The crown-implant ratio was measured using the ImageJ version 1.38 software measuring tool in conjunction with a magnification tool. Each implant was measured from its bottom to the crown base and then from the crown base to its highest point (Fig. 4).

All measurements were performed by a dento-maxillofacial radiologist who was blinded to patient information. To evaluate the intra-observer correlation, 20% of the images were randomly selected, and fractal analysis measurements were performed by the same radiologist 1 month later.

Statistical analyses were conducted with the R Statistical Software version 3.0.2 (Foundation for Statistical Computing, Vienna, Austria) and TURCOSA (Turcosa Analytics Ltd. Co., Turkey). Normal distribution of the data was evaluated with the Shapiro–Wilk test. Pearson’s correlation coefficient analysis was used to evaluate the inter-observer correlations and for comparison of numeric variables. Homogeneity of variances was evaluated with Levene test. Kruskal–Wallis test, Mann–Whitney U test, Student’s t test, and Welch two-sample t test were performed to compare the numerical variables between the categorical between the successful and failed implant groups. Fisher’s exact test was used to compare categoric variables between the implant survival groups. A probability level of less than 5% (p < 0.05) was accepted as statistically significant. Power analyses were conducted for the test results that did not reject null hypothesis.