subjectively, with all the known limitations.
Therefore, the goal of this study was to develop useful and objective image quality measurements which only employ clinical radiographs. As this approach resembles the manner in which a radiologist assesses a radiograph in practice, we expect a better agreement with the “average” opinion of radiologists. Because the developed measurements are performed by means of a computer, they are within the domain of automatic image processing; more specifically within feature extraction. In this development, we confined ourselves to just one aspect of image quality, namely contrast.
Since in an objective assessment of a clinical radiograph there is no specific knowledge available about the subject (the patient), such an assessment is called an absolute image quality measurement, as opposed to a relative measurement where the features in an image are related to those of the test object. Thus, a relative measurement evaluates to what extent the test object is reproduced true-to-life, whereas an absolute measurement assesses only the end result.
There are some applications for which an absolute and objective image quality measurement is necessary.
For continuous image quality control in an integrated X-ray system (as in a Picture Archiving and Communication System, PACS) only clinical images are available. By establishing a continuous control, it could be possible to execute the more labor-intensive periodic controls, which use test objects, over longer periods of time.
In such a digital system, it should be possible to perform semi-automatically a so-called “reject-repeat” analysis. The rejected radiographs, which are directly available in digital format, could be automatically classified according to the cause of rejection (for example unsharpness or over- and underexposed radiographs).
An absolute image quality measurement can also be integrated into an image processing program. Based on a prior image quality measurement, it can be determined whether the analysis package is expected to successfully analyze the image. If this is not the case, one can take action at an early stage by improving or repeating the corresponding exposure. It is also possible to make the choice of a certain image processing algorithm dependent upon the quality of the image. In the case of high image quality a different procedure will then be employed than in the case of low quality.
The full PhD thesis can be downloaded as a PDF-document (4MB).
For further information, please contact:
Berend C. Stoel, PhD.
Division of Image Processing
Department of Radiology, 1-C2S
Leiden University Medical Center
P.O. Box 9600
2300 RC Leiden
Tel. +31 (0)71 526 1911
Fax. +31 (0)71 526 6801