Development and validation of intraoral periapical radiography-based machine learning model for periodontal defect diagnosis

Abstract

Radiographic determination of the bone level is useful in the diagnosis and determination of the severity of the periodontal disease. Various two- and three-dimensional imaging modalities offer choices for imaging pathologic processes that affect the periodontium. In recent years, innovative computer techniques, especially artificial intelligence (AI), have begun to be used in many areas of dentistry and are helping increase treatment and diagnostic performance. This study was aimed at developing a machine-learning (ML) model and assessing the extent to which it was capable of classifying periodontal defects on 2D periapical images. Eighty-seven periapical images were examined as part of this research. The existence or absence of periodontal defects in the aforementioned images were evaluated by a human observer. The evaluations were subsequently repeated using a radiomics platform. A comparison was made of all data acquired through human observation and ML techniques by SVM analysis. According to the study findings the ability of human observers and the ML model to detect periodontal defects was significantly different in comparison to the gold standard. However, ML and human observers performed similarly for the detection of periodontal defects without a significant difference. This study reveals that the prediction of periodontal defects can be achieved by combining particular radiomic features with image variables. The proposed machine leaning model can be utilized for supporting clinical practitioners and eventually substitute evaluations conducted by human observers while enhancing future levels of performance.