Influence of binarization methods on the fractal dimension of alveolar bone using digital radiographs

• Published in: Oral surgery, oral medicine, oral pathology and oral radiology oral medicine, oral pathology and oral radiology, 2023-11, Vol.136 (5), p.649-655
• Main Authors: Oliveira-Santos, Nicolly, Silva, Aline Gabrieli, Gaêta-Araujo, Hugo, Oliveira's, Matheus L., Groppo, Francisco Carlos
• Format: Article
• Language: English
• Subjects: Cancellous Bone; Fractals; Humans; Mandible; Radiography; Radiography, Dental, Digital - methods
• ISSN: 2212-4403; 2212-4411
• DOI: 10.1016/j.oooo.2023.06.008

The aim of this study was to evaluate the influence of binarization methods (BnMs) on the fractal dimension (FD) of the alveolar bone in digital radiographic images. Five identical periapical radiographs were obtained from each of 20 anatomic regions. Using ImageJ/Fiji software, a standardized region of interest was positioned in the trabecular bone of all radiographs, 14 BnMs were independently applied, and FD was calculated. The Friedman test and Dunn post hoc tests were used to evaluate the influence of BnMs on FD (α = 0.05). The intraclass correlation coefficient (ICC) was used to assess correlation among BnMs. No significant differences were observed in the threshold value between the BnMs. No significant differences in FD were observed among 9 BnMs (Default, Huang, Iso Data, Li, Mean, Min Error[I], Moments, Otsu, and Shanbhag). These methods presented significantly higher FDs than those from 3 other BnMs (Max Entropy, Renyi Entropy, and Yen), which did not significantly differ from each other. The Triangle method significantly differed from all BnMs, and the Percentile method produced significantly higher FDs than Huang, IsoData, Li, Moments, Otsu, and Shanbhag. Most BnMs presented good or excellent correlation (ICC ≥ 0.76). Binarization methods influence FD of the alveolar bone from digital periapical radiographs; thus, this technical aspect is relevant for comparison purposes and should be standardized. For comparison of published FD research, the Default method is recommended.