The performance of a new computer-aided image analysis method of detecting and measuring approximal radiolucencies depicted on bitewing radiographs was evaluated. Image magnification and intensity calculations were found to be reproducible and the magnification calibration routines were sufficiently accurate. Randomly selected clinical radiographs and radiographs of extracted teeth were used; serial sectioning of the extracted teeth allowed histological validation of the results for 24 surfaces. Although lesion detection and the measurement of radiolucency depth were reproducible, initial estimates of area were less satisfactory and some radiolucencies visible to the naked eye could not be detected. Modifications to the system software were made and tested. The problems were largely due to the presence of radiopaque islets within radiolucencies and were overcome by the introduction of a ‘peak-merging routine’. This modified software, however, proved to be too sensitive and required adjustments to the threshold of detection which would appear to have remedied this difficulty. Further work employing the developed software with larger numbers of radiographs is justified. The performance of this system to date is promising; particularly when compared to the variability reported for repeated visual diagnoses and estimates of radiolucency s
展开▼
