Accuracy of periapical radiography with improvement filters in the detection of bone graft loss around dental implants

Objectives

To evaluate the accuracy of radiographic filters to assess bone graft loss adjacent to dental implants.

Methods

Twenty-one dental implants were inserted into dry mandible alveoli. The interface between the implant and the alveolar cortex was filled with particulate synthetic bone graft. Three levels of bone graft loss (GL) were simulated: exposure of the implant-implant cover interface, exposure of the first implant thread, and exposure of the second implant thread. Periapical radiographs were acquired with VistaScan system, and all images were saved without an image filter (original) and with six filters applied: smoothing, emboss, inversion, Endo, Perio, and pseudocolorization. Three oral radiologists evaluated the bone graft loss using a five-point scale. The values ​​of area under the ROC curve (AUC), sensitivity, and specificity were calculated. One-way analysis of variance compared the diagnostic values for different levels of GL between the different filters considering a significance level of 5 %.

Results

There was a significant difference in AUC (p < 0.001), sensitivity (p < 0.001), and specificity (p < 0.001) between the filters. There was no difference between smoothing (p = 0.390), Endo (p = 0.404), and Perio (p = 0.178) compared to the original images. Emboss and pseudocolorization generally present less satisfactory values.

Conclusions

The application of smoothing, Endo, and Perio filters in periapical radiographs does not compromise diagnostic accuracy for GL. However, the original images still demonstrate the best performance.

Clinical significance

Radiography is essential to evaluate bone graft around dental implants and, while original images present superior performance, most of enhancement filters do not impair the diagnostic accuracy. Clinicians may be able to select filters at their discretion for this diagnostic task.