The impact of Platelet Rich Plasma (PRP) in osseointegration of oral implants in dental panoramic radiography: texture based evaluation

Abstract

Purpose: In this study the temporal texture differentiation associated with the bone formation properties, around loaded oral implants after Platelet Rich Plasma (PRP) employment, was investigated in Panoramic Radiographs.

Materials and methods: Thirty eligible patients are randomly assigned to two groups. The test group received PRP application around new implants, while in the control group no PRP treatment was made. The bone-to-implant contact region was analyzed in a clinical sample of 60 Digitized Panoramic Radiographs, 30 corresponding to immediate implant loading (Class-I) and 30 after an 8 month follow-up period (Class-II). This region was sampled by 1146 circular Regions-of-Interest (ROIs), resulting from a specifically designed segmentation scheme based on Markov-Random-Fields (MRF). From each ROI, 41 textural features were extracted, then reduced to a subset of 4 features due to redundancy and employed as input to Receiver-Operating-Characteristic (ROC) analysis, to assess the textural differentiation between two classes.

Results: The selected subset, achieved Area-Under-Curve (AUC) values ranging from 0.77-0.81 in the PRP group, indicating the significant temporal textural differentiation has been made. In the control group, the AUC values ranged from 0.56-0.68 demonstrating lesser osseo integration activity.

Conclusion: This study provides evidences that PRP application may favor bone formation around loaded dental implants that could modify the dental treatment planning.

Keywords: Markov Random Fields; Platelet Rich Plasma; dental implants; implant segmentation; medical image texture analysis; radiography.

Figure 1

Oral implant of a dental panoramic radiograph. Zoomed image part at arrow locations indicate the bone-to-implant regions in which the texture analysis is applied. The differentiations in radiologic appearance patterns (increased gray tone illumination and variability) of the bone-to-implant area are been considered towards efficient segmentation.

Figure 2 a, b, c, d, e, f.

(a) Random image initialization, (b) MRF clusters, (c) implant segmentation (black outline), (d) areas of bone to implant contact, (e) fitted circular sampling ROIs (black color), and (f) radiologists’ selected ROI (gray color).

Figure 3 a, b, c, d

Box plots from the selected subset for both groups, (a) Angular Second Moment (ASM), (b) Correlation (COR), (c) Gray Level Non Uniformity (GLNU), and (d) Long Run Emphasis (LRE).