This project investigates interpretation tasks in which the diagnostic performance is limited by he ability of the human observer to recognize or understand the displayed information. Of interest is the development of specific image processing techniques enhancing the visual data presentation or performing automated quantitation of images. Current work has centered on methods relevant to digital subtraction radiography that could be integrated into a computerized dental x-ray system. Once specific goal was the automated recognition and delineation of areas in radiographs showing trabecular bone. Corresponding results indicate that fractal geometry provides a realistic mathematical model to recognize and characterize trabecular bone. Furthermore, the fractal dimension of the bone surface derived from the model identification process is a global image descriptor that may be developed into a measure enabling the quantitation of bone structure changes due to osteoporosis. Noise reduction in subtraction images is a processing step that precedes lesion detection and measurement. However, this digital filtering step necessarily also incurs undesirable image blurring and thus, a filter is desirable that achieves substantial noise smoothing without causing undue image blurring. A filter, modeled after neural nets in the visual cortex that display directional sensitivity, was designed that recognizes contrast edges and performs smoothing only in directions that do not cross edges. This novel filter structure was shown to preserves image sharpness substantially better than known linear or median filters when matched to attain identical noise suppression. Particular emphasis has been placed on clinical applications of the image subtraction method. One study investigated its use for following-up patients in which bony lesions were filled with atresia-lapatite implants. The method successfully demonstrated bone regeneration or loss of implant material over 4 to 6 months postoperatively. In another clinical study the efficacy of a new anti-inflammatory drug in retarding alveolar bone loss in periodontal patients is investigated by monitoring bone mass changes over a 12 months period and determining any correlations with the presence of specific pathogens sampled from gingival pockets.
