Quality assurance in oral radiography: computer-based methods for assessment and controlled improvement of intraoral...

The reported poor diagnostic quality of 30-50% of all radiographs made in dental practices suggests the need for standardized quality measures and appropriate image improvement procedures. We propose a new method to evaluate the quality of digital/digitized radiographs using feature vectors. This method assures the separation of over- and under-exposed radiographs from well exposed ones as clusters of "good image quality" and "poor image quality" in the resulting feature space. Automatic image improvement is accomplished by adaptive, piece-wise linear grayvalue transformation with regard to the image quality being measured. In particular, our method automatically detects grayvalue intervals which are likely to represent relevant diagnostic information as well as grayvalue intervals which are irrelevant for diagnostic purposes. Based on those intervals, a piece-wise linear grayvalue mapping function is calculated. These procedures may be used both on whole images and for the interactive improvement of suspected pathological regions of interest. Actual image quality is indicated by the simultaneous display of the defined cluster of "good image quality" and the position of the image within the feature space for visual control.