Radiographic Evidence of Immature Bone Architecture After Sinus Grafting: A Multidimensional Image Analysis Approach

Abstract

Background: Radiographic evaluation of bone regeneration following maxillary sinus floor elevation commonly emphasizes volumetric gains. However, the qualitative microarchitecture of the regenerated bone, particularly when assessed via two-dimensional imaging modalities, such as panoramic radiographs, remains insufficiently explored. This study aimed to evaluate early trabecular changes in grafted maxillary sinus regions using fractal dimension, first-order statistics, and gray-level co-occurrence matrix analysis. Methods: This retrospective study included 150 patients who underwent maxillary sinus floor augmentation with bovine-derived xenohybrid grafts. Postoperative panoramic radiographs were analyzed at 6 months to assess early healing. Four standardized regions of interest representing grafted sinus floors and adjacent tuberosity regions were analyzed. Image processing and quantitative analyses were performed to extract fractal dimension (FD), first-order statistics (FOS), and gray-level co-occurrence matrix (GLCM) features (contrast, homogeneity, energy, correlation). Results: A total of 150 grafted sites and 150 control tuberosity sites were analyzed. Fractal dimension (FD) and contrast values were significantly lower in grafted areas than in native tuberosity bone (p < 0.001 for both), suggesting reduced trabecular complexity and less distinct transitions. In contrast, higher homogeneity (p < 0.001) and mean gray-level intensity values (p < 0.001) were observed in the grafted regions, reflecting a more uniform but immature trabecular pattern during the early healing phase. Energy and correlation values also differed significantly between groups (p < 0.001). No postoperative complications were reported, and resorbable collagen membranes appeared to support graft stability. Conclusions: Although the grafted sites demonstrated radiographic volume stability, their trabecular architecture remained immature at 6 months, implying that volumetric measurements alone may be insufficient to assess biological bone maturation. These results support the utility of advanced textural and fractal analysis in routine imaging to optimize clinical decision-making regarding implant placement timing in grafted sinuses.

Keywords: bone regeneration; fractal dimension; maxillary sinus floor augmentation; panoramic radiography.

Figure 1

Step-by-step intraoperative stages of the lateral window sinus lifting procedure: (A) following crestal and buccal flap elevation, a lateral bony window was outlined and carefully removed to expose the Schneiderian membrane; (B) a xenograft material was placed into the newly created subantral space to augment the sinus floor; (C) placement of a resorbable collagen membrane over the osteotomy site; (D) the flap was repositioned and secured with nonresorbable sutures to ensure proper soft tissue closure.

Figure 2

The yellow square represents a manually selected 25 × 25 pixel Region of Interest (ROI) from the grafted sinus floor region. The ‘L’ represents the left side. These standardized ROIs were used to extract FD values, enabling a comparative assessment of trabecular bone complexity between grafted and physiologic sites.

Figure 3

Stepwise image processing workflow used for FD calculation: (a) duplicated image from original radiograph; (b) application of Gaussian Blur; (c) subtracted version; (d) added 128 value; (e) binarized image; (f) eroded image; (g) dilated image; (h) inverted image; (i) skeletonized image; and D; double-logarithmic box-count plot used to derive the FD value.

Figure 4

Directional GLCM framework applied to contrast-enhanced panoramic ROIs. The left panel shows a CLAHE-enhanced region from the grafted maxillary sinus area (central pixel intensity = 104), while the right panel represents a native tuberosity region (central pixel intensity = 160). In both panels, the pixel of interest is highlighted in red, and directional relationships (0°, 45°, 90°, and 135°) are marked with blue arrows. This configuration was used to calculate angle-specific GLCM features by assessing gray-level co-occurrence between the pixel of interest and its directional neighbors.

Figure 5

Pearson correlation matrix demonstrating pairwise associations among FD, FOS (mean gray level, standard deviation), and GLCM-based texture features (contrast and homogeneity) obtained from postoperative ROIs. A weak negative correlation was found between FD and contrast (r = −0.31), suggesting that decreased trabecular complexity may be associated with increased local gray-level variability. Other parameter pairs showed negligible correlations (|r| < 0.2), indicating minimal linear association between structural and intensity-based metrics in the early postoperative phase.