MACHINE LEARNING-GUIDED PREDICTION OF OUTCOMES FOLLOWING SCTG AND L-PRF SOFT TISSUE AUGMENTATION

А.С. Ажибеков, A. MenchishevaYu, Z Yerkibayeva

2026

Closed access Metadata (CSL-JSON)
DOI 10.70113/1815-9443.2026.51.29.012
Data confidence 70 medium
How this score is calculated

It is built from checkable properties of the record itself: whether the article has a DOI, how many independent sources confirmed its data, whether the required fields are filled in, and how recent the latest confirmation is. Disagreement between sources and a missing identifier lower it. A value of 80 or above means the record was published without further checking; 60 to 79 means it is published at reduced confidence, with a manual check advisable. A record verified by an editor by hand is never pushed below 90 by a recalculation. This scores our data about the article, not the article itself.

Abstract

Peri-implant soft tissue deficiency in the posterior mandible remains a significant clinical challenge. Although subepithelial connective tissue grafts (SCTG) combined with lekocyte-platelet-rich fibrin (L-PRF) have demonstrated favorable clinical outcomes, individualised prediction of treatment success remains insufficiently investigated. Machine learning approaches may provide new opportunities for personalised treatment planning in implant dentistry. Objective. To evaluate the feasibility of machine learning based prediction of peri-implant soft tissue augmentation outcomes following SCTG and L-PRF augnentation and to identify the clinical variables most strongly associated with treatment success. Materials and methods. This study represents a secondary exploratory machine learning analysis based on data obtained from a previously conducted randomised controlled clinical trial. The final dataset included 92 patients who completed 6-month follow-up after peri-implant soft tissue augmentation procedures, including 47 patients treated with SCTG combined with L-PRF and 45 patients treated with SCTG alone. Baseline demographic, clinical immunological and peri-implant soft tissue variables were included in the analysis. Predictive modelling was performed using Python and the scikit-learn library. Random Forest and multilayer perceptron algorithms were evaluated using 5-fold cross-validation. Model performance was assessed using accuracy, precision, recall, and F1-score. Results. Favorable clinical outcomes at 6 months were observed in 79 patients (85.9%), whereas 13 patients (14.1%) demonstrated suboptimal peri-implant soft tissue or aesthetic outcomes. The Random Forest classifier demonstrated more stable predictive performance compared with the multilayer perceptron model. The exploratory Random Forest model demonstrated a mean accuracy of 0.81, precision of 0.79, recall of 0.77, and F1-score of 0.78 during cross validation procedures. Feature importance analysis identified baseline keratinised tissue thickness, peri-implant soft tissue phenotype, smoking status as the variables most strongly associated with treatment outcomes. Discussion. The findings confirm the importance of keratinized mucosa for peri-implant tissue stability and support SCTG as a predictable approach for soft tissue augmentation. The adjunctive use of L-PRF may improve postoperative healing because of its regenerative properties. Exploratory machine learning analysis demonstrated the potential feasibility of individualized outcome prediction using integrated clinical and biological variables in implant dentistry. Conclusion. Machine learning-based analysis demonstrated potential feasibility for individualised prediction of peri-implant soft tissue augmentation outcomes following SCTG and L-PRF use. Baseline peri-implant soft tissue characteristics and patient-related factors appear to significantly influence postoperative healing and esthetic outcomes. The present findings should be considered exploratory and hypnotises-generating. Further multicentre studies with larger independent datasets and external validation are required before routine clinical implementation of machine learning-based predictive models in implant dentistry.