Infrared Spectroscopy for Determining Bone Graft Healing and Readiness in Tissue Models

Successful dental implant placement relies on the proper integration of bone graft material into an area of deficient native bone. Traditional radiographic methods provide only qualitative assessments, making it challenging to accurately determine graft maturation and the appropriate time for successful implant placement. This study investigates the potential of visible near-infrared (VNIR) spectroscopy as a noninvasive, quantitative method for assessing compositional changes that potentially correlate with intraoral bone graft healing in novel tissue models.

Six tissue models representing progressive stages of bone graft (BG) healing were developed using Straumann Xenograft BG, Jason Collagen Membrane, porcine mandibular bone, blood, and gingiva. Spectra were collected using an ASD LabSpec4 spectrometer with an infrared fiber-optic probe. Preprocessing techniques were utilized to enhance spectral resolution and enable analysis based on absorbances associated with hemoglobin, protein, lipids, and water. Principal component analysis (PCA) was performed to assess compositional variation across healing stages. Machine learning was applied to PCA scores for classification of stages using linear discriminant analysis (LDA) followed by k-nearest neighbors (KNN).

VNIR spectroscopy successfully distinguished bone graft from native bone and captured spectral changes associated with models of progressive graft maturation. Notably, C-H-related absorbances increased from early to late model healing stages, with advanced spectral models more closely resembling healthy bone. Classification using LDA discriminant functions as features for the KNN algorithm achieved a 97% accuracy for prediction of the healing stage.

VNIR spectroscopy shows strong potential as a noninvasive, quantitative tool for monitoring intraoral bone graft healing. By identifying compositional markers such as C-H absorbances related to lipid or protein in intact bone, this approach may help determine optimal implant timing. Coupled with machine learning, this technique can also offer a user-friendly approach for clinicians. Continued pre-clinical validation with histological studies and animal models is needed to support clinical trials with the eventual aim of enhancing treatment planning and patient outcomes.