Prediction of oral implant prognosis through machine learning methods

Abstract

Purpose: Despite the high success rate of oral implants, usually described as osseointegration, the practitioner occasionally encounters difficulties caused by trade-offs between various factors during surgery. The purpose of this study was to discover the most significant factors predicting implant success using non-traditional statistical analysis. Materials and Methods: The study was based on a systematic search of chart files at Seoul National University Bundang Hospital from June 2004 to October 2005. Oral and maxillofacial surgeons inserted 667 implants in the mouths of 198 patients after consultation with a prosthodontist. The implants were judged as favorably or unfavorably placed and the associated outcome was assessed 1 year after treatment by a prosthodontist from a biomechanical point of view. We processed descriptive evaluations for several features in binary form for the analysis. Unfortunately, some cases were excluded due to lack of information during this process. In this study, we used the machine learning method of a decision tree model and a support vector machine for analysis. Results: We identified mesio-distal position of the fixture as the most significant factor determining the prognosis of the implant. Both of the machine learning methods yielded this result. Discussion: The strength of the machine learning method is that it can be applied to a small sample size. To verify the conclusion of this study, traditional statistical tools could be applied with large samples