Detection of Inferior Alveolar Nerve Canal by Feature based Machine Learning Approach

Abstract

For pre-diagnostic surgical process in dental implantology, detection of Inferior Alveolar Nerve Canal (IAC) in dental images is highly essential to avoid surgical complications and injury. In this paper, a feature based machine learning model is developed for the detection of IAC from the mandible regions of dental OPG images. Initially, the soft tissue regions are enhanced by S-CLAHE (Sharpening based Contrast Limited Adaptive Histogram Equalization). Subsequently, Shape features of the IAC using Histogram of Oriented Gradient (HOG) and texture features using Local Forward Rajan Transform (LFRT) are extracted. These features are considered as an input for Machine Learning classifier. From the trained results, the feature points of IAC region are detected by polynomial curve fitting approach. The performance of the classification technique is evaluated with existing machine learning classifiers. Adaboost M2 Ensemble classifier achieves the best accuracy of 96% compared to other state of art techniques such as Naïve Bayes, KNN, SVM, and Decision Tree. Therefore, the proposed method has high potential in IAC detection and avoids complexities in dental implant surgery.